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.
Realistic and unrealistic expectations of 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:
- You don’t do business both online and offline at multiple points of sale.
- You identify your customers across only a few online or offline channels.
- Your metrics on transitions and microconversions are still in development.
- You don’t include real revenue numbers in your calculations.
- Your marketing strategy still needs improvements like including all promotional activities or analytics set up.
- Your ad campaigns are run by outsourcing agencies without a unified standard for gathering data and estimating efficiency.
- You don’t have a separate paid advertising budget.
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:
- lower their marketing expenses
- make their whole marketing strategy smarter
- apply all marketing data from every channel
- move resources from inefficient channels to efficient ones
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:
- Сomprehensive data from all the channels you use, without limitations and sampling. Anything that can be measured should be measured. And the data must be gathered all in one place, organized, and ready to use. That’s why companies typically use in-house solutions for attribution experiments.
- Historical and real-time data. As marketing modeling is based on different mathematical and statistical methods, you should use all the data you have to increase your statistical accuracy. The more data you have, the more reliable your attribution results will be.
- A data set that excludes data from peak periods. Remove data that reflects seasonal preferences, possible errors in the analytics system setup, one-off advertising events, technical errors, and big holidays.
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.
To sum up
- Prepare for attribution thoroughly. It’s never too soon to start preparing your data for future attribution experiments by collecting and storing it properly.
- Pay attention to the model you choose for attribution. It should fit your business or else it will drain your budget even faster than before.
- Avoiding attribution won’t save you money. But getting attribution insights for your business is worth all the expenses.
Try attribution yourself, and may your company prosper!