Left Side Right Side What the Colors Mean in FouAnalytics

Left Side Right Side What the Colors Mean in FouAnalytics

Given the influx of new advertisers and new websites using FouAnalytics, it's time for a refresher. BuiltWith shows a portion of the number of sites using FouAnalytics to see and manage bot traffic on their sites. FouAnalytics is usually used together with Google Analytics (GA) or Adobe Analytics (AA), not in place of. FouAnalytics shows the site owner details about bot traffic that GA and AA don't. In fact, FouAnalytics is used to troubleshoot discrepancies in GA and AA because both of those platforms filter SOME bot traffic. The problem is they don't report WHAT bots they were or HOW MUCH was filtered, leaving discrepancies that were not explained. Adding FouAnalytics to sites allows the site owners to troubleshoot those discrepancies themselves.

No alt text provided for this image


Left side vs right side donut charts

When you log into FouAnalytics, you will first see a time series chart. Below that, you will see a pair of donut charts. The left side -- "Live data loaded into browser" -- means the data that is loaded into the browser. This data populates the data grids below. The right side -- "Historic total data for date range" -- is the total data for the time period selected (usually the last month, by default). The reason I have two donut charts side by side is so you can see if there's anything out of the ordinary happening right now (live data) compared to historical.

No alt text provided for this image

Looking at this example, the ratios of blue and red are very similar between left and right, so there is nothing major going on right now -- i.e. we are not under a bot attack. As for the colors, let me take a step back and explain how we do the flagging -- red flags or blue flags.


Red flags versus blue flags

When I started building these tools 11 years ago, to help me audit campaigns for advertisers, we started with deterministic javascript parameters. We collected every parameter that was possible to collect using javascript -- from screen resolution to user agent to window size, etc. We looked for discrepancies that told us bot makers messed up and made mistakes. For example, if they lied and said the user agent was an iPhone, but we detected the screen resolution to be 1920x1080, that is a red flag (because that is not an iPhone screen resolution). If the IP address were a datacenter, that's another red flag because humans didn't access the internet through datacenters (some do now, when using VPNs). If we accumulated 3 or more red flags for a single pageview or single impression, we marked that as dark red (confirmed bots). Otherwise it is light red (suspected bots). Obviously bots have gotten more sophisticated over the last 10 years, so I have continued to tune and update the algorithm to account for new characteristics, as they appear in the data that I scrutinize.

Blue flags work the same way and are typically based on human interaction events. For example, when humans move the mouse, scroll the page, click on something, or touch the screen on smartphones. If we have 3 or more blue flags we mark that pageview as dark blue (confirmed humans), otherwise it is light blue (likely humans). There's obviously a lot of other stuff happening behind the scenes which I won't get into here. But if any of you are interested, feel free to ask me privately.

Gray means we don't have enough red or blue flags to mark it either way.

No alt text provided for this image


Yellow, Orange, and White

You may also see yellow, orange, and white in your data. Yellow means search engine crawlers. These are the bots that come to your website to index the content for search engines. These are good bots and are fine when you see them coming to your site. But we also want to make sure they don't come in too large quantities. Be sure to look for false flags too -- last year, we noticed bad guys were naming their bots "googlebot" in the user agent. This is a "false flag" meant to deceive analytics into thinking it was Google's search crawler. After discovering this fake googlebot, I updated the FouAnalytics algorithm and it is now categorized as dark red, fraud.

Orange means declared bots -- ones that say their name honestly in the user agent. They tell you they are bots, like "headless chrome." Headless chrome is an automation tool used by developers to test their own websites. But as a browser that can be remotely controlled and automated, it is also used to crawl and scrape sites. But as long as they disclose they are a bot in the user agent, we label these as orange.

Red means bad bots. Obviously the automation tools like headless chrome are used for ad fraud -- creating fake traffic to sites to load the pages and generate ad impressions out of thin air. The latter are marked as dark red because they falsify the user agent deliberately to avoid getting detected. That is why I have said those legacy fraud verification firms that rely on user agents for bot detection are not catching most of the bots.

Finally, the white area means "incomplete js" or "noscript." The former means the FouAnalytics tag was loaded but no data was written back. The latter means the javascript tag was not called because javascript was not available (e.g. in certain environments like CTV).


Data grids

When you scroll further down the dashboard, you will see data grids. They are populated with the data from the 10,000 pageviews/impressions loaded into the browser. This corresponds to the left side donut chart -- "live data."

No alt text provided for this image

Let me draw your eyes to 2 things. See the yellow highlight at the top of each data grid. That tells you the prevalence of that variable in the live data. For example, out of the 10,000 pageviews loaded into the dashboard, 6.8% have a utm_source in the url, 1.1% have a utm_campaign, and 1.5% have utm_medium. You can also see that source (what page or where the ad ran) is detected in 99.5% of the data; that means we don't have a measurability issue.

Then in each data grid itself, you will see a percent column and a count column. This represents the further breakdown of the values of that parameter. For example, in the UTM_SOURCE dat grid, you see a total of 676 pageviews. Out of that 494 (73.1% of 676) have fbclid ("facebook click ID"), etc.

Finally note the color coding under each row. This gives you a way to glance at the relative quality of that row versus the others.

For further background and examples of how publishers (site owners) and advertisers use FouAnalytics to troubleshoot website traffic and programmatic campaigns, see the following:


Tour?of FouAnalytics Dashboard and Campaign Checklist

https://www.dhirubhai.net/pulse/tour-fouanalytics-dashboard-campaign-checklist-dr-augustine


How?Site-Owners?Use FouAnalytics to Troubleshoot Bot Traffic

https://www.dhirubhai.net/pulse/how-site-owners-use-fouanalytics-troubleshoot-bot-dr-augustine


How to use FouAnalytics to Scrutinize Clicks from Programmatic Campaigns

https://www.dhirubhai.net/pulse/how-use-fouanalytics-scrutinize-clicks-from-programmatic-fou


If you have any further questions, please ask me.





要查看或添加评论,请登录

Dr. Augustine Fou的更多文章

社区洞察

其他会员也浏览了