COOKIELESS and CLICKLESS attribution with FouAnalytics

Executive Summary

Individual ads do not cause individual conversions/sales. This is why the cookie-based attribution methods used for the last 10 years were flawed and over-attributed sales that were not actually caused by said ad impression. Furthermore, various forms of fraud, the largest of which is affiliate fraud, use cookies to steal affiliate commissions by falsely claiming credit for sales it had nothing to do with. Cookieless and clickless attribution eliminates affiliate fraud and increases the accuracy of attribution. FouAnalytics is cookieless analytics. By using geolocation, timestamp, and device-browser fingerprints, FouAnalytics enables advertisers to calculate "conversion velocity" and "attentiveness." This answers business questions like 1) which channels and tactics drive higher "conversion velocity," 2) how do I reallocate dollars among different channels (marketing mix) to maximize outcomes with the same budget (zero-based cost management), and 3) how do I improve ad creatives and tactics to maximize "attentiveness" even if conversion events happen offline or are not easily visible online. Even advertisers that do TV and CTV advertising can do proper cookieless attribution with FouAnalytics. See the automotive, financial services, and tourism examples below.

Original Full Article (brace yourselves)

Hard to believe, but we have been doing this for years; we just didn't call it "cookieless attribution" until recently. Advertisers have always wanted to know whether their digital ads drove sales and outcomes. Yes, their digital ads do drive outcomes, and here's how advertisers have been using FouAnalytics to do attribution over the years.

Individual ads do not cause individual conversions

Let me reiterate something that is extremely important for advertisers to understand, before getting into attribution analyses with FouAnalytics. "Individual ads do not cause individual conversions." Yes, read that one more time, "INDIVIDUAL ads do not cause INDIVIDUAL conversions." Furthermore, individual TYPES of ads don't cause conversions. All types of ads contribute something to the customer journey that leads to the conversion. The problem is that it is very hard to give credit to each ad or each type of ad (display ad versus search ad) or even know how many ads it took to drive a conversion for different customers. All the current attribution models that use cookies "effectively created a fictitious cause and effect that everyone happily bought into." If someone was exposed to an ad, a cookie was set to mark that user as "exposed." Then any conversion that happens after that is claimed to be caused by that ad. This includes "foot-fall" in offline stores -- i.e. if an exposed device walks into a store, the attribution model claims that sale as if it were caused by that 1 ad impression. In this way, most current attribution models vastly overstate the number of conversions they caused. This misleads advertisers into thinking certain channels and tactics produced so many outcomes that they over-invest in said channels and tactics that didn't actually drive that many conversions.

See: You don't need cookies to track conversions, why?

One other note for advanced practitioners -- if you understand the above, you will also realize that you don't need "identity resolution" (which means figuring out which user made the purchase). In other words, advertisers don't need to know which individual person bought their product; advertisers just need to know that more people who were exposed to their ads converted (i.e. higher "conversion velocity") than people who were not exposed to their ads. Advertisers were convinced they needed "identity resolution" services (for targeting ads and for attribution) by adtech companies selling them those services over the last 10 years. These didn't make advertising better; in fact, it caused more wasted ad dollars.

See: Cookie-based targeting is inaccurate at best

Affiliate fraud uses cookie-based attribution

Even more dollars were lost to one of the most egregious and long-running fraud schemes that use cookie-based attribution, "affiliate marketing." While the original idea was valid -- set an affiliate cookie on a user that clicked through from a useful piece of content and then later completed a purchase -- fraudsters have gamed the system for the last 2 decades to steal money at-scale. They did this simply by "cookie-stuffing" as many users' browsers as possible. If any of those users proceeded to buy anything from any of the advertisers, the fraudsters would get affiliate commissions under false pretenses. The advertiser would think that affiliate marketing was working so well, they'd pour more money into it, not knowing that those affiliate payments were unnecessary costs because the sales would have happened anyway. All of these fraud schemes were possible because (affiliate) cookies were used to track conversions. An individual fraudster, Shawn Hogan, amassed $28 million in affiliate commissions fraudulently in 2007-08, on an estimated $2 billion of sales on eBay. Source: [1] This was done through a massive cookie-stuffing scheme that planted affiliate cookies on tens of millions of users' browsers, unbeknownst to the users, so that when any of them completed a purchase on eBay, the affiliate commission would be paid to Hogan. To this day, browser toolbars and extensions, hidden iframes on webpages or even ad slots, and stacked redirects stuff thousands of affiliate cookies on millions of users' browsers continuously to rake in affiliate commissions at-scale (fraudulently).

