Edit Article

How to Understand Your Website Traffic Variation with Time

Eight Parts:Identifying Ways to Use the DataUnderstanding Strategies and Pitfalls when Exploring the DataUnderstanding the Different Kinds of MeasurementUnderstanding the Weekly Traffic CycleUnderstanding the Annual Traffic CycleUnderstanding the Daily Traffic CycleUnderstanding the Traffic Pattern Following Push FactorsUnderstanding the Long-Term Trends for Your SiteCommunity Q&A

This guide explores how to understand the way that your website's traffic varies over time, and how to use this information to better understand your website's users. The goal is to separate periodic patterns, one-off anomalies, random noise, push and pull factors, and long-term trends. Even if you are interested in only one of these (such as long-term trends) it is still important to get a rudimentary understanding of the others, so that you can filter them out to focus on what you care about.

Most of our data and examples are in English and most relevant to the United States, but some principles generalize. We cover websites in a wide range of domains, but familiarity with the subject matter is not necessary to understand the guide. We use Google Analytics (GA), Quantcast Measure (QM), Google Trends,[1], Wikipedia pageviews, and other sources for data.

Part 1
Identifying Ways to Use the Data

  1. 1
    Identify timing choices related to outreach efforts that may be influenced by your analysis of traffic variation.
    • For instance, you might use the information to help figure out when to send a newsletter: the time of day, day of week, or day of month.
    • You might use the information to better time your social media posts for maximum impact. For instance, do you want to post to Facebook, Twitter, and Pinterest in the morning, during the day, or in the evening? The reason it's important to understand the situation for your own website better is that there is no robust global heuristic. For instance, a blog post by CoSchedule considers sixteen different sources of advice on the best times of post, each suggesting different strategies and some of them even contradicting each other.[2]
    • You might use the information to buy advertisements, as well as boost existing social media posts, to show at the appropriate times: the time of day, day of week, or day of month.
  2. 2
    Identify choices that you may make based on what you infer about users from traffic patterns.
    • For instance, if your site sees a lot more traffic during the academic year than the summer, that is an indication that it is primarily used by students in connection with their coursework or academic life. This might be interesting information to you. It might be a relationship that you did not expect but that makes more retrospective sense. For instance, content on how to improve memory might be something you'd expect year-round traffic to, but you might discover that its main target audience is students during the academic year.
    • Similarly, if your site sees a lot more traffic near the end of the calendar quarters, that shows some relation to businesses or sales teams reaching quarterly goals. This might be interesting information that you did not expect but that makes retrospective sense. For instance, a website on how to motivate yourself to reach your goals might see this sort of spike.
  3. 3
    Give some weight to the validation you receive by seeing user behavior.
    • Creating and maintaining a website that attracts enough users to see clear patterns is an impressive accomplishment. You can use the trends you see to confirm that you have had an impact and are meeting real human needs.
    • Looking at these patterns can also help you satisfy your curiosity about your visitors, many of whom you may never meet.

Part 2
Understanding Strategies and Pitfalls when Exploring the Data

  1. 1
    Understand the concepts of time range (date range) and granularity when graphing.
    • You will generally study traffic by making graphs of metrics (such as sessions, pageviews, and users). The time range defines the starting and ending point in time for the period for which you are plotting the graph. This may also be called a date range. In GA, the Date Range can be selected on the top right.
    • You can also control the granularity of the display. Granularity is the length of the time interval covered by each individual data point. A daily granularity means that you will see one data point for each day, representing the overall value for the day. A weekly granularity means that you will see one data point per week, representing the overall value for the week. Smaller granularities are called finer and larger granularities are called coarser. In GA, you can choose granularity within the chart area itself, with options "Hourly", "Day", "Week", and "Month". The "Hourly" option may not be available in all parts of GA.
  2. 2
    Choose time range and granularity based on the cycles and trends you want to focus on.
    • A longer time range allows you to see trends and cycles over a longer period of time, whereas a shorter time range allows you to zoom in on a specific time period more clearly (and gives you more screen space for finer granularity).
    • A finer granularity allows you to see trends and cycles at smaller scales. For instance, if you want to understand how traffic varies within a day, you have to plot at a granularity finer than a day.
    • A coarser granularity allows you to see trends and cycles at a larger scale while reducing the effect of noise as well as smaller cycles. For instance, plotting at a weekly granularity helps you get rid of weekly cycles and focus on how traffic is changing between weeks.
    • A general rule for plotting to be able to capture cycles: plot at a granularity that gets rid of smaller cycles than the one you are trying to discover. Also, plot over a time range that covers at least three periods for which you are trying to determine a cycle. For instance, if you want to understand the annual cycle, plot at a weekly granularity (to get rid of the weekly cycle) and plot over at least three years. Similarly, to understand weekly cycles, plot at a daily granularity (to get rid of the daily cycle) over at least three weeks.
  3. 3
    Do not choose too fine a granularity for the level of traffic that you have.
    • Random fluctuation as a proportion of total traffic goes down as total traffic goes up.
    • Therefore, the more traffic you have, the finer the granularity at which you can do analysis. An approximate rule-of-thumb: you should not go to a finer granularity than one where you have 100 pageviews (or about 50 visits). This leads to the following approximate thresholds of minimum traffic you need for various granularities. Note that these are just approximate guidelines; your website might have much more stable traffic, allowing you to use finer granularities even at lower traffic levels.
      • Hourly granularity: At 1,000 or more daily pageviews (or about 30,000 monthly pageviews), this becomes reasonably robust during peak traffic hours. At 4,000 or more daily pageviews (or about 120,000 monthly pageviews), this becomes reasonably robust at all hours. Note that since weekday and weekend traffic can differ, hourly granularity may be fine on some days of the week but noisy on others.
      • Daily granularity: At 100 or more daily pageviews (or about 3,000 monthly pageviews), this is reasonably robust.
      • Weekly granularity: At 15 or more daily pageviews (or about 450 monthly pageviews), this is reasonably robust.
      • Monthly granularity: At over 100 monthly pageviews, this is reasonably robust.
    • If traffic to your website is heavily influenced by viral news events (in the world) or the virality of specific posts you make to it, then analysis at short timescales may be hard. In other words, the rule of thumb above might break down. A site driven by viral news events or viral social media posts can have fairly unstable day-to-day and hour-to-hour traffic patterns despite having a fairly high level of traffic.
  4. 4
    Keep in mind the role of location and the timezone aggregation problem. This becomes particularly relevant if your traffic spans multiple timezones.
    • The main problem is as follows. Analytics providers are reporting traffic from across a bunch of timezones. In principle, there are two ways of doing this: use a single fixed timezone, or use local time. Most analytics tools, including GA, use the single fixed timezone approach.
      • The single fixed timezone approach: When reporting traffic for the calendar day June 1, report total worldwide traffic during that calendar day in the specific timezone. Therefore, this counts traffic over a contiguous 24-hour period only.
      • The local time approach: When reporting traffic for the calendar day June 1, sum up, across timezones, the total traffic during the calendar day June 1 using local time in that timezone. Therefore, this counts traffic over what is potentially a 48-hour period, though in any given region it counts traffic only over a 24-hour period.
    • GA displays all results in a single fixed timezone. The timezone is part of the settings for your Google Analytics property. This can be changed one-time but cannot simply be changed as part of the display setting. You can find and change the timezone used in your View Settings. Time zone changes are applied only going forward and not retroactively.[3]
    • The timezone Quantcast uses for QM is Central Time (as observed in Mexico City). This timezone is chosen to best approximate the local times of the majority of end users of websites that use QM.
    • Google Trends also uses a single fixed timezone, albeit one that varies with the observer. It always displays data in the local time of the person using the service. For instance, if you are in California but you filter Google Trends to show data for New York, it will still show results using California's local time (Pacific Time). Therefore, you will need to explicitly adjust for the timezone difference when interpreting Google Trends data.
    • The single fixed timezone approach makes some types of analysis easier. In particular, understanding the traffic effects of specific news events or social media posts is easier with this approach. On the other hand, understanding the daily cycle and people's consumption habits becomes trickier. A few key points:
      • The single fixed timezone is most challenging when trying to understand the daily cycle. It can also complicate the analysis of the weekly cycle. It is generally not an issue for understanding annual cycles.
      • One way of getting around the problem is to filter to specific geographic locations. However, traffic level in individual regions might be too low for meaningful analysis. This is particularly the case for traffic at hourly granularity, which is where filtering by region is needed most.
  5. 5
    Be cautious about confusing between periodic variation, random noise, and long-term trends. Some examples of ambiguous traffic changes are below.
    • It can be hard to know how much of an increase in traffic from December to January is a result of a long-run growth trend and how much is recovery from a Christmas dip.
    • Similarly, for a site that's growing a lot every day, it is hard to know how much of an increase from Sunday to Monday is a result of a weekday-weekend differential and how much is due to the overall growth trend.
  6. 6
    Filter traffic based on loyalty, age, gender, and location to more clearly understand your traffic patterns.
    • Filtering by loyalty (e.g., new versus returning visitors, or people who have viewed at least a certain number of pages) can help you determine if traffic patterns are driven by your regular users or by "drive-by traffic". In general, drive-by traffic tends to account for a larger share of variation.
    • If your hypothesis about the reasons for variation in traffic center around traffic of a particular type, then filter for that kind of traffic. Note that the traffic segments responsible for most of the variation may be different from the traffic segments responsible for most of the traffic. For instance, a website that gets a lot of traffic from both students and professionals might see an annual traffic cycle primarily due to the academic year of students, although the majority of its users are professionals. An example (discussed later) is programming Q&A website Stack Overflow.[4]
    • In cases where you cannot directly filter for the attributes you care about, use proxies. For instance, to verify that the variation in your traffic is driven by the academic year, filter to large university towns as your locations. You can filter for different kinds of university towns to better understand the underlying sources of variation.

