Prospect Segmentation: The Data Fields That Move Pipeline

The most valuable data fields for prospect segmentation are the ones that predict buying behaviour, not just describe company characteristics. Firmographics get you in the right postcode. Intent signals, technographic fit, and engagement depth get you in the right room. Most teams build segments from the data that is easiest to collect, not the data that is most predictive, and that is where the gap between average and effective segmentation opens up.

This article breaks down which data fields consistently drive better segmentation outcomes, why some widely used fields are weaker than they appear, and how to think about layering data types to build segments that marketing and sales can actually use.

Why Most Segmentation Starts in the Wrong Place

When I was running agency teams, we would inherit client CRM data and the first thing I would ask was: how was this data collected, and what decisions was it built to support? The answer was almost always the same. It was built to support reporting, not prospecting. Fields were populated because someone had to fill in a form, not because the field predicted anything useful.

That distinction matters enormously. Segmentation built on reporting-grade data gives you clean-looking lists with poor conversion rates. Segmentation built on prediction-grade data gives you smaller lists that actually move.

The instinct most teams have is to start with the data they have, then work out what to do with it. The better approach is to start with the question you are trying to answer, which is usually some version of “which prospects are most likely to buy, and when,” and then work backwards to identify which data fields answer that question most reliably.

If you want a broader frame for how segmentation fits into go-to-market thinking, the Go-To-Market and Growth Strategy hub covers the upstream decisions that shape how segmentation gets used in practice.

Which Firmographic Fields Are Worth Keeping

Firmographics are the baseline. You cannot do meaningful B2B segmentation without them. But not all firmographic fields carry equal weight, and some that get used routinely are weaker signals than most teams assume.

Employee count is the single most consistently useful firmographic field across most B2B categories. It is a reasonable proxy for budget, procurement complexity, and the likely number of stakeholders involved in a buying decision. A company with 50 employees and a company with 5,000 employees are not just different in size. They are different buying entities with different sales motions, different approval chains, and different product requirements. Segment them accordingly.

Revenue band is useful when you can get it, though it is harder to source accurately for private companies. Where it is available, it adds a dimension that employee count alone misses. A professional services firm with 200 people can have very different revenue profiles depending on its model, and that difference affects willingness to pay and budget authority.

Industry vertical is useful but only at sufficient granularity. “Financial services” is not a segment. “Independent wealth management firms with under £500m AUM” is a segment. The broader the vertical, the less predictive it becomes. I have seen campaigns targeting “retail” perform badly because the segment included everything from grocery chains to luxury fashion, which have almost nothing in common from a buying behaviour perspective.

Geography is often over-relied upon as a segmentation variable. It matters operationally, for territory management and language, but it rarely predicts buying behaviour on its own. A mid-market SaaS company in Manchester and one in Munich are more similar to each other than either is to an enterprise in the same city.

Company age is worth including in your data model but rarely as a primary segmentation field. It can be a useful filter, for example, excluding companies under 12 months old from certain outbound sequences, but it is not a strong predictor of fit or readiness on its own.

Technographic Data: The Underused Signal

If I had to pick one data type that most B2B teams underuse relative to its predictive value, it would be technographics. Knowing which tools a prospect already runs tells you more about their operational maturity, their likely pain points, and their openness to your category than almost any firmographic field.

Consider a company running Salesforce, Marketo, and a data enrichment tool. That technology stack tells you they have made meaningful investments in their go-to-market infrastructure, they probably have dedicated ops or RevOps resource, and they are likely to evaluate new tools through a structured procurement process. That is a very different prospect from a company running HubSpot with manual processes bolted on.

Technographic signals are particularly powerful in three scenarios. First, when your product integrates with or displaces a specific tool, knowing who runs that tool is a direct proxy for fit. Second, when a prospect’s current stack has a visible gap your product fills. Third, when a competitor’s customer shows up in your pipeline, because their existing tool tells you something about the problem they have already decided is worth solving.

