Media Planning & Buying Mobile Marketing

From Attribution to Optimization: Interactive Marketing Mix Systems

April 30, 2015

One of the underlying goals of any marketing effort is to set and spend within budget in the most efficient way possible, defining success metrics along the way. There has been much debate over cross-channel attribution and the importance of choosing the right multi-channel marketing mix optimization strategy. The use of revenue attribution systems alone do not provide the answers marketers are looking for, and in many cases give the wrong answer. During any ad campaign planning or analysis, advertisers should view media channels holistically and use advanced marketing mix systems to optimize budgets, messaging and tactics. In the following post we dig into why marketers should be flexible in their thinking on attribution.

Attribution: A Relic of the Past

To understand which channels or tactics generate the corresponding credit for which portions of the revenue, digital marketers in the past have used a revenue attribution system – but this method has now become a relic of the past. 360i’s SVP of Advanced Analytics and Decision Sciences, Kevin Geraghty, made the case in an article for AdExchanger that any attribution scheme that is used (e.g., first-click, last-click, equal or weighted distribution, dynamic distribution, etc.) will lead to inefficient budget allocation and lost opportunities.

We pose the question: During your next trip to Disney World in Orlando, which of the car’s six cylinders gets the most credit? To most people this is a nonsensical question to ask. The important metrics are (A) how much gas you put in (cost) and (B) how many miles you have traveled (return). The car’s internal computers will set those cylinders to optimize the overall performance, but the metric “miles per gallon for each cylinder” would not be defined.

Similarly, advertising channels do not live in a vacuum – they all work together (or against each other in some cases). There is no reason, for example, to separate display advertising from paid search when budgets are optimized. There should be holistic measure of success (e.g. advertising ROI, CPA, etc.). Thus any optimization scheme based on ROI of individual channels will produce subpar performance, which is why marketers must change the paradigm from attribution to marketing mix modeling.

From Attribution to Marketing Mix Modeling

Marketers should not rely on attribution modeling alone when it comes to budget optimization. There are two main approaches to budget optimization which are used in the industry by vendors such as Adometry, MarketShare, Visual IQ, Convertro, Abakus, etc. These are “bottom up” and “top down” approaches.

The Bottom up method relies on cookies and device tracking, customer-relationship-management (CRM), and other user level data to recreate (or simulate) individual converting and non-converting paths.

In the modern connected world, many vendors tend to create models by connecting exposure paths for each individual consumer. It is a noble idea, but even in the digital ecosystem, the method fails to account for a large part of customers journeys. Due to the nature of tracking, many paths can be lost, an issue largely due to the usage of multiple devices by the same user or use of the same device by multiple users, cookie deletion issues, difficulty tracking online and offline interaction, in earned media and walled gardens where tracking systems may not be allowed (e.g. search views).

In addition, it is very easy for aggressive publishers to manipulate the system. Top down approaches use statistical methods on aggregated data to infer the relationships between different channels, tactics, etc. The main drawbacks of this approach are poor handling of coincidences (e.g. increased seasonal spends in multiple channels do not signal interaction) and inability to correctly determine the mechanics of channel interactions that lead to incorrect recommendations. For example, improvements in performance require a different response than increase in demand.

Deploying an Interactive Marketing Mix System

As noted, both bottoms up and top down methods have respective drawbacks. We recommend marketers use interactive marketing mix modeling, combining both techniques with mandatory testing thrown in to budget the most efficiently. Marketers should use consumer level data when available, append the results with regression analysis on aggregated data when tracking is not reliable or subject to manipulation, and always (especially on the cookie level data) use multivariate testing to determine true interaction between different channels, tactics, etc.

The last item is the most crucial piece – without the feedback loop that continuous testing provides, the statistical modeling and/or customer tracking can (and will) lead to lost opportunities and wasted media dollars. By changing the paradigm from attribution to optimization, advertisers will be able to allocate media dollars in the most efficient ways.

Cover photo via Pinterest.