Cookieless attribution increases accuracy and eliminates affiliate fraud

If you now understand that cookie-based attribution was flawed, because it dramatically overstated success and has been rigorously exploited by fraudsters over the last decade and a half to steal billions from advertisers, let's ask how to do cookieless attribution. Not only are we compelled to do this (since Chrome is deprecating third-party cookies soon), cookieless attribution also makes sense because it reduces false "cause and effect" (this ad caused this sale) AND reduces the forms of fraud that rely on setting cookies to claim credit under false pretenses, like affiliate fraud.

See: FouAnalytics is a GDPR-Compliant Alternative to Google Analytics

FouAnalytics is cookieless and is used by advertisers to do cookieless attribution to determine if their digital marketing is working, how well it's working, and what optimizations can be made. FouAnalytics never sets cookies, never have and never will, because cookies are unnecessary for fraud detection. FouAnalytics will not be affected by third party cookies going away and has not been affected by Safari, Firefox, and Brave initiatives that further protect users' privacy. FouAnalytics is privacy-preserving analytics, and the data is not used for any other purpose, unlike Google Analytics and other Google advertising products.

Cookieless attribution for ecommerce advertisers

Onward to cookieless attribution. Let's start with a simple use case -- advertisers that have ecommerce sites (where conversion events can be seen). The ecommerce site has FouAnalytics on-site tags installed. We start by filtering down to those user sessions that ended up on the thank you page after checkout -- i.e. a conversion. Then we run a report that shows state (GEO_REGION_0) and city (GEO_CITY_0), along with fingerprint and timestamp. Fingerprint is an anonymous representation of a device-browser combination -- i.e. a user. For example, on the same PC, different browsers like Chrome and Firefox would have different fingerprints. The fingerprint in the data grid below is used to group pageviews together into a session and confirm that it is the same user using the same browser to complete the purchase. The timestamp is the Unix Timestamp (in seconds). In the several examples below, you can see that each session that ended in a conversion included about 6 - 8 pageviews (yellow highlight).

Advertisers can then pair the data above with UTM_SOURCE, UTM_MEDIUM, and UTM_CAMPAIGN to see which channels (source), media (search vs display), and campaign were driving more conversion events and increase budget to those channels, tactics, or campaigns.

Cookieless attribution for advertisers running TV and CTV ads -- conversion velocity

The example above is easy, because we get to see UTM parameters and also ecommerce transaction events (conversions). How do we do cookieless attribution for advertisers running TV ads or CTV ("connected TV") ads? In this example, the advertiser still has a website that is measured by FouAnalytics on-site tags. The advertiser may use a QR code on the TV/CTV ads which lead to the website (like Intuit TurboTax did on their Superbowl TV ads). Human users that saw the TV ad might directly visit the website OR might google the brand name and click through to the site. All of these scenarios can still easily be accounted for in FouAnalytics.

In the 3 examples above -- automotive, financial services, and tourism -- the advertisers had TV campaigns running in specific markets (cities) and not in others. In most of these campaigns, the advertiser did NOT have UTM codes to work with since users saw the TV ad and came directly to the site or arrived on the site after "Googling" something. But even without the UTM codes we can still do attribution, using time and location (see REGION and CITY columns). Users in the markets where TV ads were running would come to the site, look at X-number of pageviews (yellow highlight), and then convert. Using a similar methodology as in the ecommerce section above, we isolate those sessions that ended in a conversion and track back to the users who were in the markets where TV advertising (or CTV advertising) was running. Note that this is correlation and not causation. But this is OK because advanced advertisers understand that one ad impression does not cause one conversion. But you can use the concept of conversion velocity to gauge the relative impact of advertising. Over the same period of time, if you divide the number of conversions by the number of users, you get the rate of conversions or "conversion velocity." By comparing markets in which you are running TV ads to ones where you are not running TV ads, you can see whether running TV ads increased conversion velocity and by how much.