Part 3
Understanding the Different Kinds of Measurement

  1. 1
    Use traffic measurement including users, sessions, pageviews, and unique pageviews.[5]
    • Pageviews: This measures the number of times people open a page on the website.
    • Unique pageviews: This measures the number of times a visitor first visits a page. In particular, refreshes or revisits of the same page are not counted.
    • Sessions (also known as visits): This measures the number of times people visit a site. Every visitor must have at least one visit. The way an analytics tool separates sessions/visits for a single visitor from each other can vary based on the analytics tool. GA resets a session after 30 minutes of inactivity.[6][7]
    • Users (also known as visitors): This measures the number of distinct people who visited your website, as best as can be judged by the analytics service. Note that there is some ambiguity regarding the granularity at which users are deduplicated. For instance, if a user visits twice with a year of separation between the visits, is that user counted as a single user? Therefore, when reporting users, a granularity within which users are deduplicated is specified. For instance, if users are reported in terms of daily uniques (DU), then a user who visits multiple times in the same day is counted as a single user, but visits across days get added up. Similarly, weekly uniques (WU) and monthly uniques (MU) are also used. It's important to remember that weekly uniques would be less than the sum of daily uniques through the week.
    • Since calculating uniques across an arbitrary period is a computationally intensive process, the general practice with reporting uniques is to do so only for fixed time periods (such as daily, weekly, and monthly). For instance, QM reports uniques only for 1-day, 7-day, and 30-day periods.[8]
    • An analytics service may be unable to identify the same user across browsers or devices, or if the user clears cookies. For this reason, unique user counts as estimated by analytics services are generally overestimates.[9]
  2. 2
    Use derived metrics to assess traffic quality.
    • Pageviews/user (also known as pages per visitor) and pageviews/session (also known as pages per visit) are the two common metrics. These numbers can range from 1 to about 20, with most websites seeing pageviews/session numbers between 1 and 4 and pageviews/user numbers between 1 and 8. As a general rule, higher values of pageviews/user or pageviews/session are considered "better" but there are many exceptions. For instance, for a transaction-based site, getting things done more quickly might mean a better user experience, so smaller values of pageviews/session might be better. Similarly, for a site that a user should ideally use once and then not need again, smaller values of sessions/user or pageviews/user may be better.
    • New versus returning users is also used. Note that it is not completely obvious whether a higher percentage of users who are new is a good or a bad thing. Rather, looking at the absolute numbers and trends for both is more helpful than simply looking at the percentage.
  3. 3
    Keep in mind the following correlational rules of thumb.
    • Pageviews, sessions, and users should generally go up and down in tandem. In other words, days when there are more pageviews will also have more sessions and users.
    • However, the metrics will not necessarily change proportionally. In other words, the pageviews/session and sessions/user metrics could fluctuate over time. In many cases, days with higher traffic see lower values of pageviews/session and sessions/user, because the extra traffic is more shallow and less loyal. In particular, for sites that have lower traffic on weekends than weekdays, pageviews/session is usually higher on weekends than weekdays. However, there are exceptions: in case of a viral and rapidly developing news event, pageviews/user can go up (as people keep poking around the site to find new coverage of the developing event). Also, on big holidays (as opposed to simply weekends), pageviews/session might go down along with overall pageviews if very little new content is being released.
  4. 4
    Include conversion metrics if they matter to you.
    • If you are selling stuff directly on your site, then sales you make through the site (purchases and subscriptions) are your conversion events. This is typically the case for Business-to-Consumer (B2C) websites which include e-commerce as well as online subscription services. It may also apply to low-end Business-to-Business (B2B) product websites.
    • For more high-end product websites, including expensive B2B products as well as in-person services (like counseling services), the final sale does not usually occur on the website, but the website is used as a starting point for showcasing and explaining the product. In this case, "conversion" on the website is usually defined in terms of initiating contact through a contact form or call now button.
    • Keep in mind that trends for conversion can differ from trends for pageviews. In cases where there is a weekly or annual cycle, sessions and pageviews start rising before conversions start rising.

