The path to integrating intelligence throughout the martech stack
- Summary:
- Marketing and AI are going to get a heavy does of hype this year. But Barb Mosher Zinck does see some progress. Turns out that machine learning can help with of the most vexing martech problems: lead attribution. A more interesting play is coming: look out for the "integrated intelligence layer."
A critical element of that intelligence is attribution. Attribution is about understanding the importance of each leg in the customer journey, each touchpoint, each interaction, each piece of content, and then optimizing the right ones. This is an area many marketers struggle with and many more are focusing on this year.
A growing focus on attribution
IAB and the Winterberry Group surveyed a group of digital marketing and media managers on their use of audience data, and they found that marketers are evolving in their approach, but still struggling with implementation for a few reasons.
Look at how approaches have slowly evolved:
- In 2015 the focus was on programmatic media buying
- In 2016 it was general audience analytics
- In 2017, the focus is on cross-channel measurement and attribution.
Essentially, marketers are shifting from single focus activities to improve conversions to understanding how each piece in the marketing activity bundle is contributing to the overall picture. Consumers don’t typically purchase after one initial connection; they look at a few things, are exposed to advertising and emails, social media and other tactics that keep a product top of mind.
Why is 2017 all about attribution? It’s not new, but it is getting easier to figure out. In the IAB study, factors such as the increasing volume and quality of first-party data (52.6%), the growing emphasis on measuring investments (50%) and improvements in the technology to support attribution (46.1%) are top reasons. But the most important factor is that customers are demanding this capability clearly showing that marketing is a revenue-generator and it has to prove exactly how.
Let’s talk challenges though, because even if there’s a demand and a desire to show marketing contributes and how, it’s still not always easy to do. In this study, the challenges are straightforward:
- 3% have problems proving ROI
- 3% Lack internal experience at the operational/functional level
- 7% don’t have the right technology
- 36% struggle with siloed organizational structures and poor data-sharing protocols
So 2017 will see a strong focus on attribution, but it will be a struggle. It’s mid-April, plans should be in place to identify the best strategy and implement the right technology.
The role machine learning plays in the intelligence market
With attribution moving up the ladder in the digital marketer’s toolkit, it makes sense that technology providers like Conversion Logic are doing well. It recently closed Series A funding to take its offering to the next level. I spoke with CEO and co-founder, Brian Baumgart, about the marketing intelligence market, the role of machine learning and the future of intelligence throughout the marketing technology stack.
First, machine learning (ML), which everyone is talking about, on its own doesn’t mean a great deal. As Baumgart points out, the challenge with ML is that there aren’t any standards, so there’s no baseline understanding of what ML means and there is no common thread to approaches or application. This is especially true in martech, he said.
It’s also important that marketers understand the difference between ML and Artificial Intelligence (AI), the latter of which is - at this point - aspirational for marketing. Machine learning is a foundational aspect to AI, and you need to ensure the technology you purchase is implementing ML right.
Functioning properly, machine learning provides the algorithms that make machines smarter but keep in mind that a lot of these algorithms are commoditized - many products use them. It’s how they are used in combination that often differentiates a solution from others. Baumgart said that Conversion Logic is very transparent about the algorithms its product uses, noting that it’s key to explain the methodology, implementation and how it applies understanding to what the algorithms mean that gives customers confidence in the analytics and what comes out of ML.
When the algorithms are black-boxed, and you are supposed to trust what they are doing and the results, that should make you question what’s really going on. Some companies are fortunate to have data scientists in-house that can validate the algorithms and how they are used, but this isn’t always the case.
There’s a reason ML is used for attribution - it helps improve the accuracy of the analysis. Consider the amount of historical data you need to analyze; no human can accurately analyze all that data and come back with realistic insights. The same for predictive analysis. And ML can learn as the data grows.
Intelligence tools are one part of the marketing stack, and that can make it challenging to do attribution and other analysis. This is starting to change. Baumgart said that Conversion Logic is good at leveraging machine learning to better understand attribution for the purpose of marketing efficiency using historical data. The next step is to incorporate predictive capabilities to improve how you can look into the future and optimize your next dollar spent.
But there’s a bigger, more important future for marketing intelligence.
The integrated intelligence layer is coming
Right now, marketing intelligence technology, like Conversion Logic, is shifting to a top layer that sits on top of the marketing stack. It’s not completely integrated, but it’s moving closer, and the flow of data into the intelligence layer is improving.
Part of the problem is the best of breed marketing technology stack that most organizations are putting in place. This mixed bag of tech makes integration more difficult, but Baumgart said it also makes the solution more valuable. As fragmentation continues to increase, MTA (marketing technology attribution) solutions can offer a holistic intelligence layer that informs not just on efficiency and optimization but also deeper insights into the entire portfolio including tools, technology, vendors, and customer experience.
But the future, which may not be that far away, is when that ML intelligence is implemented throughout the stack. When this happens, you’ll be able to leverage full stack AI analysis on the entire dataset.
What that full stack analysis does is enable you to look at the entire customer lifecycle, not just parts like the purchase journey. Marketers spend a lot of time improving the purchase journey, but companies that analyze the complete lifecycle, understanding each leg of the journey - including different touchpoints and interactions with the company - can make improvements across the entire customer lifecycle.
My take
Data continues to be a primary challenge for marketing intelligence and proper attribution. It’s about quality data, but also the collection of all the right data across all the touchpoints and technology in the marketing toolkit. But if we’re waiting for that to be perfect, we’re likely going to wait a very long time.
Baumgart said something important. He said incremental improvements could be exponentially beneficial. We can’t wait for the full, perfect package of intelligence to arrive. We have to take the time to get it right because it is so important to the best customer experience.
If the intelligence layer sits on top and connects down through all the data in the stack, it’s going to take time to ensure that data is analyzed correctly and supporting the right insights and actions. That’s when AI comes into play. Let’s get good at the first part though. Fully integrated intelligence will come.