Advertisers can also easily run tests using conversion velocity. For example, pausing advertising in specific markets, turning on advertising in other markets while noting the change in conversion velocity in each of those markets. Simple. Using conversion velocity to gauge the impact of advertising his helps advertisers break out of the bad habit of using cookies to attribute conversions AND also entirely avoid the forms of fraud that use cookies to falsify attribution (because fraudsters can no longer use cookies to claim credit for conversions they had nothing to do with).

This also solves the kind of fraud where remarketing vendors write false data into their own customers' Google Analytics, to make it appear the CTV ads purchased from them caused lots of conversions (when those conversions were actually organic and direct).

See: The Beautiful Illusion of Digital Marketing Success

Cookieless attribution for FMCG advertisers -- attentiveness

Does the above work for FMCG advertisers and others that don't sell anything online? Yes. How? Remember my article on attentiveness?

See: From Viewability to Attention to Attentiveness

Ads of all types (display, video, search, social, etc.) make people aware of your product or remind them to buy your product before they go to the offline store. Some of those people may still visit your site to look up nutrition information, allergen information, information on where to buy, etc. They visit your sites without clicking on an ad, so they appear to be "direct" traffic. They may also google something specific and click through to your site, so they appear to be "organic" search traffic. They may see your ad on Instagram, not click the ad itself, but visit your site at a later time to get more information. All of these scenarios can still be accounted for by using the "attentiveness" concept. With FouAnalytics on the website, we can first check if the users are human. Then we can look at their attentiveness while on the site -- e.g. did they move the mouse, scroll the page, touch the screen, click something, etc. Humans looking for more information usually do these things. Google Analytics, Adobe Analytics and legacy fraud verification vendors don't tell you these things. More attentive humans lead to more conversions, even offline. Is there going to be a one-to-one relationship between ads and visits and conversions? No. But there is "line of sight" between the attentiveness of the human user and future conversion events, even if those conversions cannot be "seen" online (because they occur in offline stores).

What can you do with the concept of "attentiveness"? You can use it to judge the relative quality of your marketing channels and tactics. In the grid above, with 5 examples, you can see how each of the sources (e.g. google search, bing search, facebook, PMAX, linkedin, display, etc) compare to "direct traffic" in terms of attentiveness, the components of which are valid-clicked:1, mousemove-exists:1, valid-scrolling:1, touch-exists:1, etc. If paid search ads leads to more attentive humans than display ads, spend more on paid search ads. This directly informs your marketing mix, in a way that MMM models do not (because those models use performance data like clicks without scrubbing for bot activity).

Does cookieless attribution answer the same business questions?

Anyway, this is a lot to chew on on a Saturday, especially if you're been using cookie-based attribution for years. But, let me know if the above makes sense. Challenge me on the thinking. Ask questions in comments. Share the post with other folks who might chime in, challenge, clarify the concepts. Let me know if you want to test this out and try doing cookieless attribution for your own campaigns with FouAnalytics. Will YOU use cookieless attribution to improve accuracy and eliminate affiliate fraud?

Does the cookieless attribution method described above answer the same business questions as the attribution models you use today?

Let me leave you with a final thought. Does the cookieless attribution described above answer the same business questions as the attribution models you use today? If your business questions include: 1) does my digital advertising drive conversions? 2) which channels drive more conversions? 3) how do I re-allocate spend across channels (i.e. marketing mix) to maximize conversions for the same overall budget? The cookieless attribution above should help answer these questions and provide you with the insights to optimize your marketing mix, get more conversions for the same budget, and reduce waste and unnecessary costs (e.g. paying out affiliate commissions unnecessarily).

Happy Saturday Y'all!

Further reading: 625 other articles by me on analytics, digital strategy, and ad fraud