Part 4
Understanding the Weekly Traffic Cycle

  1. 1
    Keep in mind the timezone aggregation problem when doing your analysis. The timezone aggregation problem was discussed in Part 2, Step 4.
    • Before proceeding, please make sure to check the timezone GA uses for your website, so that you can correctly interpret all the remaining steps![3]
    • GA uses a single fixed timezone, rather than the local time of regions. This can cause effects to appear to spread out over more days. The problem is bigger the more worldwide your traffic is.
    • For instance, let us say your website's traffic dips in each region during the weekend in that region, and that GA uses Coordinated Universal Time (UTC). Then, in your GA, you will start seeing a dip on Friday afternoon UTC (because that's already the weekend in far-east timezones like Australia and New Zealand). Moreover, the dip will appear to continue till the middle of Monday UTC, because it's still the weekend in far-west timezones (such as California's timezone, Pacific Time).
    • Filtering your traffic based on region (and then making the timezone adjustment between GA's timezone and the region's local time) can be one way to get around the timezone aggregation problem. Plotting at an hourly granularity may also help with understanding the subtleties of the day boundaries more clearly.
  2. 2
    Use the general heuristic that traffic dips on weekends.[10]
    • Depending on how worldwide your audience is, and your timezone, traffic will be least at the time that is the overlap of weekend time for all relevant timezones.
    • The general heuristic is that daily weekend traffic is somewhere between 50% and 80% of daily weekday traffic.
    • The heuristic is strong for work-based sites that get much of their traffic from direct and search interest. For websites related to home maintenance (such as cooking and gardening) the pattern is usually reversed. For websites that get traffic primarily from social media, the difference between weekdays and weekends is generally smaller in magnitude, and there may even be no clear pattern. More details based on the type of website are discussed Step 6 onward.
    • Desktop traffic generally goes down on weekends whereas mobile traffic (both mobile web and apps) goes up.[10]
  3. 3
    Keep in mind the cultural and geographical limitations of some of the rules.
    • Most of our data is based on sites that get a significant share of traffic from the United States. Moreover, many of our conclusions are based on US traffic, for which clearest information is available. We therefore rely on known facts about the definitions of weekday and weekend by country in different countries.[11]
    • Many countries follow a similar weekday-weekend pattern as the United States: Monday to Friday are workdays and Saturday and Sunday are holidays. In particular, Anglophone countries and European countries, as well as China, most East Asian countries and most African countries, follow the same weekday-weekend pattern. India as well as some South American countries have some places open Monday to Friday and some places open Monday to Saturday, though the trend is toward a 5-day week.
    • In Israel as well as many countries where Muslims form a significant share of the population, the week is from Sunday to Thursday and the weekend is Friday and Saturday. Moreover, in Israel, Saturday (the Sabbath day) is likely to see much lower web usage, whereas in high Muslim population countries, Friday, the day of prayer, is likely to see lower traffic except for websites that directly cater to related needs. Unfortunately, we do not have reliable publicly available data to confirm these plausible hypotheses.
  4. 4
    Use Google Trends to figure out traffic variation in the topics of interest to you.
    • Go to the Google Trends main page, enter a search term in the domain you are interested in, and then select a time range of 30 days. This will display data at a daily granularity and cover a little over 4 weeks. This should be enough to get a sense of the weekly pattern.
    • To get a finer sense of the weekday/weekend differences, you may wish to just plot for the past 7 days. If you do that, data will be displayed at an hourly granularity.
    • You may wish to filter by location to get around the timezone aggregation problem.
    • Keep in mind: Google Trends displays times using your local timezone, even if you filter to a different location. You will need to adjust the time yourself. For more discussion, see Part 2, Step 4.
  5. 5
    Use Wikipedia views (via the WMF Labs tool) to verify the weekly trend.[12]
  6. 6
    If your site caters to work needs of professional audiences, use the heuristic that traffic is roughly constant during weekdays, and a much lower constant level during weekends.
    • More explicitly, the traffic pattern is as follows: reasonably steady traffic from Monday to Thursday (with traffic possibly a little lower on Monday due to the effect of timezone aggregation), then a slight drop on Friday, and a significantly lower level (between 40% and 70% of weekday traffic) on Saturday and Sunday. The data you see in GA may show the weekend spread out more, covering Friday and Monday, due to the timezone aggregation problem discussed in Step 1.
    • Examples of heavily work-driven websites are listed. Keep in mind: you don't need to understand the subject matter of these sites to get the illustrative benefit of these examples! Examples are: programming question-and-answer website Stack Overflow[4], Security Stack Exchange,[13] ServerFault,[14] Engineering Stack Exchange,[15] SuperUser,[16] Ask Ubuntu,[17] Game Developer Stack Exchange,[18] Network Engineering Stack Exchange,[19] Programmers Stack Exchange,[20] Database Administrators Stack Exchange,[21] the Graphic Design Stack Exchange,[22] the Webmasters Stack Exchange,[23] and the Personal Finance & Money Stack Exchange.[24]
    • You can also see the weekly cycle (lower traffic on Saturday and Sunday) by checking out Google Trends for specific keywords that are related to work queries. Keep in mind: you don't need to understand the meanings of the terms to see how they illustrate the weekly cycle! Examples are "jquery" (related to Javascript, used in web programming, part of software engineering),[25] "constructor" (a concept used in object-oriented programming, also part of software engineering),[26] and "dhcp" (a term related to network engineering, in the context of Internet connection protocols).[27] The use of specific keywords can help you better understand traffic patterns in your particular work domain.
    • You can also verify these cycles using Wikipedia pageviews. In particular, the reasonably consistent trends among weekdays and the much lower levels on weekends are easy to see. Since the UTC timezone is used, Monday traffic is pulled down a bit due to the timezone aggregation problem. In general, Wikipedia views are more robust than Google Trends, particularly for the long tail, because no confusing normalizations are performed. Examples from the English Wikipedia: JQuery,[28] "Constructor (object-oriented programming)"[29] and "Dynamic Host Configuration Protocol".[30]
    • Some websites release more fine-grained activity data that you can process to verify weekly patterns. For instance, question-and-answer website StackOverflow's data dump has been used to identify weekly patterns in the number of questions asked on the website.[31] The weekly cycle in number of questions asked closely mirrors the weekly traffic cycle.[4]
  7. 7
    If your website primarily addresses the academic needs of students, use the heuristic that traffic starts dropping Friday, bottoms out on Saturday, and rebounds on Sunday.
    • More explicitly: traffic is fairly consistent from Monday to Thursday, lower on Friday, lowest on Saturday, and then higher on Sunday. Sunday and Friday traffic levels are usually comparable. This is explained by the fact that many students celebrate the weekend starting Friday afternoon/evening, but resume academic work on Sunday in order to meet homework deadlines. In particular, the rebound of traffic starting Sunday is a key indicator that the site is used more by students than by people at work. Note that the data you see in GA may see the weekend effect spread out more, due to timezone aggregation, as discussed in Step 1.
    • Some examples of academic websites with this pattern are the Math Stack Exchange,[32] Physics Stack Exchange,[33] Chemistry Stack Exchange,[34] Biology Stack Exchange,[35] Economics Stack Exchange,[36] GoodReads,[37] and algebra.com.[38]
