Performance vs. Proxies and How Conversion Data Fits In

By Paul Cannon
Media Center of Excellence 

As we get closer to the end of the year, all of us are about to see a bevy of holiday-themed ads come our way, as brands and their agencies get ready to launch their holiday campaigns. Whether it is retailer ads about the hottest gifts or food product ads for upcoming holiday meals (I hope someone does a fun fruitcake ad this year), the advertiser’s arguably favorite season is almost upon us.

While some of the ads will hit their mark – reminding us about particular items to add to our giving list, or whetting our appetites for a new food (perhaps even reminding us our homes need a deep cleaning before guests arrive!), many of these holiday campaigns will fall short of their goals.

And many of the advertisers will not even know it.


In short, they are not using the right metrics to prove whether or not those campaigns did what they were supposed to.

Let me explain.

When it comes to CPG products, at least, it’s well established that a past purchase is a good indicator of a future one. For example, if you buy a certain kind of bread or soap, it’s a safe bet that you will buy it again, especially if you have bought it in the past.

But many advertisers don’t connect ad viewing data to sales data to know if their ads made a difference in the place where it most matters – at the checkout.  

Many marketers and their agencies are basing success on what those of us in the industry often call proxy metrics – clicks, interactions, reach, impressions, etc. – rather than performance metrics, such as sales. With so many ad dollars spent with the ultimate goal of increasing sales, marketers need to quickly know that they are getting good returns and where those returns are coming from.

Just think of how you could potentially improve your campaigns if you had that performance information as the campaign was still running. You might make some different choices.

Crazy, right?

Of course, measuring sales while a campaign is underway has historically been difficult and/or impractical for many businesses. This has been due to a lag time in data availability or the data was too complex to work with and/or it was expensive to acquire and manage large data sets. 

The ad industry responded to the challenge by introducing a variety of proxy metrics. While far less accurate, these proxy metrics are easier, faster and more efficient to gather and analyze. Unfortunately, it’s also true that proxy metrics vary widely in their correlation to sales and don’t accurately predict results. This impacts the quality of business decisions and, ultimately, business outcomes.

By focusing on performance, such as sales, you can also begin to solve some of the broader issues related to media quality. Bots don’t buy products, and ads that are not viewed by humans can’t produce sales. If you are measuring sales as the ultimate indicator of campaign performance, then the bot or fake computer program clicking your ad won’t matter. Bad players ultimately get squeezed out. 

Today, many offline sales lift studies still happen after the campaign has ended.  Although they provide great information about overall campaign success and some insights around what to do for the next campaign, measurement isn’t a static one-time event. It’s an ongoing process that helps advertisers make improvements along the way.

In-flight measurement and optimization solutions that leverage offline sales data are newer to the industry, but they can provide the necessary insights around which tactics are driving sales early enough in a campaign and far enough upstream that they can provide marketers the opportunity to maximize sales driven by the campaign.

Here are some tried and true examples:

  • Fuel programmatic in-flight optimization via sales data versus awareness metrics. By supplying your DSP with a weekly data feed of all the households that have purchased the advertised product during the last week, you can optimize your campaigns based on actual conversion. This not only helps shift ad impressions to your desired consumers but also eliminates ad spending waste since now you won’t be serving up ads to households that have just purchased. 
  • When samples sizes are small and or measurement is needed early within a campaign, Sales Index provides a lighter alternative to Lift. While it provides a test and control audience, it is a non-modeled approach, providing the directional trend of campaign performance. Think of it more like a thumbs up or thumbs down signal that provides insights for overall campaign direction and which tactics are aligning to converted households.
  • Use a fully modeled measurement and optimization technique. This has the most statistical rigor and granular tactical results. Historically, because of the limited size of offline conversion data sets, sales lift analysis has typically only been possible at the end of a campaign due to the feasibility needed to generate results that statistical models deem as confident. But, today, offline sales conversion data is available at scale, so marketers can get the best data to drive campaign optimization and measurement decisions.

No matter how big or small your campaign budget is (or the measurement portion of that budget), there are solutions available in the market to meet your ongoing measurement and optimization needs. Making this approach a standard part of your ongoing marketing business processes helps advertisers make more data-driven decisions and change the trajectory of in-flight campaigns towards a more positive outcome.

Now that we’ve discussed that, who’s ready for a little holiday fruitcake?