Have you ever looked at your AdWords or Facebook reporting to see lots of conversions, feeling great about your hard work, only only to be surprised that the actual results aren't nearly as good? Where did all those conversions go? We've been there too, it's scary. Let's take a look how this happens and what you can do about it.
This guide will cover the fundamentals of how ad network conversion pixels work and why they can often show you misleadingly optimistic results.

Under the Hood

Nearly all ad networks support a 'conversion pixel' that they use to show you conversion events that their traffic drove (Facebook, AdWords, Twitter). These conversion pixels are designed help show marketers the results of their investment. Each network has a time frame (aka window) where they will honor conversions, usually 30 days.

Network metrics 0

Here's how it works:

  • 1. Visitors see an ad
  • 2. Visitors click on the add and a cookie is set
  • 3. Visitor converts and the conversion pixel is reported back to the network
  • 4. Visitor arrives at your landing page
  • 5. When the visitor converts, the ad network receives the pixel and reports it as a conversion

Trusting Pixels

The conversion pixel functionality is reasonable at first glance: the networks are helping their customers understand the results their paid traffic caused. They are justifying the marketers spend on the network. Unfortunately each networks pixel is showing only a small portion of the full story behind each conversion.

Let's take a look at two hypothetical campaigns on Facebook and Google AdWords. Here's a summary of the ad network-provided conversion reporting:

Network metrics 1

The conclusion from this reporting is that both campaigns are doing great! They both show solid returns and it seems like both campaigns should get increased budgets to keep building the business.

The Problem with Network Pixels

The problem with pixels is that they are very greedy: a single conversion can be reported by many pixels. This can cause tons of double-counting if your audience interacts with multiple campaigns, which is very, very common. Let's take a look:

Network metrics 2

After investigating the results we found 4 conversions that clicked on both ad networks and were reported twice: once to Facebook and once to AdWords. Without credit for those 4 conversions, both networks are ROI negative and the entire effort has lost money. The conclusion is wildly different now: focus on optimizing the campaigns and funnel before expanding the budgets. We also have these 4 conversions that we need to handle, but which network deserves credit?

Using Attribution Modeling to De-Duplicate Conversions

Instead of combining metrics directly from each network, let's see what happens when we use an independent service to run attribution models that give partial credits to each network. This way the network pixels are not used and conversions are tracked independently:

Network metrics 3

In this case, you can see each network gets credit for their original conversions and shares partial credit for the four shared conversions. In this case, Facebook got credit for another 0.5 conversion and AdWords got credit for 3.5 more conversions.

Now that we've accounted for every conversion and given partial credit back to the networks the results look positive again. Each network has a positive return and can be scaled up, but they don't look nearly as amazing as the did when we looked at the network metrics in isolation.

The networks aren't lying to you, they just aren't telling the whole story. Their intent is simply to justify their value and they end up over-reaching a bit. It's worth noting that the rest of the network data is perfectly valid to use: clicks, ad views, eCPC, cost, geo reporting, etc.

Crediting Ad Networks

You might be wondering, how we knew to give Facebook another 0.5 conversions and AdWords another 2.5. This is just a fictional example, but the results are common. Credit is assigned by an attribution model, a standardized definition of how to give credit to various channels that a user encounters. For example, they may say that only the last paid ad network they visited before purchasing gets all the credit.

There are lots of various attributino models for different types of busineses. For more details read more about attribution models and how they work.

TL;DR: Don't use Network Conversion Pixels to Justify Spend

Ad Network pixels are intended only to tell you which conversions clicked on an Ad. Don't use them to justify returns from that network, as these conversions can be quickly double-counted. Combining conversion events from ad networks is both time consuming and dangerously optimistic. Instead, use an independent 3rd party to measure your results.

Attribution modeling can quickly show you which traffic sources, networks or campaigns are performing best, without complicated hours in Excel, without inflating conversions and without complicated setup.

Now What?

Ready to get started with attribution modeling?

Start Free Get a demo