5 Keys to Unlock AI’s Potential for B2B Marketing

I used to dread hearing the objection “garbage in, garbage out.” It almost always came from a savvy marketing operations pro, and white it felt like a cop-out response, I couldn’t deny there was definitely credence to the comment.  Today, however, I have turned the table and now lead with the “garbage in, garbage out” message when describing the value of AI to generate meaningful and trustworthy recommendations using comprehensive customer signals and more advanced attribution techniques.   This does not have to do with data hygiene, but rather, the quality of the customer signal the AI is sitting on and how well models are trained in B2B marketing.

Here are 5 areas that need to be solved in order to trust AI powered recommendations:

1) Customer Signal

Many legacy and even some current marketing attribution software providers still rely only on hand-raiser or first-party data to fuel their models.  But in today’s B2B landscape—where buying decisions involve 10-15 influencers (or more)—this approach is fundamentally flawed. Few decision-makers are filling out forms or downloading whitepapers, yet traditional attribution tools still base their insights on less than 1% of account signals from form fills. That’s a dangerously incomplete picture.

This is where de-anonymization of web traffic with accurate, cookieless, account identification is so important.   While 30% used to be considered best-in-class back in my ABM days, 60% is the new gold standard and it takes a multi-source, privacy-forward model to get there.

2) Correct Attribution of Direct Traffic

A problem plaguing B2B marketers, and the reason I have been an “attribution hater” for years, is that the majority (75%+) of B2B traffic comes from an unknown source and gets thrown into a bucket called “direct.”  Reality is that it was likely driven from other digital programs, events, social posts, your sales teams and other sources.   But if the source(s) of your largest customer signal are ignored, how can AI or any B2B attribution model generate the proper guidance on where to invest more or less.  Fear not, there are new view-thru technologies and API’s in the works to finally solve this issue. Read more about my thoughts on the actual source of “direct” traffic.

3) Predictive Attribution 

Legacy attribution solutions that rely on visits or touches are historical at best—they provide insight into first-touch and marketing’s influence on journey stages but lack predictive capabilities.  But if you want to know where to invest your next $100,000, you might get stuck.  More modern solutions leverage machine learning and advanced modeling techniques to predict how changes in investment or channel mix impact key business outcomes such as pipeline growth, ACV or even sales cycle length.  Fortunately, this has become easier with simple integrations into the major media platforms and CRM systems to draw these correlations using machine learning.   However, for predictive attribution to work, #1 (Customer Signal) and #2 (Direct Traffic Identification) must be in place or you risk insufficient data and unreliable recommendations—AKA as “garbage in, garbage out.”

4) Unified Data and Common Measurements or KPI’s

I’ve seen growing frustration across marketing ops and B2B digital teams struggling to match and synchronize campaign results from different vendors with their web analytics and CRM systems.  No surprise here as the numbers can be quite different and KPI’s are all over the map.   To simplify and build trust in the results, I’ve found it best to synchronize the data at the account level (vendor data to digital engagement to CRM) following impressions to interactions to business outcomes.  Once the data is unified, you can then establish a common set of KPI’s to measure all channels and vendors on a level playing field.   I’ve collaborated with a dozen agencies and large enterprises and settled on reach, engagement, $ efficiency and business impact which is typically a measure of pipeline influence.   For a deeper dive into KPI alignment and campaign benchmarking, check out my findings in a blog post here.

5) AI Models Trained by B2B Experts

AI might be the easy part compared to getting #1-#4 nailed down, but it will only be effective if it starts with quality data and a deep customer signal.  Additionally, it must be trained specifically for B2B marketing, learning from historical actions and their resulting outcomes.  Some vendors claim to offer marketing AI functionality, but in reality, their capabilities are often limited to basic customer support recommendations.  Learning HOW to measure $ per visit is simply not that interesting. However, AI becomes truly powerful when it can suggest a series of investment changes and understand how pipeline will be impacted, or ask how to optimize your Google Display campaigns to better target certain enterprises.

In the next week, we will publish part 2 of this blog, diving into the phases of AI in B2B marketing that go way beyond writing blogs like this one (yes, I tried, but the copy was just average). Stay tuned!