VP Growth at OWOX
Attribution is the scariest beast to wake up in the world of marketing analytics. Only about half of companies in the US with more than 100 employees are using attribution modeling, but this number is expected to rise to 88% in 2020 according to eMarketer estimates.
Nearly 43% of businesses that have already applied attribution use the first click model, which is one of the simplest to implement. About 45% have tried the multi-touch model. But does the multi-touch model meet the expectations of marketers?
“Previous multi-touch attribution technologies and companies failed to deliver on the promise of generating value in the form of measurable incremental sales and profit through digital and paid social media optimization. For us to be able to successfully deploy and use such a platform, we had to be confident that the attribution impact on sales from the rest of our marketing and operational programming was integrated in a way that told us the true impact of our digital campaigns.”
Jon Francis, Senior Vice President of Starbucks
Source: The Next Generation of Multi-Touch Attribution
Half of those who have tried a multi-touch model have perceived gaps in the services and technologies they’ve used to implement it. But despite these challenges, companies who are trying to build attribution still develop faster than other companies and get insights that are unavailable to those who aren’t using attribution.
If you have no idea if you really need attribution, here’s a list of indicators that maybe you don’t need it right now:
Indicators that you’re ready to experiment with attribution are quite the opposite. If at least three or four of the points above don’t apply to you, then you should consider attribution. Remember that each business is unique and develops at its own pace. Attribution requires lots of resources to implement, so you should prioritize your tasks carefully. Let’s see how attribution can help companies who are ready for it.
Improving the conversion rate and decreasing spending on inefficient channels are perhaps the most convincing reasons to use attribution. Everybody expects that attribution will help them:
But assuming that attribution is a magic wand will definitely lead to disappointment. You should understand attribution modeling as a way of measuring. If you’re doing it correctly, you’ll get the right results. The main question is how to do it correctly.
A lack of preparation or having the wrong aims for attribution can result in your efforts leading to a dead end. The first step in preparing for attribution modeling is checking that you have all the necessary data:
This is all the data you could ever need for attribution. So it’s worth collecting it even if you aren’t performing attribution just yet. In order for your attribution modeling to be statistically reliable, you should build your model on:
In short, don’t start working with attribution till you have a lack of data or all your data is collected in separate databases and services.
But gathering data is only the first challenge. For your attribution modeling to be meaningful, you need a sufficient level of reliability. Because without proof of reliability, you’ll waste your money. Make sure you can reprocess data. This is necessary for correctly calculating returns. And in perfect conditions, you should ensure efficient integration with ad services to quickly optimize ad campaigns based on attribution results.
After you’re done preparing, it’s time to decide which model suits you best. This is a whole new topic that we’ll help you with in our next article. After you’ve chosen your attribution model, you can start performing the attribution modeling itself — and don’t be afraid to try as long as you need. It really takes time to get attribution just right. But the harder you fight for results, the more success you’ll have.
Try attribution yourself, and may your company prosper!