    • Websites aimed at a more advanced audience (such as graduate students and people with doctoral degrees of equivalent) generally see a less sharp Sunday rebound, because these audiences do not experience the homework deadline effect on Sundays. Examples include MathOverflow (a math Q&A site for students at the graduate student level and professional mathematicians)[39] and the TeX/LaTeX Stack Exchange.[40]
    • Here are some websites that deal with academic content but do not themselves cover academic subject matter. An example of a site where there is a Sunday rebound, but Sunday traffic is still clearly less than Friday traffic, is Student Doctor.[41] College Confidential, a website with advice and discussion on getting into college and doing well in college, also exhibits a Sunday rebound almost at par with academic information websites.[42]
    • You can also verify the phenomenon of the weekly cycle using Google Trends. You need to use search terms that get sufficient traffic volume. Examples of such search terms are "derivative" (a key concept in differential calculus, part of high school and college mathematics),[43] "ethanol" (used in chemistry and biology),[44] and DNA (used in biological sciences).[45] Keep in mind that the trends are clearer and sharper during the academic year, because traffic volume is greater, and some of the phenomena (like Sunday rebound) are specific to the academic year.
    • You can also use Wikipedia pageviews to verify the weekly cycles in data. Wikipedia pageviews are likely to be more robust than Google Trends, particularly for the long tail, because of the absence of confusing normalizations. They are reported in UTC. The Saturday dip is clear across a wide range of examples. Examples from the English Wikipedia: "Derivative" (used in calculus, a part of mathematics),[46] "Ethanol" (used in chemistry and biology),[47] "Hydrochloric acid" (used in chemistry; more noisy),[48] "Normal subgroup" (used in advanced college mathematics),[49] and "Law of demand" (used in early college economics, but also in high school economics classes for students who take them).[50]
  8. 8
    Keep in mind that websites related to home improvement, cooking, and similar activities have traffic peaks on Sunday.
    • Explicitly, the pattern is as follows: traffic reaches its lowest level on Friday, starts going up on Saturday, peaks on Sunday, and then steadily declines till Friday. In some cases, the peak is on Monday rather than Sunday.
    • Examples are the Cooking Stack Exchange,[51] home improvement site Apartment Therapy,[52] cooking site The Kitchn,[53] vegan recipe site Oh She Glows,[54] the Home Improvement (DIY) Stack Exchange,[55] and the Gardening Stack Exchange.[56]
    • You can also use Google Trends to verify the weekly pattern in topics related to cooking and home improvement. Example search terms include "eggplant",[57] "tomato",[58] and "stew".[59]
    • Unfortunately, Wikipedia pageview trends do not display the same weekly cycles. Rather, they are all over the place. This is probably because most people who are looking for cooking recipes and advice don't go to Wikipedia for it, and variation in Wikipedia pageviews for these pages is governed by other factors.
  9. 9
    Keep in mind the following heuristic for websites related to eating out.
    • Relative search interest related to eating out (as shown in Google Trends) goes up during the weekend, peaking on Saturday, somewhat lower on Sunday, and lowest from Monday to Thursday.[60]
    • However, websites related to eating out see their lowest traffic on Sunday. The highest traffic is generally during the week, between Tuesday and Friday. Traffic dips partly on Saturday and goes down more on Sunday.[61][62][63]
    • The higher traffic on Saturday than Sunday is likely because people are more likely to go out to meet friends on Saturday and more likely to stay at home on Sunday. This also explains the opposite pattern for home maintenance and cooking websites discussed in the preceding step.
    • The difference between the observations for search trends and website traffic could be due to a combination of two factors:
      • Search interest is relative to overall search interest, and people search less on the whole on weekends.
      • People are more likely to use the mobile app on weekends.
  10. 10
    Use other heuristics based on the type of site. Keep in mind how the timezone aggregation problem will affect the interpretability of numbers you see in GA, as discussed in Step 1.
    • For websites in some categories, traffic is lowest on Saturday, rises on Sunday, and peaks on Monday, after which it declines through the week. Sunday traffic is generally between the level of Thursday and Friday traffic. Examples are the Parenting Stack Exchange, the Travel Stack Exchange, and the Health Stack Exchange.[64][65][66]
    • Gaming websites see slightly more traffic during the weekends, and substantially more engagement. In other words, there are more sessions, and pageviews/session and total time on site go up. This is because more people have time to play games (and catch up on gaming news) over the weekend, and they have larger blocks of free time.[67][68]
    • For news and commentary sites, regular variation within the week can often be swamped by other fluctuations, such as changes in traffic in response to trending stories. In general, the social media component of traffic sees less of a weekend dip than the search-driven component of traffic. The release cycle for new content as well as the timing of newsletters and social media posts can also affect the weekly cycle. Moreover, since the sites get a lot of their traffic from casual mobile browsing, this also reduces any weekend dip penalty. For some sites, like Refinery29, The Chive, Betty Confidential, or Uproxx, the weekend dip is statistically clear, with the lowest value on Saturday and the highest value on Tuesday or Wednesday.[69][70][71][72] For Vox, the lowest value is on Sunday, but there are a lot of exceptions due to trending news events.[73] For others, such as Upworthy and The Zoe Report, any weekend dip is swamped by other sources of traffic variation.[74] Note that financial news sites do not fall in this category, as they are mainly used by people when markets are trading, so they exhibit traffic patterns similar to work-related sites.[75]
  11. 11
    Keep in mind how the weekly traffic cycle can interact with monthly and annual phenomena.[76]
    • The number of weekends in a month can significantly affect the amount of traffic a website receives in that month. This is particularly important since people often report traffic on a monthly basis.
    • The day of week during which a particular holiday (like Christmas or New Year) falls can affect the traffic patterns for the month. For instance, in the United States, if January 1 falls near the end of the week, people are likely to get to work only the next week. On the other hand, if January 1 falls near the beginning of the week, people are likely to start work right after it.
  12. 12
    Keep in mind the following heuristics for how conversion events can vary between weekdays and weekends. We also discuss the relationship between conversion and traffic, where data is available.[10]
    • For B2B websites, there is much more traffic on weekdays than weekends. However, conversions are even more heavily skewed toward weekdays. In general, people make B2B decisions during working hours. This may be because such decisions generally involve mutual consultation and approval by multiple people, who are more.
    • For B2C websites, traffic is generally higher (in terms of unique users) during weekdays. Conversions are higher during weekdays for simple purchase decisions, including most retail e-commerce. However, pageviews/session and conversions are higher during weekends for cases where the purchase decision involves more complex research that needs to be done on personal time. An example of the latter is the purchase of medical products based on one's personal condition. The details can vary based on the nature of the website.[10]
    • For nonprofits, total donations are highest from Monday to Thursday, lower on Friday, and lowest on Saturday and Sunday. For instance, Network for Good's 2015 Digital Giving Index shows that 12%, 7%, and 8% of total donation amount was donated on Friday, Saturday, and Sunday respectively.[77] Average donation amounts are also lower on weekends than on weekdays, though the margin is pretty small.[77]

