Blog

Comparing AI-Derived Signals and TechTarget’s Directly Observed Intent Data

AI vs. purchase intentWith something like 10,000 RevTech solutions out there, it’s become super hard to understand who really does what, and most importantly, whether or not what is being said really stands to help you deliver substantive value to your company at the end of the day. At TechTarget, we realize that while there’s very little real mystery in what we do, there’s absolutely magic in how it impacts our clients’ results. Toward helping you and ourselves navigate through the functionality fog that, at times, threatens to obscure what’s really necessary to make steady progress, we’ve put together a series of pieces that examine recent claims and present our case for a more transparent and pragmatic way to understand the issues involved.

Claims of “the best AI in the business”

According to McKinsey’s latest research The State of AI in 2021, B2B marketing is not even among the top 10 use-cases for AI today across the 8 functions they surveyed. Drilling into the Sales and Marketing functions specifically, the top two use-cases for AI are Customer Service Analytics and Customer Segmentation, with only 17% and 16% of all companies surveyed attempting to apply AI to their work. To us this raises the question of why, in fact, are RevTech companies actually touting AI, when the experts are only just beginning to find any use for it, and the cases are largely from the high-volume, high-cost or high-transaction worlds of customer service and consumer products? We believe that vendors are touting AI for two simple reasons: Because it makes them sound smart and because it is difficult for potential customers to inspect. We don’t believe that in the ways it is currently deployed with respect to contact identification, that AI can deliver substantive value to our customers in comparison to other more transparent methods.

What they’re saying, trying to claim, and hoping that you will accept.

The claim of having the best AI in the business is easy to make because none of us have any way to evaluate it. The question you should be asking is: what is this AI good at? And how can I determine that it’s better than other methods?

What data are they using for this in the first place? Importantly, AI requires a relevant data set to run on. For our purposes, that means one that must necessarily include connection between people and behaviors. If a vendor doesn’t have such information, they have to get it from you to run their models. But as recent history has shown, modeling B2B clients’ first party data yields very little, if any lift (the companies that tried to build a business around this have all either been absorbed by others with more successful business models, or they have added potentially more impactful capabilities to their portfolios).

So what’s going on here? Like everyone, an “AI-powered” offering can scrape ‘open’ (and therefore low-value) domains from the internet, or they can buy data from sites willing to sell it to them, (which begs the question why a site would do this. We think, frankly, it’s because it’s weak in the first place and that these are sites that have difficulty monetizing it on their own due to quality or scale issues).

Black boxes that you can’t see into to evaluate for yourself

The theory seems to be that they’ll pick up enough of this low-ranking, low-quality traffic, put it all together using a data lake, analyze it with an AI model, and make a pretty good guess at what you should prioritize. But these super weak signals that they are aggregating are commonly wrong with respect to the existence of demand, because they start from a very weak inference in the first place – an inference that often uses only a general topic key word as the connection between a search and your solution. And furthermore, such signals aren’t additive – in data, just as in mortgage-backed securities (you do remember 2008 still don’t you), two weak things can’t be added together to make a strong.

At best, this method delivers only an indication of weak general interest in a topic, and that’s why they so often have to add in feeds from G2/TrustRadius as an account data source – it’s to create the appearance of some sort of verification of their initial “surge”.  As you probably know, those two suppliers make their money from vendors who pay for promotion (which puts even their signals’ veracity in doubt), but at least with them you know the account was looking at a solution like yours. Unfortunately for you, an AI-based vendor can also use this to backtrack to their other signals and then claim that any surge that matches the G2/TrustRadius or Gartner signal must therefore be real. Again, a false positive from G2 used to verify that a false positive from AI is real should not convince you in our opinion.

TechTarget’s approach, our solutions and directions

Years of academic research, and both our own analysis and personal observations of the markets around us, have proven that buyers largely cluster in locations where they can easily find fulfillment of their needs. (In our category, this comes in the form of the decision-support information they require to make enterprise tech purchases). This is why we built our company – to create such clusters for our clients (and ourselves) to then monetize. It’s because of our huge content volumes and SEO power, that we can reliably continue attracting these tech buyers to us versus other suppliers of similar material.  We’re the #1 publisher of buyer decision-support for enterprise tech by far: A million 1st page organic rankings that lead to massive engagement and click-throughs! With our scale and precision, we reliably intercept the vast majority of enterprise tech buyer research occurring on the web at any given time.

We know the content so we know exactly what each person truly cares about

And since we make all the editorial content and can therefore successfully utilize an opt-in membership model, we know exactly who’s reading about what and why. This means you’re able to then personalize to the specific individual — yielding better conversion, more meetings, and better conversations.

At TechTarget, we aren’t interpreting a random scatter of signals and asking you to believe. We’re delivering to you dozens of relevant interactions per user and active buying team; interactions you can easily inspect and understand as the outputs of a transparent process. And with BrightTALK signals now added, the breadth and depth of our data can literally mean hours of engagement with exceedingly granular material about your category and products. There really is no other source of insights out there that comes anywhere near this strength of signal, its accuracy at showing real buyer’s journeys, or it’s precision in understanding each individual that makes up the buying group.

Directly observed intent: A transparent method that you can inspect, understand and evaluate

As a marketer, you know well, that each element within this data concept can deliver critical advantages in your efforts at successfully maximizing go-to-market performance – from strategically how you understand your markets and targets, how you position and message to them, and executionally, from your product concepts to your content, right from click all the way through the close. Despite everything you might be hearing, in the category of data for enterprise tech marketing and sales, no data- generating process is currently able to produce an output that’s stronger, more accurate or more precise than Priority Engine Prospect-Level Intent™.

This blog is one small part of my ongoing effort to both provide useful information and clearly explain why we’re so confident that Priority Engine will deliver fewer false positives, fewer false negatives, and many more opportunities into your pipeline, faster and more reliably at scale. As always, my team and I are here to help and to discuss this material at your convenience. I can be reached here. I hope very much to talk to you soon – Steve

More Posts On