Proxy World: The Need to Evolve How We Measure Success

patrick-jones-datalogix“Data Driven Thinking” is written by members of the media community and contains fresh ideas on the digital revolution in media.

Today’s column is written by Patrick Jones, head of Facebook partnership at Datalogix.

Earlier this year I had a meeting with a major brand’s head of digital to explain that it was now possible to tie that brand’s online ad spend back to its sales transaction data. The marketer’s response? “That is amazing; I can see huge value in doing that. However, I am tasked with driving leads, not sales. I need solutions that will drive me X amount of online leads this month.”

I shouldn’t have been surprised. We have worked our way down the funnel in this fashion: A click is a proxy for interest in my ad, a visit is a proxy for interest in my company, a lead is a proxy for interest in my product, and so forth. We’ve even created proxies for our proxies. For example, because a lead doesn’t always equal a sale, advertisers assess their average lead value and apply a lead-to-conversion ratio to calculate their expected revenue from their total leads. Using this logic, advertisers are creating a revenue calculation based on assumptive calculations instead of real concrete data.

This proxy creation has been happening in siloes, resulting in each industry vertical developing their own set of proxies as standards. CPG uses engagements, fans, microsite visits, store-locater usage, “in-target” impressions, and other measurements as proxies for actual in-store purchases. The automotive vertical uses dealer-locator visits, build-a-model usage, and request-a-quote forms in lieu of car sales. We have created proxies for nearly everything except for goods that are actually sold online and only online.

Of course I’m simplifying this to make my point, but here’s why all this matters: We have now come to a time where it is possible to eliminate these proxies, and yet instead of switching to measuring what really matters — sales — we only continue to add more proxies and build processes and standards around them.

Entire sub-industries have been built on the back of this proxy world. Hundreds of companies thrive off the ability to drive engagements, “conversions” (the word conversion is even used interchangeably with proxy metrics that are not dollars spent), or fans. As we have discovered problems with this model, instead of re-evaluating the model itself, we’ve built new solutions to patch the holes.

Here’s one example: Because we measure view-based conversions, some people started gaming the system and running ads that loaded (and therefore were considered as “viewed” by an adserver), but were never seen by the user. The thinking goes that it’s cheaper to run more ads at the bottom of pages than to buy premium content — more ads equals more ad exposures and an increased likelihood of picking up a view-based conversion credit. In reality, of course, unseen ads don’t increase sales.

To help fix this problem, we have created a new metric called “in-view,” and companies have created new products to help clients determine what percentage of their ads were actually seen by a user, rather than just loaded onto the page. That’s great, but the real lesson here should be that it’s time to evaluate the usefulness of view-based conversions as a proxy.  Why are we spending money to improve view-based metrics instead of focusing swapping out a “conversion” for an actual sale?

Meanwhile, analytics companies have begun building models to help determine what views or engagements had the most impact and how to attribute value within an ad-exposure path. Again, this is a great use of math to help patch a hole in the model. But the fix just ignores the bigger leak. Even if you figure out the perfect attribution between all the exposures and engagements leading to a user becoming a fan or requesting a quote, that person may not have spent a dime on the advertiser’s products. So how much is that information really worth? If the ad-exposure path on which the attribution model is based leads to a proxy and not an actual sale, the whole concept breaks down.

And it doesn’t have to be this way. It’s now possible to bring offline data online. If an advertiser owns the point-of-sale data, it can be tied to ad exposures and used to measure actual sales transactions. Even in the case of the CPG, in which advertisers often don’t own their own sales data, this is possible. Offline transaction data from some of the largest CPG retailers in the world is already being matched to online ads for measurement.

If measuring against actual revenue and sales is now possible — even when products aren’t sold online — why are advertisers investing so much in driving and analyzing engagements, conversions, and other online proxy metrics? Don’t get me wrong, there’s definitely value in consumer interactions outside of purchases. But now that true sales data is available, it no longer makes sense to use those interactions as proxies to calculating ROI.

Why are we hanging on to all of this? Is it simply because too many people have built their careers around these proxy metrics? Is it because too many people have agreed to business goals focused on driving engagements, leads, and quotes? Is it because driving any of these proxies is significantly easier than driving true sales?

Whatever the reason, it’s something we need to fix. If we continue to build more and more solutions on top of the broken model of online proxy metrics instead of focusing on sales, we’re all going to be in trouble. We need to hold our marketing efforts to a higher standard.

The industry as a whole will benefit from a paradigm shift toward better metrics and true ROI. If we accept the challenge, we can only expect further growth in digital advertising spend.

Enjoying this content?

Sign up to be an AdExchanger Member today and get unlimited access to articles like this, plus proprietary data and research, conference discounts, on-demand access to event content, and more!

Join Today!


  1. Patrick – thanks for the interesting column. What tool are you using to track ad engagement back to the POS? I’m trying to get an understanding of whether or not you’ve seen this work for someone spending $10k-$30k per month, or if it requires custom POS integrations, larger spends, etc.


  2. Patrick


    Thank you. I currently work for a company called Datalogix that enables measurement back to POS transaction data. There are other tools in the space as well that clients can utilize.

    There are a few things at play when measuring back to sales: feasibility and scalability. Feasibility is how we determine statistical significance as well as take the pre-campaign steps needed for creation of a forensic control group (a control group of users who have the same demographic traits, purchase patterns, geo, etc… for a 12 month period pre ad exposure). Scalability is where we take a look at your target audience and determine if there are enough users in that audience as well as what is the minimum number of unique users that need to be exposed to meet measurement minimums.

    In short, as long as you are taking the necessary steps to ensure your methodology and campaign setup are correct, it is possible to run and measure a campaign at your scale. Datalogix has clients who spend in that range and have driven successful results when measured back to ROI. Optimal exposure frequency is actually an area we are researching now so that we can also help our clients understand exactly how much they should be spending and when they hit the inflection point and ROI starts to shrink.