Part 5
Understanding the Annual Traffic Cycle

  1. 1
    Identify the broad domain within which your website falls and how that might affect usage.
    • For instance, is your site primarily targeted at students, people in specific jobs, holiday activities, home improvement activities, or something else?
    • For the region and type of activity that your site is targeted at, what is the pattern of use through the year? For instance, if your site targets people at work in specific jobs, what are the peak seasons and the lean seasons for that job?
  2. 2
    Use third-party tools to get a good baseline for seasonal fluctuation.
    • Use Google Trends for seasonal fluctuation in keyword interest. The default display in Google Trends displays data at a monthly granularity since 2004. However, it only displays relative search interest, rather than absolute search volume. Therefore, it is not very good for identifying long-term trends. But it can help with understanding seasonal variation. Note that the default display might offer too little granularity, and a 5-year display might be better to see the finer pattern in the annual cycle.
    • Use QM data for the Stack Exchange in the subject closest to yours, where applicable, or any other site with publicly available QM data. Note: only use sites that have verified Quantcast metrics, as indicated by the checkmark next to the site's name; data for other sites is too unreliable. When using QM data, it is generally preferable to look at data over 3 or more years, and pick websites that were not undergoing rapid growth. This is so that the pattern you see is, for the most part, the periodic trend rather than the long-term growth trend. If you do have to use a site experiencing rapid growth as a benchmark, you will need to control for this growth. Controlling for growth is a little challenging, so do it only if needed.
    • You can also use Alexa or SimilarWeb data for websites similar to yours, to look for annual fluctuation in their traffic rank. It's important to remember, however, that this measurement is reliable only for websites that have fairly high traffic, or those that have certified metrics.
  3. 3
    Use drilldown dimensions to verify hypotheses around the reasons for trends.
    • Drill down by location to verify seasonal trends and annual trends that are location-specific, including academic and work years as well as holidays. You can drill down data by region for your own site in GA. You can also drill down by region in Google Trends.
    • Drill down by referral type (e.g., the default channel grouping in GA) to further verify hypotheses. As a general rule, annual cycles are stronger in search and direct traffic and weaker in social media traffic. This is because people's use of social media fluctuates less throughout the year.
    • Drill down by device type. Holiday dips are likely to be experienced more on desktop than on mobile and tablet.
  4. 4
    For academic sites, keep in mind the following guidelines for the traffic pattern.
    • If the website primarily serves a United States college audience, the pattern is a combination of two kinds of annual academic year structures:
      • The semester structure, that includes a Fall Semester that runs about 16 weeks. It starts somewhere between the second week of August and the first week of September, and ends somewhere between the middle of December and the middle of January. The other semester is the Spring Semester, that runs from the second week of January to the end of May.[78][79][80] There is also a summer term that starts shortly after the conclusion of the Spring Semester and ends shortly before the Fall Semester.
      • The quarter structure, that includes three quarters each of which runs 11 weeks (10 weeks of classes and 1 week for finals), plus a summer quarter. The fall or autumn quarter starts around the last week of September and ends around the second week of December. The winter quarter begins in the first week of January and goes on till the first or second week of March. The spring quarter begins after a one-week break after the winter quarter, and goes on till early June.[81]
    • Non-US traffic will depend on the structure of the academic years in the respective countries. Most Northern Hemisphere countries have an academic year beginning some time in August or September and ending some time in May or June. Therefore the traffic pattern is reasonably similar across countries. If the website attracts significant traffic from school students (secondary education or lower) you will also need to look at the academic year pattern for school students.
    • A good general heuristic about traffic is that it rises for the first few weeks of an academic term, then stays roughly constant during the term. However, after finals are over and people go on vacation, it drops a lot. Note that the pattern within the term can depend on subtle aspects of the website. Some websites can see steady growth within the term because they become more useful to people as they accumulate more material to learn and master. Other websites are more useful during the early stages.
    • By combining the above heuristic and the mix of semester and quarter patterns, the following traffic picture emerges.
      • Summer dip: As summer vacations start in the Northern Hemisphere, traffic takes a dip. The dip begins with the beginning of summer vacation under the semester system (in May) and there is a further dip once those on the quarter system also go on vacation (in June). The dip lasts till the middle of August, when people on the semester system start coming back to work.
      • Autumn traffic: Traffic starts picking up in the second half of August, as people on the semester system get back to classes. It sees another boost in the end of September, as people on te quarter system get back to classes. It then stays steady, growing slightly. There is a brief Thanksgiving dip in te United States, after which traffic returns to old levels.
      • Christmas dip: Traffic drops precipitously around the middle of December, as the academic term (both the quarter and the semester) ends and people go on the Christmas break.
      • Winter/spring traffic: Traffic resumes its growth around the middle of January, as people on both the quarter and semester systems get back to work. It grows till the end of the month. After that, with the exception of a brief dip for spring break, traffic stays mostly steady till the summer dip hits.
    • You can use Stack Exchange websites for academic subjects as benchmarks for the annual traffic cycle. Plot over three or more years so that you can clearly discern the long-term trend of growth (particularly applicable to newer sites) and the annual cycle. For instance, on the Math Stack Exchange, you can select a custom range from August 24, 2010 to August 22, 2016 to see both the annual cycle and the long-term trend. To get rid of weekly fluctuation, select Week under "Show By".[32] For the Biology Stack Exchange, you can go back four years. Since the year-over-year growth rate of this has been higher, the summer dips were obfuscated by this rapid growth for 2014 and 2015. However, the summer dip is clearer in 2016.[35] Other examples include the Physics Stack Exchange[33] and Chemistry Stack Exchange.[34] The Economics Stack Exchange is a little more noisy due to less overall traffic, but exhibits the same general pattern.[36]
    • Filtering location to a university town might be a good way of understanding your annual cycle better. For instance, filter to Stanford or Ann Arbor for a United States university with a quarterly system. Filter to Berkeley for a United States university with a semesterly system. However, traffic to any individual location might be too little. To make out variation clearly, plot at a granularity where you see at least 100 sessions per point plotted. Individual cities might attract too little traffic for this to work, so you may need to look at several of them to get a good picture. Another source of variation, when you are down to the level of the individual university town, is that specific features of the courses taught in that year at that university can affect traffic patterns. This becomes more of an issue the more specialized and less diverse your content is.
