How your multi-channel strategy takes the struggle out of proving marketing’s true impact.
If you boil down the marketing mix[amsive_tooltip term=”marketing-mix”], you’re left with how a customer came to know and buy a product or service. It’s the job of the marketer to read the multi-channel tea leaves and relay to the client just how that journey happened from start to finish.
The buzzed-about phrase of the moment that describes all these ingredients is marketing attribution[amsive_tooltip term=”attribution”], or the way that marketing professionals assess ROI[amsive_tooltip term=”return-on-investment”] in channels that connect to potential customers. The ability for attribution to illuminate true impact is an integral part of the future of digital marketing, and to crack the code means huge dividends for organizations worldwide.
To get a better sense of how attribution connects to marketing efforts, we spoke to Amsive’s VP of Data Analytics, Ray Owens.
Amsive: Between last-touch, first-touch, and linear-touch, there are so many different approaches to attribution floating around — where do you fall in this debate?
Ray Owens: We found when you’re putting out multiple marketing messages, plus different ad rotations happening with creative execution, those heuristic methods — first-touch, last-touch — produce a kind of false-positive indicator for actual path-to-conversion[amsive_tooltip term=”conversion-path”].
Certain consumer exposures to a particular ad may get the highest ROI simply because it was a first-touch KPI. We initially observed this as a go-to method for many of our urgent care clients because it’s such a needs-based service. Unlike real estate or finance with long-tail transactions, these sorts of urgent care needs happen quickly because of immediacy. People decide on their provider with a shorter transaction window. So we ask ourselves, “Was it really the first thing they saw that drove the conversion?”
The best approach is use as much data as you can.
The creative rotation and the multiple signals that are exhibited to drive an actual conversion was something we really needed to dig in on for better reporting. If I were a marketing professional that just looked at my results generated by large conversion volume with only first-touch, I might say, “Optimize and reallocate all my spend towards these one or two messages.” But I would have really handicapped my results by pouring money into a metric while potentially losing site of other signals that could’ve contribution to the conversions.
Staying with urgent care — we quickly realized that five or six different signals could factor into whether a conversion truly happened versus the simple math of first or last-touch. It took some machine learning muscle from our analytics team to prove this out, and today we can confidently jettison touch-based attribution from the mix. Once you can demonstrate the effectiveness of a multi-touch journey and how we actually to a conversion, they’re by far the strongest indicators.
Most types of attribution models do not account for other external factors that influence sales and marketing effectiveness. How do we account for that?
The best approach is use as much data as you can, have as much transparency as you can, and realize there are intangible things that will affect a response that you just can’t account for.
What we do to understand that effect is use extremely rigorous control design. We try to implement control groups[amsive_tooltip term=”control-group”] as much as we can in different channels. And while we can’t get there a hundred percent, because not all the platforms allow trackable control groups, we can get pretty close to the truth doing it that way.
By controlling our audiences in different channels, we’re actually able to account for other noise, especially with a mature brand. So while we can pinpoint how our programmatic, our search, or our social media drove responses, we bake in the other intangible factors by normalizing for control.
How can top marketers convince eager leaders and decision-makers that attribution doesn’t necessarily tell the whole story, but is still effective?
When the client asks how we know my marketing spend worked, we have to be able to prove it because we do very rigorous feedback loops to show that channel effectiveness. While we all want fully integrated analytics[amsive_tooltip term=”analytics”], measurement and attribution, it’s hard to get it in a complete sense.
One of the problems with attribution across multi-channels is the fact that so many of these platforms are siloed off. We’ve heard the term “walled gardens” ad nauseum but it’s a real problem because you can’t get the transparency needed to prove each channel’s results separate from one another. One way to combat this is we have to build our own audiences — we don’t use Google’s programmatic audiences or Facebook’s audiences. We use those platforms for their reach and ad delivery algorithms because our clients demand more transparency.
We think the combination of advanced analytics and machine learning will differentiate the true marketing organizations out there.
Every attribution model has built-in biases and difficulties. How do we apply common sense to attribution?
You have to have discernible lift over control, incremental lift in the campaign, all those things. The reason we’re doing rigorous random control design is that we realize other factors like brand awareness, price-promotions, or other marketing efforts outside of these channels are actually driving the conversion. We have to account for these types of effects.
For example, we do a good bit of work in the telecom space, and when the industry introduces the latest new phone and features, our control response percentage jumps to 50 percent — meaning half of all of our marketing conversions would have happened anyway because they’re clearly driven by a new product promotion. However, when there’s a lull in new products, the control drops back down to around 35 percent. But we still have to account for the common sense shifts in that regard and not arbitrarily take credit where we shouldn’t.
What is the next step in terms of marketing attribution?
One of the largest growth factors that we’re seeing in the MarTech stack is in AI[amsive_tooltip term=”artificial-intelligence”] and the machine learning I was talking about earlier. While marketing isn’t necessarily the first thing that comes to mind when thinking about AI, ingesting millions and millions of data points leading to attribution can actually be easy math using these advanced tools.
It makes perfect sense to implement AI and ML in the marketing world given so many touchpoints increasing from ad platforms, social media sites, different connected devices, and more. We know it will require some heavy-lifting to sort out and we think the combination of advanced analytics and machine learning will differentiate the true marketing organizations out there.
