It’s 7:30 a.m. on a bright, chilly Thursday. Gwen, the EVP of marketing for a large, regional bank settles in at her desk, coffee in hand. Ready to attack the day, she cracks open her laptop and fires up her email. A report from the VP of marketing analytics catches her eye and she clicks on it.
Her brow furrows as she reads. Once again, the team is recommending a shift in media mix for checking products — away from upper-funnel channels to lower-funnel workhorses — email and paid search. The attribution model that her team is using shows that the bottom-of-the-funnel tactics are highly effective at generating significant incremental deposits with an attractive return on investment. But Gwen’s gut says that too much of the budget is already allocated to this type of direct-response media.
She fires back an email to the team, asking for a deep dive.
Gwen is stuck in a classic tail-wagging-the-dog scenario. Marketers have increasingly powerful measurement and attribution tools at their disposal, which is a good thing. These tools, however, come with built-in assumptions that make them anything but neutral. If marketers aren’t vigilant about the inherent biases of various marketing-optimization strategies, they could fall into one of these common traps.
Here’s how to avoid them:1Assume Outputs Are Neutral
Different attribution approaches will naturally push the business to select specific channels. That’s why it’s critical to understand the impact of different model types on the outcome before committing to an approach. For example, last-touch models will prioritize lower-funnel tactics. Marketing mix models tend to only include direct mail and most digital channels in a cursory way. This means the outputs provide insight on the effectiveness of the largest traditional media channels. Beyond that, they are less reliable. 2Eliminate Channels Without Understanding Relative Performance
Just because your spending is inefficient does not necessarily mean the channel itself isn’t efficient. In fact, marketing spending data analyzed by Novantas indicate some channels are highly efficient for all but a few outliers. Without reliable (and relevant) points of comparison, you won’t know if your underperformance is driven by marketing execution problems, competitive dynamics or the inherent inefficiency of the channel. 3Insufficient Context
Attribution models are useful at indicating where you will get the highest return from your next marketing dollar. But this model isn’t always right. For example, if your model indicates that it’s inefficient from a cost-per-acquisition standpoint to invest in a particular market even though the bank has a much lower fixed-cost base there (fewer branches and ATMs), it might make sense to tolerate a higher CPA in that market because the total cost is lower. 4Generalizations
When it comes to media mix, ROI isn’t the only number that you need to keep in mind. Digital advertising might be only 35% as impactful as TV in building awareness for a new checking proposition, for example, but internet ads might be four times more impactful than TV in driving CD sales. A relentless focus on efficiency can cause significant waste on cheap impressions that deliver little or no impact for a particular product. Marketers need to stay away from blanket conclusions and consider the implications at a more granular level. 5Lose Sight of Quality
Driving efficient acquisition is generally a good thing. However, calibrating your CPA performance against a quality metric such as cost-per-retained dollar (CPRD) is an essential step. Filling your funnel with cheap conversions that drive limited long-term value keeps you stuck on a treadmill to nowhere.
BANK MARKETERS have more powerful tools at their disposal than ever before. But just like artificial intelligence can’t be used in a vacuum, attribution models are really just a starting point and any conclusions must be applied with caution. Human insight and experience will always be valuable companions to technological outputs.
Director, New York