    • Another valuable heuristic for discerning the annual cycle for your website, and for specific content on your website, is Google Trends. Google Trends at the level of individual search queries can be even more granular, but also more noisy. It is more granular because it can capture information about the annual trends in specific search queries. Search queries related to concepts generally covered more in the fall quarter or semester will see a bigger peak in the fall months than in the winter/spring months. In contrast, search queries more customarily covered in the winter/spring quarter or the spring semester will see more traffic in those months. For instance, "derivative" is the key topic of differential calculus and is generally covered in the fall quarter or semester. It sees its search traffic peak in October, and a much smaller peak in February.[82] In contrast, "Taylor series", a topic generally covered in more advanced winter and spring courses, sees a slightly bigger peak in April.[83] You might be able to use Google Trends to better understand what the expected annual cycle should be for the mix of content that you offer. Keep in mind, though, that search volumes for highly specialized or obscure terms may be too low for the patterns to be statistically robust.
  5. 5
    Keep in mind the following heuristics for professional information websites.
    • The typical traffic pattern is as follows:
      • Worldwide Christmas dip: There is a sharp but brief dip during the week between Christmas and the New Year.
      • Regional dips for regional holidays: For instance, United States traffic sees a dip during the United States Thanksgiving, but there is no similar dip in non-US traffic.
      • Slight summer dip (Northern Hemisphere): Work-related sites see a slight dip in traffic during the summer, but not as pronounced as for academic websites. There could be two reasons for the dip. First, some of these sites are also used by students, although less than by people at work. For instance, StackOverflow is primarily used by people who code for a living, but is also used somewhat by students taking programming courses. The latter could account for the summer dip. A second reason for the summer dip is that professionals are more likely to take time off during the summer months (July and August). However, they do not all take time off at the same time, so the effect of this dip is mild but spread over a longer time period.
    • You can use data from the following sites to get a sense of the annual traffic cycles for professional websites: StackOverflow,[4] ServerFault,[14] and Security Stack Exchange.[13]
    • You can see similar patterns using Google Trends. Since the Christmas dip is brief and may be hard to see, plot with a time range of 5 years so that you get weekly granularity. Also keep in mind that because Google Trends is relative, only very strong trends can be discerned. For instance, for "jquery" you see the sharp Christmas dip and a very slight (but hard to discern) drop every summer. The analysis is complicated by the fact that traffic is also going down steadily.[84] On the other hand, "dhcp" sees a Christmas dip but no summer dip.[85]
  6. 6
    Use the following traffic pattern heuristics for business-to-business (B2B) websites. Some of the details of these traffic patterns are specific to the United States, so do not blindly apply to different geographical locations.[76]
    • The annual traffic cycle is as follows:
      • January traffic can be slow to start, and depends on whether January 1 was early in its week (in which case it would start earlier and go longer) or late in the week (in which case people would start work next week, shortening the effective month).
      • February is shorter by two or three days.
      • March and April traffic can be erratic due to spring break and Easter.
      • May and June see consistent traffic.
      • July and August see lower inbound traffic due to vacations.
      • September and October see the highest traffic.
      • November traffic is decent till the week of Thanksgiving. It starts dipping on Monday and falls even further Wednesday through Sunday.
      • December traffic is low and falls through the month, with the second half having lower traffic. The day of the week that Christmas lands can affect traffic levels.
    • Unfortunately, very few B2B websites use Quantcast (since Quantcast is primarily used by websites who monetize by showing advertisements, and B2B websites are selling their own product). Therefore, we don't have examples of publicly revealed traffic data.
  7. 7
    Keep in mind some heuristics for annual traffic cycles specific to the type of site. Here are some heuristics for sites other than the academic, professional, and B2B categories already discussed above.
    • Websites related to outdoor activities and home improvement generally see lower traffic in the winter and higher traffic in the summer. The pattern can vary depending on the extent to which winter impairs activity. For instance, the Gardening Stack Exchange peaks in May and June, then steadily falls till December and January, and then again steadily rises till May and June.[56]
    • Websites related to travel see steady year-round behavior. Luxury travel goes up during vacation times but business travel goes down, with roughly cancelling effects. The Travel Stack Exchange has a remarkably steady year-round behavior, but does see a slight summer increase. It also sees less of a Christmas dip than professional, academic, and B2B websites.[65]
    • Websites related to cooking are quite stable year-round. The main anomalies are around specific holidays. Interestingly, in the United States, cooking websites see an increase in traffic around Thanksgiving and also sometimes the week before Christmas, but a decrease around Christmas. Some example websites to look at are the Cooking Stack Exchange,[51] Oh She Glows,[54] and The Kitchn.[53] Note that the Thanksgiving increase is stronger for Oh She Glows and the Kitchn than for the Cooking Stack Exchange. This is probably a result of the first two sites (that are publisher-driven) putting extra effort to promote their content in the run-up to Thanksgiving, whereas the Cooking Stack Exchange (that is community-driven) does not see this sort of concerted effort.
    • Entertainment and gossip websites generally do not see a holiday dip. Some of them may even see a traffic increase during holidays, particularly if they cover feel-good stories and help people pass their vacation time. In general, their traffic pattern does not see discernible annual cycles.
    • News sites generally do not have discernible annual patterns, but rather, their patterns are driven by the nature of events in that year. They may, in fact, even see patterns across time periods longer than a year. For instance, politics news sites see more traffic closer in time to elections, which may occur once every few years.
    • Websites related to tax information for the United States see high traffic near the calendar year transition (December and January) as well as during tax filing season (beginning in the middle of February and proceeding till around the middle of April in the United States). Websites with similar tax information in other countries will see traffic patterns governed by the tax year and filing deadlines of those countries.
  8. 8
    Keep in mind the following annual pattern heuristics for conversion-related traffic. We consider the cases of charitable giving and e-commerce.
    • Charitable giving in the United States is highest in the month of December, and second highest in January.[77] Tax optimization as well as festival season generosity are contributing factors, but these are amplified by nonprofits engaging in marketing and fundraising drives at this time of year. Websites of charities as well as those related to charity evaluation see more traffic in December (and after that, in November and January) than in the rest of the year. Examples include charity evaluator GiveWell (based on their Google Analytics and Clicky web traffic data, as well as donation metrics as well as Alexa data)[86][87] and Charity Navigator (as recorded in Alexa data).[88]
    • In most countries, retail spending is highest at the turn of the year for that country. In the United States and in most countries that follow the Gregorian Calendar, the retail peak is experienced in December. In the United States specifically, sales are greatest between Thanksgiving and Christmas. Particularly heavy days for sales are Black Friday and Cyber Monday.[89] In China, the biggest day for e-commerce is November 11, known as Singles Day,[90] whereas in India, sales are highest during Diwali.[91] With that said, if your website does not explicitly offer special deals for these occasions, traffic and conversions on your website are not likely to go up much.
  9. 9
    Use age, gender, and location filtering where relevant to verify hypotheses.

