Tl;dr - It's common for people to wonder about the accuracy of third-party intent data. There's a certain element of mystery to most data. For some that creates stronger faith in the info since there aren't clear points for quibbling. For others, in contrast, that introduces more doubt. What's important to understand are the various areas of potential inaccuracy - including how you interpret and use it!
How Will We Define Accuracy?
Third-party intent data includes a lot of explicit and inferred information. So to answer questions around how accurate intent data is, we need to clarify the elements of accuracy we are trying to quantify. For example:
- widespread account-level activity
- specific individual activity
- prospects' understanding of their objectives
- accuracy of elements behind an opaque topic taxonomy
- precision of natural language processing algorithms
- your understanding of the market, personas, and signals of intent
- contact details
- interpretation of signals
In other words there are a number of points in the intent data process where accuracy can be impacted.
- Do you really (like, REALLY) know your ICP, who your buyers are, what their journey is, and what actions indicate intent for each permutation?
- Can you clearly convey this to the third-party intent data provider you're using?
- Is your data provider able to accurately translate that profile into their data collection methodology?
- Does your provider's technology accurately execute on their conceptual methodology?
- How many 3rd parties does the data result depend on?
- Does your team accurately interpret the data? In other words, are they accurately understanding what it's telling them? Or what they want it to be saying?
- Is the marketing and sales follow up "accurate" to the prospect's situation?
- Are details like email address and job title current and correct?
And of course there are two high-level types of inaccuracy. Most people focus on false positives. The other to consider is whether you're missing signals that you should have captured.
Intent data is complex technology and the answer, therefore, is complex. That doesn't mean you can't make some informed decisions, however. Let's unpack the question of third-party intent data accuracy.
What's the Underlying Intent Data Methodology?
Let's start with the basics. How is intent data collected?
Different methodologies include exchange of anonymous information among publishing websites, bidstream data via DSPs (programmatic advertising,) and crawling technology. Some rely on 3rd party cookies and reverse IP lookups (IntentData.io doesn't use either of those methodologies.)
Those different methods raise various questions. What is the accuracy of data shared between different entities (e.g. account level activity shared between publishers)? And how accurate is bidstream data? Some estimates of bidstream data accuracy indicate that 80, 90 or even 99% of the data may be inaccurate.
Second party data (where a publisher, which owns content, monetizes the behaviors and activities of known users with that content) is likely the most accurate - assuming that the publisher's tracking methodology and technology are current.
Is Content Engagement Accurate?
If it's important to us to know that a company is taking action around a topic - let's use "unified communications" as an example - then an important factor in accuracy is how content is identified as relevant.
Natural Language Processing
If someone engages in an interaction which has a hashtag of "#UnifiedCommunications" that's unambiguous. But that's also a relatively small percentage of the 3rd party purchase intent data signals that are collected. More common is the use of natural language processing (NLP) technology to determine which content is relevant.
In other words, is some content really focused on the topic/product "unified communications" or not? Obviously just having both words in some content doesn't mean it's the same.
"The CEO and board agreed that they needed to be more unified in their shareholder communications" is very different than "The most important trend in small business telephony in 2020 is likely to be the shift toward fully unified communications platforms."
So how robust and accurate is the NLP? That's frankly an unknown for buyers of intent data.
Content & Topic Taxonomies
Let's assume for a moment that the NLP perfectly and consistently identifies the real essence of content and only informs real, legitimate and relevant engagements.
Now consider what it's searching for.
How clear and specific are the criteria? Are they unknown elements of an opaque topic taxonomy? Or are they specific, granular terms, competitors, people and events that you've identified as strings, handles and URLs? Are they searching for an "exact" match or a string of various terms in some content.
In other words, are "IP telephones" and "virtual PBX" part of "unified communications"? Should they be? Do you consider them relevant? How about the switches and software that are used by providers of virtual PBX service? You probably have an opinion - and certainly will when it comes to those sorts of details around your product/service.
Some related terms will provide helpful marketing and sales intelligence related to selling your products. Some will not. The same applies to the content behind them.
So it's worth asking how much granular detail you'll have to manage the signals you get and discern the relevance.
Are People Looking for the "Right Things" or Their Version?
Talk to any good physician or consultant and you'll hear about patients/clients who self diagnose and are almost always wrong. Skilled, customer-focused sales people often find the same. Really understanding the problem to solve and outcome to achieve often take perspective that comes with experience.
So the next area of accuracy to consider is whether you're observing the activity that as an industry expert indicates someone is in market for a product or service, or are you observing what they might be searching and thinking because they don't have your industry expertise?
Obviously the right answer is to monitor both - but if you're only tracking what an expert or experienced buyer would reference, then you're missing signal.
Are Contact Details Accurate?
Some 3rd party data includes contact details. This raises a couple other issues of accuracy.
First, how confident is the contact level™ intent data vendor that the named contact is the one that actually took the action? Second, how accurate are the reported details?
The answer for the first is in the publicly available detail of how the individual's actions are observed. In the case of IntentData.io it's based on observing people taking public action (no ambiguity there) and then a powerful algorithm to resolve from a public profile, who the person is.
The second piece depends on a number of factors related to data enrichment accuracy and quality. Email addresses, for instance, are sourced through data partners.
So, while confidence and accuracy in correctly identifying the actual person is quite high, the accuracy of appended details like email addresses is subject to industry standards on enrichment type data.
How Accurate is Your Interpretation of the Data?
The nature of intent data is that it is sales and marketing intelligence.
Intelligence takes analysis. Analysis involves making inferences, relying on some assumptions, and piecing together lots of different data points to try to create a full picture.
You'll never know for sure - but as you collect more signals you'll make a better guess.
You've decided what it should mean before you analyze it.
An important element of intent data accuracy is actually the accuracy of your interpretation of the data.
If you believe that every signal indicates someone about to write an order for your product or service, then not only are you wrong, but your interpretation of the data will be horribly inaccurate.
On the other hand, if you see each signal as part of a tapestry into which you weave first party data and other 3rd party data sources, and you consider account and contact fit and activity, then you'll have a more accurate interpretation of what it means.
How Accurate (Appropriate) is Your Follow Up?
The final step is to take action with the data.
Even if everything else was "5-9s" (99.999%) accurate to this point, you could still make a real mess of things.
Your marketing and sales approach needs to be contextually appropriate - if it's not, then it will be completely inaccurate from the prospect's perspective!
Of course you'd never consider some clumsy version of "Since I know you're evaluating unified communications solutions you've got to look at ours because it's the best."
But, will you put prospects into narrow custom audiences for high personalized ad content? Deliver dynamic web content that's tailored to what your analysis indicates about seniority, function, stage in buying journey, size and engagement of buying team, specific competitors, etc.?
If you don't; if you just use the same generic approach for all of it, then you'll have disappointing results because in ignoring rich detail in the data, particularly contact level intent data, you'll have created inaccuracy in the application of it.
Is Accuracy Really the Question You Mean to Ask?
When someone asks me about the accuracy of intent data what I understand is that they're trying to somehow gauge what the business impact will be of allocating a portion of a finite marketing or sales budget.
That's an entirely appropriate question and one that a dive into NLP algorithms won't answer.
The answer to that is in the "lead bullets."
If you really enjoy nerding out on data details (hey, we're there with you) then hopefully this post has helped to answer the common question "How accurate is intent data".