Understanding how to build an attribution model to better calculate marketing ROI
In the past, ROI of a marketing investment was calculated based on if sales increased after the investment was made. This calculation was inaccurate because it didn’t factor in other variables that contributed to the purchase. But today, marketers are able to track an individual’s various interactions with a brand to determine what influenced the purchase.
Attribution Modeling is the process whereby a company can assign credit to each online and offline touchpoint in the customer’s purchasing process to understand how to affected the final sale.
There are four stages to develop a useful attribution model.
Stage 1: Prepare your data
About: Collect all your data around touch points and outcomes into one repository. In this way, you can easily recognize patterns and apply simple rules-based models, which can lead to immediate results.
Case Study: A company used to collect all data in multiple repositories. After organizing all data into one location, they were able to allocate resources to each touch point as a direct function of its marginal ROI.
Stage 2: Experiment
About: Now it’s time to experiment and fine-tune the attribution rules. Once you are more comfortable with the rule-based model, you can start to identify the relationship between touch points. For instance, some touch points on their own aren’t as effective than when combined.
Case Study: An insurance company conducted regional experiments to test the synergy between their television, organic search and display advertisements. When they increased TV ads in different regions, they noticed a disproportionate increase in organic visits and display advertisement click throughs. This proved that they needed to better plan their marketing campaigns across media channels to truly create a better synergy.
Stage 3: Apply statistical models
About: Once you are more comfortable with the above two stages, you can begin to employ more sophisticated attribution models. Marketers can better understand which touch points to invest in, the effect of combined touch points and how much to invest in each one. These models can both explain and predict what is most beneficial.
Case Study: A retailer employed a multivariate regression analysis and Bayesian estimation to understand the effect of their targeted offline communications. The results indicated that they should communicate in all channels but after the first three months, they can decrease the rate of communications. This realization helped boost their offline communications by at least 10%.
Stage 4: Expand the scope of analysis
About: A customer’s purchase journey is not only affected by the company but by experiences that take place outside that journey (i.e out of time interactions). Using statistical methodologies like panel vector-autoregression, can help marketers understand these effects.
Case Study: A software company developed and employed a similar statistical methodology. They assessed the combination of offline advertisements (TV and radio) and digital media (branded and unbranded search, display, etc) through the panel vector-autoregression model. This analysis helped them understand that TV advertising helped boost the clicks coming through branded search. As a result, they increased investment in TV and improved the marketing ROI for the company.
Although it may seem daunting, attribution modeling can help marketers understand where to invest. Over time, marketers can navigate through the complex cause-and-effect environments to hone in on solutions that improve ROI for their companies.
*Source: Harvard Business Review
If you’re looking for ways to make the best of your marketing investment, don’t hesitate to reach out to ConnectWithUs@centralgrp.com