Sources like BuiltWith, Bombora, and G2 Buyer Intent all surface technographic and intent data at varying levels of coverage and cost. The data is imperfect, but it is directionally useful, and directionally useful is often enough to improve segment prioritisation meaningfully.

Intent Signals and Why Recency Beats Volume

Intent data has become a significant part of the B2B data conversation over the past several years, and for good reason. A prospect who is actively researching your category right now is categorically different from one who fits your ICP but has shown no recent signal. The challenge is that intent data is often used clumsily, with teams treating any intent signal as equal regardless of when it occurred or what it actually indicated.

Recency is the most important dimension of intent data. A company that visited your pricing page last Tuesday is a materially better prospect than one that downloaded a white paper in October. The pricing page visit is a more specific signal, and it happened recently enough that the need is likely still active. The white paper download might have been research, competitive analysis, or a junior employee doing background reading.

When I was managing large-scale performance programmes, one of the consistent findings was that recency of engagement was a stronger predictor of conversion than depth of historical engagement. This connects to something I have come to believe more firmly over time: much of what performance marketing gets credited for is capturing intent that already existed. The prospect was going to look for a solution. The question is whether your segmentation and outreach got there while the intent was still live.

The most useful intent signals to track, roughly in order of specificity, are: pricing page visits, product comparison page visits, demo requests that did not convert, content downloads tied to specific pain points, and third-party intent signals showing category-level research. Each of these tells you something different about where a prospect is in their thinking, and they should be weighted accordingly in your segmentation model.

There is useful thinking on how GTM teams are approaching pipeline signals in Vidyard’s Future Revenue Report, which looks at where teams see untapped potential in their existing pipeline data.

Behavioural Data from Your Own Systems

First-party behavioural data, the signals generated by how prospects interact with your own website, content, and product, is often the most actionable data you have and the most underused for segmentation purposes.

The problem is that most teams track this data at the individual contact level and do not aggregate it to the account level. If three people from the same company have visited your site in the past two weeks, that is a stronger signal than one person visiting three times. Account-level aggregation of behavioural signals is one of the highest-value, lowest-cost improvements most B2B teams can make to their segmentation approach.

Specific behavioural fields worth capturing and using in segmentation include: number of unique visitors from the account in a rolling 30-day window, pages visited and their position in the buyer experience, time between first and most recent visit, and whether any visit has included high-intent pages like pricing, case studies, or integration documentation.

One thing I would add from experience: behavioural data is most useful when it is tied to a specific segment hypothesis. Tracking everything and hoping patterns emerge is less effective than deciding in advance what you think the behavioural signature of a high-intent prospect looks like, building a segment around that hypothesis, and then testing whether it converts at a higher rate. That is the discipline that separates teams who use data well from teams who just have a lot of it.

Contact-Level Fields That Change the Conversation

Account-level segmentation tells you which companies to target. Contact-level fields tell you who to reach and with what message. The two layers work together, and most segmentation frameworks are weaker at the contact level than the account level.

Job function is more useful than job title for segmentation purposes because titles vary enormously across companies of different sizes. “Head of Growth” at a 30-person startup and “VP Marketing” at a 500-person company might be doing similar jobs. Function-based segmentation, marketing, sales, operations, finance, and so on, is more portable across your ICP.

Seniority level matters for message design as much as for targeting. A C-suite contact cares about outcomes and risk. A manager-level contact cares about process and implementation. If your segmentation does not account for seniority, you will end up sending the same message to people who need very different things from it.

Tenure in role is a surprisingly useful field that most teams ignore. Someone who has been in a role for six months is often in a different buying posture than someone who has been there for three years. New leaders tend to evaluate existing vendors and tools. They have mandate and motivation to make changes. That is worth knowing.

Previous company is worth capturing where possible. A contact who previously worked at one of your customers brings implicit familiarity with your category and potentially with your product. That changes the conversation from education to evaluation, which is a shorter sales cycle.

