Three Reasons 'Good Enough' Analytics Is Bad For Business

With so much discussion about marketing attribution recently, agreeing on a definition for the term can be a cumbersome process these days. The IAB defines attribution as “the process of identifying a set of user actions (‘events’) that contribute in some manner to a desired outcome, and then assigning value to each of these events.” Econsultancy’s definition says, “Attribution is the practice of allocating partial value to different touch points within the customer journey influenced a sale or another desired outcome.” I like both of these for different reasons, but note both refer to crediting all touch points that contribute to conversion. Of course, within the attribution market there are also several different methodologies.

Here are three things you should know about the differences between rules-based and data-driven attribution.

#1 Simple attribution models are meant only for simple conversion paths. While simple, rules-based attribution models, such as last-click, have rapidly fallen out of favor among sophisticated marketers, there is a reason they were created. For starters, rules-based attribution models tend to be more accessible for advertisers with a modest number of monthly impressions and can be effective if the majority of conversions are single-channel.

These models offer some quick, surface-level answers about campaign performance; however, as the marketing mix becomes more complex,, they often lead to incomplete conclusions such as overstating the significance of bottom-of-the-funnel touch points while undervaluing the “introducer” touch points associated with brand awareness.



#2 Pre-determined weighted systems lead to arbitrary insights. One of the common scenarios where we see rules-based models “fail” is in identifying the influence between channels. Imagine a racecar driver taking the wheel while wearing blinders. If the race is a straight line on an empty track, the impact on his/her performance may be minimal, but introducing other cars or a turn makes it exponentially more difficult. As the marketing mix has grown in complexity, advertisers are now faced with a similar challenge. Intuitively, they know multiple touch points across various channels influence consumers in unison, but rules-based models aren’t able to “see” this impact clearly.

Because simple attribution models use predetermined rules to assign credit to every touch points, even advanced systems offer somewhat arbitrary conclusions. Without a detailed picture of the conversion sequence and relationship between each ad, marketers are left guessing how important each touch point was, relative to the others. Even worse, over time the data may lead to inaccurate conclusions about the relative value of specific ad inventory or channels, which negatively affects budgetary decisions.

#3 Data-driven attribution ensures impartiality. Conversely, data-driven attribution models rely on the data itself to determine how credit is assigned. By using a “bottom-up” approach that assesses data from every touch point, these algorithms compare converting and non-converting sequences to analyze how valuable each ad was within a particular channel. For example, let’s say we know the conversion rate for one set of customers who clicked a paid search ad for new sneakers before converting, and the conversion rate for another set of customers who saw a display ad for the sneakers first, then clicked the paid search ad before eventually converting. Comparing the conversion rates across these groups lets us isolate and assess the impact of the display activity.

Marketers can use these insights not only to understand true campaign performance over time, but also to run granular ROI analysis about each channel’s contribution, and plan more effective campaigns based on the role each channel is playing in the purchase funnel. Since all marketing data is considered, insights are comprehensive and impartial.

Parting Thought

Buyers should take the time to assess the methodologies of each vendor to make sure claims are substantiated with actual science. When you’re attempting to glean insights from copious amounts of data, the last thing you want in a partner is math that doesn’t add up.

Next story loading loading..