Last-Click Attribution: Media's version of Back to the Future

The Problem with “Last-Click Wins”

Rick Watrall, Partner

September, 2011

Hudson River Group “Last-Click Wins” White Paper

Background

The purpose of this paper is to discuss the issues with “Last-Click Wins” as a method for determining the business impact and ROI of online marketing. From the internet’s beginning, marketers hailed the advent of digital marketing as a Holy Grail delivering a promise of full “360 degree view” of the consumer.

The thought was that for the first time we would be able to get full accountability linking our marketing efforts to consumer behavior. Since theoretically a person’s digital footprints can be captured from exposure to conversion, it was reasoned a direct link between marketing and response (e.g. sales) could be achieved.

Over the years the technique that emerged as the primary attribution method was “Last-Click Wins”, where full credit was given to the final touchpoint preceding conversion. This means that no matter what other behavior the consumer engaged in prior to the conversion event (e.g. search on long-tail terms, viewing a display ad, exposure to traditional media) the impact was attributed to the last-click.

This technique has gained major acceptance since it’s easy to understand, is relatively trivial to measure (versus some methods such as First-Click – which suffers from issues such as cookie deletion) and is baked into most of the software applications designed to help marketers manage their online activities.

The Main Problem

There are a number of issues with Last-Click analysis in determining the absolute effect of marketing (and resultant metrics such as ROI) which will be detailed later, but perhaps none is as egregious as the fact that it neglects the concept of incrementality.

Anyone who has done even the most basic marketing analysis will recognize that for almost every business there is a “business as usual” or baseline level, above which marketing contributes. Some refer to this difference as “lift”, but the idea is a simple one; some degree of business results would happen even in the absence of marketing activity.

Last-Click is quite different in that 100% of the impact is ascribed to the marketing, completely ignoring this fundamentally accepted view of marketing’s effect as incremental. The degree to how much base business exists is dependent on many factors; but even highly marketing-driven companies where half the business is incremental would lead to a complete misread of its true impact under the Last-Click approach. For mature companies that are less dependent, the error is even more dramatic.

Adding to Last-Click’s acceptance as a measurement tool is the fact that some of the largest online tools for measuring performance (such as Google Analytics) are based on it, so people have been conditioned to think of it as the de-facto measurement standard and assume that 100% accountability has been achieved.

Taken as gospel, this has led to the results of Last-Click analyses as fact, and some have used this as a basis for establishing absolute metrics (such as ROIs) and assume they are directly comparable to other measurement methodologies (e.g. test/control or marketing mix modeling results).

What Others Are Saying

Recently, analytic/measurement experts and some marketers have questioned the validity of this method as illustrated here:

“Recently, there has been a spate of articles about how last-click attribution is a flawed way to measure online ad effectiveness for advertisers. Others have debated view vs. click attribution. But these debates miss a key point. The damage of misplaced attribution doesn't just fail to give credit where it's due for conversion, but also harms brands' online advertising efforts more widely - it hampers their ability to drive upper-funnel results.”
~ Richard Frankel - MediaPost

“However, we’ve been basing our decisions on the data provided to us through site analytics packages and for a long time now the industry standard measurement has been last click attribution. This method of attribution is widely regarded as a flawed metric due to its inability to take into account the many touch points that occur along the path to conversion and ultimately all end up playing a part in that conversion.”
~ Dave Freeman - Head of SEO at MPG Media

“While there is an enormous amount of data about online campaigns being collected, the current model used for attribution based on last action (i.e. Last-Click) does not always give us the clearest picture of where the credit should be directed.”
~ Brian Fitzpatrick - Managing Director, WPP

Basically, the key argument from the digital world (including those above) boils down to the point that there are many more touchpoints along the path a consumer makes before conversion, and Last-Click only represents the final event.

The issue here is that assigning full credit to the last touchpoint ignores all other factors that go into a consumer’s conversion decision, and this has the potential of over or understating the true impact. While that is true, it is only part of the problem.

Taking a more holistic view on measuring business performance and linking it to a company’s activities (in particular beyond digital) opens up a host of other issues. Digital is still typically only part of the equation, so in order to properly measure what is impacting the business we need to look beyond a siloed view that only considers the digital world. A broader perspective is needed to uncover the elements that truly drive our businesses.

Performance Measurement Checklist

The following checklist is a guideline for the characteristics a performance measurement system should have, and the arguments against Last-Click will be covered in the context of each:

• Measures incrementality, and does not merely “assign” credit
  As mentioned above, this may be the most severe strike against Last-Click. It is essentially a “100% assignment approach”, and it ignores that fact that a large degree of conversions would have happened anyway and that the final trigger (as defined by the last click) merely facilitated the event.
  Consider the example of someone who sees a display ad for a promotion with an online coupon for 10% off. Had the consumer been in the market already (i.e. if they already were going to convert before the offer), the resultant conversion would have been “subsidized.” Last-Click would assume the conversion was 100% incremental when the correct assessment would have yielded that it had no incremental effect.

• Considers all major forms of marketing, measured simultaneously
  One of the biggest issues with Last-Click is that it attempts to analyze the business in a vacuum. It is inherently a univariate analysis, and leaves out many other potential drivers (such as traditional media) which may be accounting for a large part of business performance. Like it or not, digital marketers have to accept that the majority of most major brands’ current budgets are centered outside of digital, and not recognizing that fact will tend to overstate the impact - especially when a “full credit” approach like Last-Click is taken.

• Methodology does not inherently favor one driver over another
  Most marketing analytics experts recognize that any measurement system that has an inherent bias toward one driver over others is problematic. An optimal system will treat “all its children equally”, and ideally consider all factors simultaneously. One of the issues seen with companies that have adopted a patch-work of measurement methods (including Last-Click approaches for online) is that the aggregation of all results across the different methodologies typically amounts to a multiple (2-3 times in some cases) of the actual size of the business. In essence, the methodologies are overstating the impacts, and when added together illustrate the problem that is hidden if only looking at each in isolation.

• Provides an understanding of and accounts for longer-term impact of marketing
  Marketing often has an effect beyond the immediate impact. This is pretty accepted for drivers like advertising, but even more “short-term” tools can have an impact beyond the time period they were designed to affect. For example, even promotional emails can have a “branding” effect well beyond the “call to action” window. These carry-over impacts can vary tremendously by driver, brand and even tactic; and ignoring them can underweight performance outcomes for marketing that works both short and long-term. Any measurement system that does not account for this dynamic will result in a marketing mix primarily based on factors that work more in the near term.

• Has the ability to determine the sensitivities around changes in investment levels
  Marketing effectiveness has been proven to vary by level of support. For example, at low levels there may be an increasing marginal effect while very high levels of investment may see saturation where added spending adds little or no return. Since the Last-Click methodology provides no link

© 2011 Hudson River Group. All rights reserved.

We’ve known for years (see below white paper, written around 15 years ago) that last-click is a tremendously invalid approach to measure media's impact, yet it persists.

Doing proper analytics has historically been complex and costly—and required expertise that was not widely available.

Moreover, last-click was pushed by certain platforms (which we all know were biased toward their media) and baked into their systems' reporting. And ease is a thing, especially if it supports an agenda that fits what you want to hear.

With the current wave of AI buzz, I believe that’s starting to change. Smarter tools could finally lower the barrier to doing data-driven marketing properly.

Still, we’re deep in the hype phase right now—I see it all over LinkedIn. My prediction? A lot of noise, a bit of chaos, and maybe some scorched earth before real progress begins.

Navigating that will bring some challenges. But at least the conversation about moving on may finally put some on the right path.