Measurement makes digital marketing work. Without it we’d be guessing, and that’s a poor strategy when money is on the line. If I was to tell you that “guessing” was the future measurement strategy from the advertising giants of Google, Facebook and Pinterest, how would that make you feel?

A bit nervous?

Let me quickly bring you up to date on the basics of measurement so you can see where the “guessing” part comes in.

Most measurement activity hinges on tracking “conversions”. Ideally, these are as close as possible to generating sales revenue, the next best thing being website quote requests, prospect phone calls and similar.  

There are two places where these are measured: a) within your website analytics tools, e.g., Google Analytics; and b) within the advertising software supplied by the channels you use – Google, Facebook and Pinterest.  

There’s rarely a 100% match between the two. Google Analytics might say Facebook advertising generated 15 quote requests, while Facebook’s advertising software says 25. The same gaps can appear within other channels too. There are two logical and well understood reasons why this can happen. 

One can track “viewers” of your advertising.    

Most advertising channels track prospects “viewing” your advertising (not clicking) and then proceeding to visit your website via other channels to complete a quote request (“conversion”). However, Google Analytics doesn’t have access to this view data. It can only allocate the source of the visitor from the website they were on before they arrived on yours. So, your Pinterest advertising reports might show 100 view-based conversions, but if those people left Pinterest and then used the natural part of Google to find your website, Google Analytics will allocate them to Google Organic traffic.  

One follows multiple journeys, the other only follows the last.  

Prospects may visit your website three or more times before they convert. Crediting these journeys to the one person – especially if they use multiple devices across multiple channels – is a challenge for Google Analytics. Therefore, it takes the simple route and allocates the channel to the visit that created the conversion – and ignores the rest.  

The result is that if your Facebook advertising brings a prospect to your website and they return again two days later through the natural part of Google and convert, once again, Google will be credited with the lead. Thankfully, Facebook’s advertising software is a bit smarter. It can track multiple visits that span a set time period (in Facebook’s case, say 7 days, Pinterest up to 60 days) and allocate conversions within this time.  

In both instances, the advertising software is providing some extra measurement smarts to help us better understand the true ROI of the ad spend. But what if this software started to include conversions that were based on “guessing” and not supported by actual human events?

Well guess what? That’s now happening, with Google, Facebook and Pinterest including “modelled” conversions in their reports.

Each channel has its own way of describing what they call a “modelled conversion” and most are filled with data speak along the lines of this, from Google:

“Modelled conversions use data that does not identify individual users to estimate conversions that Google is unable to observe directly. This can offer a more complete report of your conversions.

We model to recover slices of data where we know we cannot observe ad attribution due to protecting user privacy or technical limitations. We do this to provide high quality measurement so you accurately understand the impact of your marketing and maintain high quality bidding to prevent underbidding or overbidding.”

From Facebook, on the same subject:

“Our measurement and reporting tools empower businesses to assess the value of their marketing investments on Facebook platforms. We incorporate modelling so that we can better account for the total number of conversions that took place, including estimates for conversions we are not able to observe directly as a result of data limitations.”

And finally, from Pinterest:

As industry changes limit the availability of certain signals, modelled conversions are used to ensure that we can continue providing you with a comprehensive view of campaign performance. Modelled conversions utilize statistical techniques and machine learning models to predict the likelihood of your conversions being associated with people on Pinterest.

The gist is that the ad channels are having to guess when a conversion has occurred because of the widening gap between the data they used to be able to access for measurement, and the data they can now gather in a privacy-centric world.

I get it. The dramatic fall in tracking is making accurate measurement not just harder, but impossible. I’m just not sure how much I can trust the guessing game that each channel is using, when there’s so much money involved and no independent authority to validate the quality of their guessing.  

The best-case scenario is that the advertising channel models a conversion that did occur but was incorrectly attributed, for example, to Google Organic traffic. The worst case is that the conversion did not occur at all, across any channel.

We observed the worst-case scenario recently, with Facebook’s modelled conversions saying their advertising was responsible for delivering 25 sales of a particular product, when the client assured us they only sold 10 across the whole site. Not a good guess. By the way, there’s no way we can feed this data back into Facebook to let them know how wrong their guessing was.

So, what’s the way ahead?

Firstly, get used to it. This is the new, fuzzy land of measurement for digital marketing.

Try to create your own measurement strategies to fill the gaps created by these guessing games. For instance, as I mentioned in this article, start directing traffic from different advertising channels to different pages on your website. You could create a collection of pages for Google Ads, some for Facebook advertising and so on. This will let you allocate some real world lead numbers back to the channel that created them.  

Secondly, find out how much guessing is going on. You can do this by tracking what percentage of your total conversions are “modelled”, compared to those that are live. 

And finally, If you need help finding the data or working out how to tweak your digital marketing strategy to allow for modelled conversions, contact us today for a chat.