The Attribution Crystal Ball: How to Create Transparency for Your Ad Spend


Q: What are common attribution models and why did they become so widely used?

A: The majority of attribution models can be split into two categories: single-touch or multi-touch.

In a single-touch model, one specific ad event such as last-touch is attributed with 100% of the conversion value regardless of how many sequential ad events preceded it. This method became popular as it can be set up relatively easily using UTM tags and measured in Google Analytics. This model is problematic though, as it ignores all assisted events in the customer journey and can lead to a skewed understanding of performance.

Conversely, a multi-touch model such as position or time based attribution still attributes a larger proportion to the last ad event before conversion, but also accounts for all other ad events within the campaign. This model allows for better transparency into the effectiveness of each ad event and creative, as well as allowing marketers to make better decisions with their ad dollars.

Q: What are the risks of employing an overly simplified approach to attribution?

Many attribution models don’t tell the whole story even when focused on conversion. One common mistake in the realm of attribution occurs when too much emphasis is placed on click conversion while ignoring or heavily discounting view-through conversions. While the click represents an easily quantified and measurable action, studies have demonstrated that only 16% of end users ever click on ads, and that half of that group accounts for nearly 85% of all ad clicks.

So optimizing for clicks doesn’t necessarily make the most sense for a performance oriented ad campaign where the goal is effectively driving conversions. The majority of consumers need to evaluate an offer before converting. For example, they might see a compelling video ad for a new product and then search for reviews of it or ask their friends for feedback before purchasing. Of course it’s key to continually re-engage them with sequential messaging during the evaluation process, but even then it’s entirely likely that even though the campaign was effective, they will never click on an ad.

Q: What are the components of a data-driven view of the path to purchase?

A: It’s key to build rich audience data that modifies how a campaign’s delivery is weighted in-flight. A truly data-driven approach measures the way users interact with ad units and looks at how that engagement lines up with post-click metrics like average order value or return on ad spend. Those factors then correlate with granular user-level attributes like location based weather, ad placement type, number of ad units seen and time of day.

It’s also important to employ an attribution model that allows you to identify how ad events assist one another through the customer journey. It makes the most sense to use a closed-loop approach that captures both click and view-through based conversions and assigns value to each.

Q: Why should marketers challenge the use of metrics like impressions, viewability, clicks and more?

A: The problem with these metrics is that marketers rely far too heavily on them as principal KPIs when they do not drive any measurable business value. If you use clicks as your main metric by which you measure the value of a campaign, you may see a high click-through rate and determine that the campaign was successful. However, if that same campaign had a conversion rate of 0% – or conversions were far below the site’s average order value – then those clicks were actually very low value.

Similarly, impressions are a fine metric to report on, but fail to provide any qualitative data. At first glance, viewability seems like it could provide some level of qualitative data – but our data shows a statistically insignificant correlation between viewability and conversion rate, so it’s likely not the best metric to optimize your campaign around.

Q: What post-click metrics work and why?

The best post-click metrics are grouped around lead or conversion actions and the qualitative data associated with each. For example, if there was a campaign whose goal was to drive purchases on an ecommerce site, you may want to look at post-click events like ‘Add To Cart’, ‘Add To Wishlist’, account creation, newsletter signup, checkout flow progression and if the sale was completed. Then you would score each event against metrics like average order value, return on ad spend, margin on ad spend and customer acquisition cost. These metrics allow us to qualify the value of each conversion and correlate lead actions with high value customers. All this data is extremely important as it is, but it then can be used to optimize both prospecting and re-engagement portions of the campaign and deliver a far more effective ad buys for the advertiser.


Please enter your comment!
Please enter your name here