Plant, Water, and Harvest: 7 Rules for Media Attribution
The following first appeared in Response Magazine, December 2015. Some edits have been made (original article link here).
The idea of highly accurate attribution is all the rage. And it should be. Data science today makes it possible to understand more about sophisticated media mixes and their impact on sales channels.
But attribution also remains severely limited. Advertising does more than merely close sales. Advertising introduces consumers to the product or service, moves them from mere awareness to wanting the product, and then finally motivates them to act. Let’s call it the “plant, water, harvest” model.
This model leads us to seven rules for thriving in the era of attribution:
- Successful attribution requires you to model a plant, water, and harvest approach. If you stop planting, you’ll end up without a harvest. And if you stop watering, your plants die. In marketing, this means paying attention to the entire journey that leads someone to buy your product or service — not just that point where they take action. Attribution should help you understand the entire journey.
- Never rely only on “last-touch” attribution. Many attribution models only consider the media that drove final contact. That’s because last touch is far easier to calculate and track than the more important steps. But if you make important decisions based too heavily on last touch you’ll hurt your ad impact. The most effective last-touch media is often or usually quite poor at either planting or watering.
- Avoid black boxes. Many vendors and agencies claim they have black boxes that reveal everything you need to know. They don’t. Even worse, black boxes tend to make huge hidden assumptions that can damage your business. But you won’t ever know the damage because the assumptions are hidden from you.
- Statistics underestimate the importance of planting and watering. About 15 years ago while investigating major retail attribution, my agency found shocking contradictions in the data. Statistical analysis estimated one retail sale for each direct sale. But consumer surveys and sales data showed as many as 14-15 sold at retail per direct sale. Why the difference? In this typically noisy market, statistics couldn’t detect the impact of the media’s critical role planting and watering.
- Statistical models are developed for ideal situations. But markets aren’t ideal. Markets are noisy. And in a noisy market, statistics tend to overestimate the most obvious things — like impact from high-profile media airings. That means they miss the impact of important smaller actions.
- Consider television. Statistically, one large primetime TV spot airing is easily spotted by statistical analysis and, therefore, receives excellent attribution numbers.
- Yet a group of 30-40 smaller airings might spend the same money and drive far more impact per dollar. Statistical analysis will have a hard time detecting the impact these smaller airings individually or as a group.
- The result is that models often lead media teams to emphasize high profile, large budget items just because they are big – not because they are effective. I’ve seen this cause media buyers to shift huge budgets into what’s actually the most ineffective airings for that campaign.
- Don’t ask consumers to do your attribution for you. Many companies build attribution around asking consumers to enter promo codes. Except, no matter how hard you try, 70-80 percent of your consumers won’t use the codes. Why? Attribution is not their job — it’s our job.
- Avoid separate attribution silos. Separating silos can be as misleading as black boxes. For example, quite often the “cost per action” for paid search is exceptionally low compared with cost per action for other mediums, like TV. The problem is: paid search works because of the TV ads. Reduce your TV spending because it appears less efficient and soon (perhaps as long as a few months) your search will suffer badly.
- One way to make attribution more accurate is to allocate a portion of TV spending (for example) to paid search and then do the calculations. It’s a simple choice even though the exact % to allocate is tricky to estimate. Because it’s complicated, many companies decide not to take a holistic look at their media and end up making mistakes that reduce the impact of their budgets.
So go forth and build attribution models. But remember, there’s no magic box. The most successful companies use attribution to complement traditional research and analysis. And, business success requires human judgment and instinct — that’s what makes it such fun.
And the good news is that when smart attribution is combined with more traditional work, we’re able to make amazing things happen.
Copyright 2015 – Doug Garnett – All Rights Reserved
Categories: Big Data and Technology, Business and Strategy, Communication, consumer marketing, Digital/On-line, Direct Response, Media, Research & Attribution, Retail marketing, Technology Advertising
Posted: June 16, 2016 20:08
New Language: Attribution or Contribution for Advertising Impact? | Doug Garnett’s Atomic Blog