Part 6
Understanding the Daily Traffic Cycle

  1. 1
    Use hourly granularity to study your website's traffic variation through the day.
    • Keep in mind the heuristic discussed in Part 2, Step 3: you should aim to study traffic at hourly granularity only if your website gets on the order of 4,000 or more pageviews a day. Below that, random variation may make it hard to see trends. You can still try at lower traffic levels; if you do see a clear pattern that can be illuminating.
    • GA allows you to look at historical data at an hourly granularity, as well as get real-time analytics with totals for the most recent 30 minutes.
    • The analytics tool available with WordPress' Jetpack allows you to get a graph of the most recent 48 hours of data, as well as the total views so far in the current UTC day.
    • Other tools might allow you to see traffic at an even finer granularity.
  2. 2
    Adjust for the following when considering traffic variation.
    • It often makes the most sense to consider traffic separately for regions that are in different timezones, and to filter your view to only consider a specific timezone. This is particularly relevant for organic search and direct traffic that is governed by local time.
    • Keep in mind: GA uses a single fixed timezone when displaying data (see Part 2, Step 4 for an explanation). You can check the timezone in use by going to "View Settings" in GA's Admin Panel.[3] Therefore, when looking at traffic in different regions, you will need to do a timezone adjustment to figure out what the local time in those regions is.
    • If you are sending newsletters or posting to social media at particular times of day, then you should view your daily traffic patterns in relation to that time of day.
  3. 3
    Use Google Trends to figure out traffic variation within the day for search terms related to your website.
    • Go to the Google Trends main page, enter a search term in the domain you are interested in, and then select a time range of 7 days. This is usually enough to capture hourly trends on all days of the week. The nature of hourly trends may be different for weekdays and weekends, so a 7-day period is recommended. You can also filter by region.
    • One key caveat: Google Trends displays search interest in a particular topic relative to overall search interest at that time (and in that region, if you are filtering by region). The data is then normalized so that the largest data point is scaled to 100. This is not that big a deal when viewing data at the daily granularity, since total Internet use does not vary that much between days. However, it is a big issue within the day, since search traffic can differ significantly through the day. In particular, search keywords that appear to show peaks at night do not necessarily get the most traffic volume at night; they just see the largest share of traffic volume at night.
    • A second caveat: Google Trends uses a single fixed timezone, namely, your local timezone (see Part 2, Step 4 for more information). Therefore, you will need to make timezone adjustments when you filter to different regions.
    • A third caveat around using Google Trends data at this fine a granularity is that for many search terms, there is insufficient volume to get a clear pattern. Moreover, because of timezone issues, it is necessary to drill down to specific locations, which makes the data volume even lower and more noisy. You may therefore have to settle for using more broad-based search queries than the ones you are actually interested in.
  4. 4
    Keep in mind a few heuristics regarding traffic variation within the day.
    • There is a daily cycle in how people use the Internet.
      • People use the Internet less at night, because they are likely to be asleep.
      • Internet use starts in the morning but is generally lower at that time.
      • Work-related Internet use is high during work hours.
      • Overall Internet traffic in a timezone is maximum in the evening hours for that timezone, i.e., between 7 PM and 11 PM. In analogy with the rush hour for transportation, this is known as Internet Rush Hour.[92][93] Much of the rush hour Internet traffic relates to entertainment, including streaming music, movies, and porn, browsing social media, and reading celebrity gossip.
      • In particular, during Internet rush hour, social media-driven sites also see more traffic at this time since visiting social media is a common evening leisure activity.[94]
      • Organic effects of greater traffic during rush hour are often amplified by publishers who time their social media posts to Internet rush hour.[93] Search-driven traffic does not see this additional boost, and therefore sees less of a rush hour peak.
    • The daily cycle has different shapes in different parts of the world. There are a few sources of variation.
      • The day-night gap in traffic varies heavily by region. In some regions, the ratio of the daytime peak to the nightly low can be as low as 3, whereas in other regions it can be higher than 5.
      • You can get a crude estimate of the shape of traffic in different continents using Akamai's real-time web metrics tool. This shows HTTP hits per second over the most recent 24 hours (ending now) with times displayed in UTC. Unfortunately, the free tool only offers granularity at the level of continent, and a single continent can span many timezones, so it is not a very good indicator of day-night variation in a single timezone.[95]
    • Note that if your traffic is global, then it can appear to be fairly even throughout the 24-hour period because of different timezones. You should therefore filter by region to see the daily cycle more clearly.
  5. 5
    Keep in mind the following other general heuristics.
    • Some professions have a larger share of "night owls" than others. For these professions, even though night-time traffic is lower, it does not fall too low. In fact, Google Trends data for related searches may even show late-night peaks. Absolute traffic is lower but the share of traffic could rise. Programming-related professions have a high share of night owls. For instance, for the keyword "jquery" in California, Google Trends shows peaks (after adjusting timezones) at 11 AM, 2 PM, and 2 AM, with the 2 AM peak the highest.[96]
    • Websites that cater to needs specific to particular times of the day see traffic peaks at those times of the day. For instance, websites that provide news and information for day traders see a peak around the time of the opening of the market, and another peak around the time of the close of the market, since trade volume is largest at these times. In the United States, the main stock markets are located in New York, and open around 9 AM Eastern Time. Therefore, even in California, interest peaks at 9 AM Eastern Time, or 6 AM Pacific Time (the local time in California).[97]
    • Websites that people may visit for planning their day see a peak early in the morning, or in the evening, or at both times. This includes websites related to transportation and retail (e.g., the stores people plan to visit).[98][99]
    • For some other examples of analysis of traffic fluctuation within the day, and how these within-day trends differ between weekdays and weekends, see Pornhub's 2015 in review[100] and the New York Times article on minute-by-minute Google Trends.[101]
  6. 6
    Use additional activity data where available to discern the daily cycle.

