How To Use Google Analytics Conversion Segments To Attribute Product Listing Ad Performance

by Katerina Naddaf

Is a picture worth a thousand [key]words? In this post we explore how you can test whether Google’s Product Listing Ads (PLAs) are a better driver of conversion than Branded Paid Search.

This question was of particular interest to one of our clients. They wanted to know how these search campaigns were performing against one another to determine where best to invest ad spend.

Given the rising popularity of PLAs, and the many ways to reach a customer (prospecting and existing) through online advertising, we thought it would be helpful to share with you how this question was answered using the Top Conversions Path report, a default report in Google Analytics (GA).

What You’ll Need

  • Beginner to intermediate understanding of attribution reports in Google Analytics (selecting conversion types, interacting with conversion segments)
  • Paid Spend on a few different online marketing efforts (we’re advising on a few specific efforts to use in this example)
  • A solid understanding of your UTM tags and their meanings
  • A granular UTM strategy that allows you to separate the different marketing efforts

Before we dive in, let’s first cover some basics:

What Are PLAs?

PLAs are cost per click (CPC) ads enabled through AdWords and Google Merchant Center. When a product search is performed on Google, they appear on the top or right side of the search engine result pages (SERPs).

How Do PLAs Differ From Paid Search?

PLAs differ from traditional Paid Search ads in two main ways:

  1. They feature a product image rather than text.
  2. Bids for visibility on SERPs are made on products and product categories rather than keywords.

Now that we’re all on the same page, let’s dive into why this question came up and how it was answered.

Default Attribution Models

In our experience, most reporting tools and software don’t use sophisticated attribution models and are prone to “most recent touchpoint receives 100% credit” logic.

It’s very common for marketers to be faced with the question: “Should I believe [tool] when it tells me that [marketing method] is producing the most conversions?”. In general this decision is difficult, but in practice it does make sense to question these attribution models and play with the levers you have access to in order to explore optimized spend options.

While migrating your organization to using a new attribution model is a difficult thing to do, what we are doing is an ad hoc application of attribution models to help support a hypothesis and a test. This is attribution analysis in service of a specific marketing spend optimization, not developing a new model for overall performance measurement of all marketing channels.

Start With A Hypothesis

Our initial hypothesis was that the client’s ad management platform was giving credit to Branded Paid Search that could justifiably be given to PLA with their default last-touch models.

We knew our hypothesis was practical because, if we validated it, we would have an associated action. We would then be able to begin testing budget optimization by moving spend from Branded Paid Search to PLA placements within the ad platform and then reviewing the aggregate impact in Google Analytics.

Method Part I: The Logic

The logic of this analysis is as follows: If we assume that PLA interactions are a more valuable indication of interest than a Branded Paid search click, then we can conclude that some credit being given to Branded Paid Search would be better attributed to PLA whenever a PLA interaction preceded the Branded Paid Search interaction that received credit.

Google Analytics will allow us to analyze the proportion of cases wherein Branded Paid Search interactions occurred after PLA interactions in order to suggest a reasonable proportion of Branded Paid Search budget to re-allocate to PLA efforts.

The plain english version we got from the client is: “I think people who interacted with PLA and then searched my brand name were prone to purchase anyway, so I want to move some of my Branded Paid Search budget towards PLA if there’s a significant amount of cases where Branded Paid Search steals the credit in default reporting.”

Method Part II: Conversion Paths Reporting In Google Analytics

The Top Conversion Paths report under the Conversion report in Google Analytics was used to distinguish PLA and Branded Paid Search conversion paths.

The goal was to understand the proportion of conversions Branded Paid Search received credit for in default reporting, which were preceded by a PLA touchpoint. To do this we used Conversion Segments. Unlike their big brother “Segments”, Conversion Segments are only available in the attribution reports and receive much less attention in analysis blogs.

Conversion Segments

Two conversion segments were created.

The first segment isolated for conversion paths where a Branded Paid Search campaign was the last interaction before a conversion and PLA was an assisted interaction in the conversion. This was created to understand how Branded Paid Search initiatives delivered as the “closing” marketing touchpoint when PLA was a part of the conversion path.

The second segment captured how Branded Paid Search performed in isolation. The goal is to understand general performance of Branded Paid Search.

 How Much More Credit Should We Attribute To PLA?

Here is a simple calculation to determine the range of budget from Paid Search that is acceptable to allocate to PLA for hypothesis testing:

X ➗ Y = Z

Where

  • X is the total conversions where Branded Paid Search was the last interaction and PLA was an assisting touchpoint in the conversion path (Total Conversions in Segment 1).
  • And Y is the total conversions where Branded Paid Search was the last interaction (Total Conversions in Segment 2).
  • The range of 0 to Z is the potential value we could attribute to PLA which is being accredited to Branded Paid Search in default reporting.

For example:

Branded Paid Search Budget is $100

X = 4

Y = 20

4➗20 = 0.20

$100 x 0.20 = $20

A range anywhere from $0.1 to $20 is an acceptable range for testing the hypothesis.

Things To Consider Before You Try It Yourself

Reporting in Google Analytics is susceptible to sampling. This is especially true when employing conversion segments. A way to mitigate sampling is to reduce your date range for analysis. Sampled data or not, it’s always important to test out insights from any GA report before making any major changes to your marketing strategies.

What Happened To Our Client?

In the real life application of this method, the client moved the maximum amount of the acceptable range of budget and we saw an increase in overall transactions and return on ad spend when we isolated for all PLA, Branded Paid Search, and Organic Traffic.

And there you have it, a simple way to uncover how different marketing strategies are driving conversion to optimize ad spend.

For more ideas on getting the most out of your attribution data check out our post on Marketing Attribution Reports Your Executive Cares About. Or, give us a call and we’ll be happy to chat marketing attribution with you!

Katerina Naddaf

Analyst

As an Analyst at Napkyn Analytics, Katerina works with our largest customers to help them understand the performance of their marketing efforts and ecommerce stores.