Building Segments Backwards from Closed-Won Data

The most reliable way to identify which data fields matter most for your specific business is to analyse your closed-won deals and identify the fields that most clearly distinguish customers from prospects who did not convert. This sounds obvious. It is surprisingly rarely done with any rigour.

Early in my career, I was more likely to build segments based on intuition about who should buy than on evidence about who actually did buy. The two are often similar, but the gaps between them are where the insight lives. I have seen teams discover that their best customers were consistently in a specific revenue band they had been underweighting, or that a particular technographic combination was a stronger predictor of fit than the industry vertical they had been using as their primary filter.

A practical closed-won analysis for segmentation purposes should look at: employee count and revenue band distribution across won deals, industry vertical at a granular level, technographic stack at the time of purchase, seniority and function of the primary champion, time from first engagement to close, and the specific content or touchpoints that appeared in the engagement history of won accounts. That last one is often the most revealing.

BCG has written usefully about how go-to-market strategy in B2B markets requires careful attention to segment-level economics, not just top-line targeting. The same principle applies to segmentation design: the fields that matter most are the ones that predict commercial outcomes, not the ones that make your lists look tidy.

The Fields That Look Useful but Rarely Are

A few fields that appear in most CRM setups and segmentation frameworks are weaker signals than their prevalence suggests.

Lead source is useful for attribution reporting but is a poor segmentation variable. “Organic search” and “paid social” tell you how a prospect arrived, not whether they are a good fit or likely to buy. Segmenting by lead source often produces segments that are more about your marketing channels than about your buyers.

Broad industry codes, particularly SIC or NAICS codes at the two-digit level, are too coarse to be useful for most B2B segmentation. A two-digit SIC code for “manufacturing” covers everything from food production to aerospace. If you are using these, go to at least four digits, and ideally supplement with your own industry taxonomy built around how your buyers actually describe themselves.

Social media presence is occasionally cited as a segmentation variable, typically LinkedIn follower count or posting frequency. In my experience, this has almost no predictive value for B2B buying behaviour. A company that posts frequently on LinkedIn is not a better prospect than one that does not. It just has a more active social media team.

Job posting activity is a more interesting signal than social presence, particularly for specific roles. A company hiring a Head of Revenue Operations is signalling investment in their GTM infrastructure. A company hiring multiple SDRs is signalling a push on outbound. These signals are worth monitoring for specific use cases, but they require interpretation and decay quickly, so they work better as trigger signals than as static segmentation fields.

How to Layer Data Fields Into a Working Segmentation Model

The practical question is not which individual fields matter most in isolation, but how to combine fields into a segmentation model that is both predictive and operationally usable. A model with 40 fields is not better than one with 8 fields if the 40-field model cannot be maintained or acted on.

A workable layered approach uses three tiers. The first tier is fit: does this account match our ICP on the firmographic and technographic dimensions that correlate with our best customers? This is a binary or scored filter that determines whether an account is in scope at all. The second tier is timing: is there evidence of active need or intent right now? This is where behavioural and intent signals come in. The third tier is access: do we have a contact at the right level and function who we can reach through a channel we have access to?

Accounts that score well on all three tiers are your highest-priority segment. Accounts that score well on fit but poorly on timing are worth nurturing but not worth aggressive outbound. Accounts that score well on timing but poorly on fit are worth a quick qualification conversation but should not consume significant sales resource.

This kind of tiered model is not complex to build, but it does require discipline to maintain. The most common failure mode I have seen is teams building a rigorous segmentation model and then abandoning it six weeks later because it was not producing enough volume. The answer to that problem is usually not to loosen the segmentation criteria. It is to expand the addressable universe by reaching audiences that have not yet shown fit or intent, which is a different problem with a different solution.

For more on how segmentation connects to broader growth decisions, the Go-To-Market and Growth Strategy hub covers the strategic context that makes segmentation decisions meaningful rather than mechanical.

About the Author

Keith Lacy is a marketing strategist and former agency CEO with 20+ years of experience across agency leadership, performance marketing, and commercial strategy. He writes The Marketing Juice to cut through the noise and share what works.

Frequently Asked Questions