This article was originally published in Martech Advisor
In the last five years, marketing attribution has risen quickly to prominence in the MarTech landscape. This is not surprising. Marketers now create and consume more data than ever. We advertise across myriad channels, with countless vendors, and round after round of creative. We target consumers with startling precision. We expect to be able to take that data, crunch it against those channels and consumers, and receive valuable, actionable insights that help us do more with our limited resources.
This is the promise of attribution: attribute credit where credit is due, so you can make better decisions. It goes beyond the capabilities of basic marketing measurement, like site-side analytics, and fills a gap that marketing mix modeling (MMM) can’t reach. With shorter timeframes and user- and channel-specific outputs, attribution delivers answers in weeks, days or even hourly, rather than quarterly or yearly.
Except when it doesn’t. Early attribution solutions took months – up to a year in some cases! – just to be deployed. Once they were up and running, some still only delivered insights a month after a campaign had run. Not only did that delay optimization, it frequently meant that the analysis wasn’t even valid. Media insights are highly perishable; the market moves too quickly for data from a month ago to offer any real value.
For most attribution practitioners, those days are happily behind us. But that doesn’t mean all methods are created equal. New attribution offerings seem to come to market daily and many companies choose to explore building something in house. Either way, it can be difficult to understand the nuances when they all claim to offer the same magic potion. With that in mind, here are nine things to consider to get started with an attribution initiative for your business.
1. What data model should you use? This is the big one. “Attribution” can be as simple as last-click – now largely outdated – or as complex as multi-algorithmic modeling. While the latter, also known as the Ensemble Method, sounds intimidating, it combines calculations from several popular algorithmic models for better accuracy and customization to your specific business objectives. Most solutions fall somewhere in between the two; if you choose a solution with just one algorithm, like logistic regression or game theory, make sure that it’s the right fit for your organization.
2. How long will it take to implement? As mentioned above, many solutions come with a long, steep learning curve. Attribution tools that are tough to deploy not only waste time, but are rarely used to their full potential, diminishing ROI. Avoid the shelf-ware curse with a solution that you can implement in six weeks or less and put the proper training in place so users get the most out of the insights that unified attribution can provide.
3. How fast is data delivery and analysis? Timeframes around attribution data insights can vary wildly from tool to tool and channel to channel. Some still turn around analysis in a matter of weeks, especially for TV, while others claim to operate in “real time.” Consider what timeframe is ideal for you to turn insights into optimization and ensure that the pace of the data matches the pace of your optimization.
4. Define your cross-channel requirements. Like “real time,” cross-channel insights are often promoted but rarely perform as advertised. A true understanding of inter- and intra-channel dynamics is critical to today’s marketing strategies. Attribution solutions should not only measure the appropriate channels, but also translate interactions between channels into actionable insights. If the results aren’t unified, you are looking at silo-ed data and you aren’t getting an accurate read on channel performance.
5. What type of optimization do you need? Different solutions specialize in different forms of optimization. Do you need to understand which vendors deliver more for your money, which channels drive the highest conversion rates, which creative performs best with your target audience, or all of the above? Ask for examples or seek out case studies of how industry peers use technology to optimize marketing spend, operate more efficiently, and scale more effectively.
6. How easy is it to use? Once the solution is live, it must be user-friendly. Yes, attribution is a complicated science. That doesn’t mean that the user interface should be. Attribution tools should be built for the people that use them. Have your media practitioners take the software for a test drive. If they can’t get in and out with the information they need in five minutes, move on or ask for a redesign.
7. How robust is the data? Simple to use shouldn’t mean simplistic. Your solution should serve up the most relevant information in a meaningful way, but also allow the full breadth of data to be extracted for deeper analysis. Consider how deep the data needs to go, and how it can be liberated from the system if need be.
8. Is your solution media agnostic? As the mar-tech industry moves toward consolidation, many attribution providers now report up into media partners like AOL and Google. This is an inherent conflict of interest and a classic example of walled gardens. Tread carefully.
9. Can your solution play well with the rest of your stack? Attribution is only one cog in the mar-tech machine. Both technology and business partnerships are important, from software integrations to relationships with agencies. Your ideal solution will be even more valuable if it collaborates with your other marketing technology, processes and partners.
Unified marketing attribution provides the insight that media-centric brands need to understand what’s working and what’s not. The right solution is less about bells and whistles and more about a good fit: does the attribution approach offer data and analysis that are useful and usable to your business? In the end, it often comes down to the Goldilocks effect. Attribution should not be too simple, or too complex, but just right for what you need it to do.
This article was originally published in Martech Advisor