How We Count Our Users

We want our count of unique users that the AddThis network sees per month to be one our customers can believe in.  This post explains our methodology for generating that number.

There are 2 numbers to consider:

  • Raw uniques – The raw number is really a count of all cookies, including bots and everything.
  • Good UID – We use a “good UID,” a number that we know is lower than our actual reach, but better to be conservative.


Here’s how we get to our “good” UID:

  • Start with time component:  We start with every cookie that has been seen at least twice over a period greater than 24 hours apart because we need to be sure that we do not include bots and consumers who constantly clear their caches.  The time component is crucial to an accurate count because there are many cases where transient consumers appear in the log files multiple times in a narrow time period but then disapear from the logs a short while later.  A filter that counts any consumer seen multiple times would count this activity incorrectly and cause the unique user count to be inflated.
  • Count Unique Monthly Users: Then we take a de-duplicated count across all of the days in the month we are measuring to get the total count.  As we have discussed in another blog post, this is not a trivial task and it is tempting to take the easy way out and simply sum the total unique users by day in order to get the final unique user count.  Doing so would overestimate the actual number of unique users counted.
  • Consider multiple devices: Unfortunately there is no reasonable way to detect when the same person uses multiple devices to access the internet, this is very common in the US and other westernized countries.  For this reason, we take the floor of our total unique user count to reduce the impact of overcounting multiple devices.


Even with our conservative counting approach, we’re proud to be on sites that reach 1.3B users on a monthly basis, making us the largest social sharing platform on the open web. We’d love your feedback and would be happy to expand on any aspect of our methodology.

  • Dominic

    Curious how you make use of the word “user”.

    99.99% of the impressions addthis collects and tracks.. most likely the… “user”… has no idea addthis even exists or they have been flash cookied.

    So isn’t it a bit misleading to ever talk about… “users”… when you talk about addthis?

  • Matt Abrams

    AddThis provides light weight features that do not require a end user login. When we discuss unique users we are referring to the billion plus unique vistiors that use web pages with AddThis tools on them every month.

  • http://pro100blogger.com vvovvo

    Thank you. I’ll try to add it to my blog. Can I add buttons Russian social network?

  • Dominic

    Matt – so the term “user” technically is not valid. You could say, “1 Billion Unique People were ‘used’ by the AddThis tracking engine”.. but in reality probably 99.9% of the “users” as you define them, have no idea that AddThis even exists.

    No biggie, just a funny use of the term “user”.

  • Matt Abrams

    Dominic – Of course you are welcome to your opinion. At AddThis we work hard every day to provide pleasant social experiences to our publishers and end users.

  • Corbin

    So what do you mean by “For this reason, we take the floor of our total unique user count to reduce the impact of overcounting multiple devices.?”

    Usually floor is used in finding the nearest whole integer?

  • Matt Abrams

    Corbin – I was using it as a general term in this case. Basically we round the number down significantly and apply some other basic sanity functions to get a conservative result.

  • Corbin

    Matt that makes sense. How did you calculate this divisor? Just taking a guess maybe you can look for user-agents per ip with some URL traffic analisys and determine if it’s the same user on a different device.

  • Matt Abrams

    Corbin – What you suggest is possible but not practical at our scale and is not full proof. The fact is not method is full proof (even if you have users log in directly to your application because they could create multiple accounts). We do count the unique IP/UA on occasion to help validate our number. This number usually comes back much higher than our count. To account for multiple devices we take a heuristic of the average number of devices per user per geographical region to try to take a conservative value. Do you face this issue with your application? What is your approach?

    matt

  • http://bloger-man.ru/ Юрий

    1.3B, seriously? Why so much?

  • http://www.addthis.com/ AddThis

    Totally. AddThis is on 14 million websites worldwide. :)