Part 7
Understanding the Traffic Pattern Following Push Factors

  1. 1
    If posting to Facebook, keep in mind the way Facebook circulation of posts changes over time.
    • For an initial period after the post (that could last anywhere from a few minutes to an hour) Facebook circulates the post within a select subset of the potential audience, gradually ramping up the rate of circulation. The extent to which it receives a positive reaction from that initial subset determines how much further Facebook ramps up the post.
    • After the post is ramped up, Facebook exponentially decays the circulation of the post. In particular, if the post does not receive a lot of organic re-shares, traffic to the post from Facebook decays exponentially. The traffic graph shape is concave up. This sort of traffic shape is seen when the driving factor in the post's circulation is the huge reach. Without any special boosting, this is determined by the number of people who have liked the page. With special boosting, it depends on how heavily the post was boosted.
    • If, on the other hand, the post receives a lot of organic re-shares, the decay pattern will not be exponential. Traffic shape will be concave down. This sort of traffic shape is seen when genuine engagement rather than raw volume is that driving factor for the post's circulation.
    • In general, keep in mind the following heuristics for the ballparks of ratios. The number of people who see your post will usually not be more than 20% of the eligible population, but it can be much lower, particularly if you have a large number of people who have followed and liked the page. A typical range is 0.5% to 10%. The click-through rates for people who do see the post can vary from 0.5% to 10%, and usually from 1% to 5%. The ratio of likes + reactions to clicks generally varies from 1% to 5% (though some posts get a lot of likes without clicks, and others get a lot of clicks without likes). Overall, a page with a few million likes should expect to get in the tens of thousands in pageviews from an average Facebook post.

Part 8
Understanding the Long-Term Trends for Your Site

  1. 1
    Identify the various parameters that would affect the way your site would grow.
    • Long-term trends for Internet access for your target audience
    • Long-term trends for the domain or topic area: Some domains are experiencing overall growth, and if your site is in one of those domains, then it should grow organically as a result of that.
    • Changes to search algorithms (especially Google Search) and social media algorithms (especially Facebook).
    • Your own strategy for growing content on your website and promoting it.
  2. 2
    Use the following types of sources to get a sense of global patterns in web usage.
    • comScore is a global media measurement and Internet analytics company. Based on its many measurement tools, comScore provides regular estimates of the extent to which people are using the web on various devices (personal computer and smartphone) as well as their use of apps. Its focus is on United States traffic. comScore data has been cited for estimates of the relative use of mobile and desktop,[102] the relative use of apps and the webs,[103] and stagnation in desktop use.[104]
    • Survey research organizations and nonprofits occasionally conduct surveys on people's Internet and website usage. The Pew Research Center's Internet, Science and Tech section (pewinternet.org) contains a number of such surveys, focused on United States audiences.[105][106] Another source is Public Knowledge (at publicknowledge.org).[107]
    • You can get statistics on Internet penetration in various countries on the International Telecommunications Union (ITU) website.[108] You can get more references for current Internet penetration and trends worldwide and in specific countries by looking at the references from the associated Wikipedia pages.[109][110]
  3. 3
    Keep in mind the following broad trends in Internet use for the United States.
    • People's consumption of the Internet has been shifting from desktop to mobile and from web to apps. According to comScore data, mobile Internet use accounts for about double of desktop and laptop Internet use[102] and app use accounts for a little over half of Internet use (so, a little more than web use).[103] Moreover, comScore data also suggests that desktop use peaked in 2015.[102][104] comScore data on the use of Wikipedia, the world's foremost information site, shows a decline in unique visitors starting in late 2013, with the decline sharpest on desktop (with mobile numbers staying roughly constant or going up) and biggest in North America.[111]
    • User surveys of people's changes in Internet use, as well as changes they are considering, show that a minority (but a non-negligible one) of users have either stopped or are considering stopping their home broadband subscriptions because smartphone plans meet their Internet use needs. Users who are seriously considering this tend to have lower incomes and education levels. They cite both the cost of home broadband and the fact that they can accomplish the most important things through a smartphone as factors that influence the decision.[106][107] Most users, however, see value in both desktop and mobile Internet. Mobile Internet is primarily used for communication, quick information consumption, and social media whereas desktop Internet is used more for video consumption and buying things.[107]
    • Advertising spend is another proxy for the growth in Internet use. In the United States, desktop advertising spend has been mostly stable since 2012, whereas mobile advertising spend has approximately doubled year-over-year from 2010 to 2015 (however, its growth rate is now falling).[112]
  4. 4
    Keep in mind the following heuristics regarding website traffic growth patterns.
    • Websites that deal with work-related topics catering to a tech-savvy audience are likely to have seen their growth and peak earlier. In particular, organic growth rates for established websites in this category since about 2014 would be relatively small, under 10% per year in the United States and under 20% per year worldwide. Examples are websites in the Stack Exchange Network such as Stack Overflow,[4] ServerFault,[14] AskUbuntu,[17] and Math Stack Exchange.[32]
    • Women's magazines have, since 2014, seen a slight decline in desktop web use and an increase in mobile web use, for a slight net increase.[113][114][115]
    • Websites with social media-driven traffic driven by "clickbait" have generally seen an increase in traffic till somewhere between 2012 and 2014, but a decline in traffic since about 2014. In particular, a serious decline begins around the middle of 2016, as Facebook rolls out new ways to combat clickbait in the news feed.[116] However, this rule has a lot of exceptions, as different kinds of websites and different kinds of social media strategies are affected in different ways by changes to Facebook's news feed algorithms and users' changing social media habits. An example of a social media website that has seen steady decline in traffic is Upworthy.[74]
    • New genre websites in topics with less of a tech-savvy bias have much more room for rapid growth in recent years, but this growth is not guaranteed. Examples of such domains include sports, cooking, and fashion.
  5. 5
    Keep in mind the following information sources for search and social media updates that might affect your website traffic.
    • For search updates, use announcements on algorithm changes to Google Search. A good place to get an up-to-date timeline of all search algorithm changes is the Moz page on Google algorithm change history.[117] Another source is Wikipedia's timeline of Google Search, but this is less likely to be kept continuously up-to-date.[118] You can compare these changes against the search traffic to your website. If you have enabled Search Console, you should also be able to see data on total impressions and clicks on Google search queries that led to your website. If you see sharp changes in the search traffic occurring at a time when Google made an algorithm update, the change is likely driven by that algorithm update. You might be able to use the direction of such changes to better understand how further planned algorithm updates will affect your website traffic. There is a lot of online discussion on Google search algorithm changes, as well as paid tools that can help you understand and optimize against those.
    • For a history of Facebook news feed algorithm changes, see the events related to Product (news feed) in the Wikipedia timeline of Facebook page, and follow the references to get more details on the updates.[119] If you see sharp changes in the reach and engagement of your posts, or the traffic to your website from Facebook, at the time Facebook makes algorithm changes, then you are likely affected by those changes.

Community Q&A

Ask a Question
If this question (or a similar one) is answered twice in this section, please click here to let us know.

Sources and Citations

  1. Google Trends Updated to Provide Minute-by-Minute Data on Trending Searches, Matt Southern, Search Engine Journal, June 17, 2015
  2. What 16 Studies Say About The Best Times To Post On Social Media, Nathan Ellering, CoSchedule Blog, April 13, 2016
  3. view settings. Update view name, time zone, Site Search parameters, and other settings. Google Analytics. Retrieved August 26, 2016
Show more... (116)

Article Info

Categories: Internet

Thanks to all authors for creating a page that has been read 781 times.

Did this article help you?