The short version, since that is usually what people want first:
Firmographic data describes what a company is. Industry, headcount, location, founded date, revenue band. It answers whether an account fits your ICP.
Technographic data describes what a company runs. CRM, cloud provider, analytics, security tooling. It answers whether your product fits their environment.
Intent data describes what a company appears to be researching. Content consumption, search behavior, review-site activity. It answers when an account might be in market.
The usual framing groups the first two as "fit" data and treats intent as the "timing" layer. That framing is correct as far as it goes, and every comparison of these three stops there.
It leaves out a fourth category, and the omission matters because the fourth is the most verifiable of the four.
The three, compared
| Firmographic | Technographic | Intent | |
|---|---|---|---|
| Answers | Who is this company? | What do they run? | What are they researching? |
| Unit | The organization | The tech stack | Behavior, usually aggregated |
| How it is obtained | Observed and classified | Detected and inferred | Inferred from activity |
| Coverage | Broad, nearly universal | Broad but uneven | Narrow by definition |
| Refresh cadence | Quarterly to annually | Monthly to quarterly | Daily to weekly |
| Verifiable by you | Partially | Often, by inspection | Rarely |
That last row is the one nobody includes, and it is the most important.
The layering everyone describes
The standard worked example, which is genuinely useful:
Firmographic filtering identifies, say, 10,000 companies matching your size, industry and geography. Technographic filtering narrows that to the 500 running a stack where your product creates value. Intent narrows further to the handful showing research activity this month.
Broad to narrow, fit then timing. It is a good model and it is why most teams buy in that order.
The problem is that the three layers are presented as equally dependable, and they are not. Each step down that funnel increases precision and decreases verifiability at the same time, which means the narrowest layer is also the one you can least confirm.
Observed, detected, inferred
A more useful way to sort these than fit-versus-timing is by how the data came to exist.
Observed data records something that demonstrably happened. A company was founded in 2011. A person holds a title. These are facts with a source you could, in principle, check.
Detected data infers a state from evidence. A script on a page suggests an analytics tool. A DNS record suggests an email provider. Usually right, sometimes wrong, and wrong in ways that are invisible from your side. A tag left behind from a tool the company stopped using last year looks identical to one in active use.
Inferred data derives an internal state from external behavior. Someone at an IP range associated with a company read three articles about a topic, therefore that company is in market. Each step in that chain is probabilistic.
Firmographic data is mostly observed with one significant classified field. Technographic data is detected. Intent data is inferred, often twice over, since identity is inferred from network signals and interest is inferred from reading behavior.
None of this makes intent data useless. It makes it a prior, not a fact, and the distinction matters because teams routinely treat all three as equally solid inputs to the same scoring model.
The fourth type: movement data
Here is the category the standard comparison leaves out.
Movement data records changes in people's professional circumstances. Someone joined a company. Someone left. Someone's title changed. A company opened a role it did not have before.
It is worth separating from the other three for a specific reason: it is observed rather than inferred, and it carries a timestamp.
When someone changes jobs, that is a fact with a date. It does not require a probabilistic chain of reasoning from network activity to company identity to purchase interest. It happened, and you know when.
Compare the two signals directly:
| Intent signal | Movement signal | |
|---|---|---|
| Example | Accounts at this company read about observability | The VP Engineering started three weeks ago |
| Basis | Inferred from aggregated behavior | Observed change of state |
| Timestamp | Approximate window | Specific date |
| Identity confidence | Company-level, probabilistic | Person-level, specific |
| Actionable on | An account | A named individual |
The practical difference is that an intent signal tells you an organization may be interested. A movement signal tells you a named person has a new mandate, a fresh budget cycle, and no loyalty to the incumbent vendor. One of those you can write an email to.
Three movement signals carry most of the value:
Someone joined. A new leader is measured on what changes in their first two quarters, which is the one window where buying something is a visible win rather than a disruption.
Someone left. Underused and the most interesting of the three, because one event tells you two things at once. The account they left just changed, so if they were your champion your renewal risk rose today. And they are landing somewhere new, carrying their tooling preferences with them.
A company started hiring for a function. A first hire in a function means a capability is being built from nothing, with no incumbent vendor and a budget that was approved rather than wished for.
The implementation of all three, including the deduplication problem that breaks every first version, is in job change signals, and the hiring side is in hiring signals.
Where each type actually fails
The failure modes are different, which means the validation work is different too.
Firmographic fails on classification. Industry labels are judgments, not facts. A software company selling to hospitals is classifiable either way and both answers are defensible. Headcount is a band pretending to be a number, with contractors, subsidiaries and reporting lag all introducing error. Build threshold rules on it and your model decides on noise. The detail is in the company data API guide.
Technographic fails on staleness and residue. A tag from a decommissioned tool looks exactly like an active one. A company can run a product in one business unit and not another. Detection tells you something was present, not that it is in use, and not at what scale.
Intent fails on attribution. Identity resolution from network signals is imperfect, aggregation obscures who inside the company was actually interested, and "researching a topic" and "about to buy" are not the same state. Coverage is also inherently narrow, since only companies generating detectable activity appear at all.
Movement fails on detection lag. Every provider learns about a change some time after it happened, and that lag varies. The signal is accurate when it arrives, but "nobody moved" means "we have not detected a move," which is a weaker claim than it sounds.
How to validate each type
Nobody writes this part, and it is the part that determines whether any of it is worth buying.
Firmographic: build a golden set. Take 100 companies you know well. Compare the vendor's headcount, industry and location against reality. Pay particular attention to where classification is ambiguous, because that is where providers diverge and where your rules will silently break if you switch.
Technographic: spot-check by inspection. This is the one type you can verify yourself. Pick 30 accounts, look at their public footprint, and check whether the detections hold. You will usually find the error rate is higher than advertised, concentrated in smaller companies and in tools with similar fingerprints.
Intent: run a holdout. Take two matched cohorts, work one using intent prioritization and one without, and compare conversion. This is the only honest test, because intent data always looks predictive in retrospect when you only examine the accounts that converted.
Movement: test retrospectively. Take twenty people you personally know changed jobs in the last ninety days. Count how many the provider has in the new role. Small sample, direct evidence, and it measures detection lag rather than coverage claims.
The general methodology, including why vendor-supplied samples are drawn from vendor strengths and why you must test on your own data, is in how to evaluate a B2B data provider.
Which to prioritize
Honest answer: it depends on your motion, and most teams buy in the wrong order.
If your ICP is not yet precise, start with firmographic. Everything else is a refinement layer on a boundary you have not drawn. The filter-ladder method for making an ICP genuinely narrow is in ICP filtering.
If you are displacing an incumbent, technographic is your lever, because knowing what they run is the whole pitch.
If your product sells into a new function, movement data is the highest-return single signal, since a first hire in a function means no incumbent exists.
If you have a large qualified universe and a small team, intent helps with prioritization, with the caveat that you are prioritizing on an inference.
If you sell to a stable, well-defined set of accounts, movement data beats intent comfortably, because you already know who they are and what you need is to know when something changed.
That last case is more common than the intent-first framing suggests. Plenty of B2B companies have a target list of a few hundred accounts and do not need help identifying them. They need to know when a champion leaves or a new leader arrives, which is an observed-fact problem rather than an inference problem.
Renidly covers the firmographic and movement sides of this, deliberately rather than by omission: resolved company records, people-level resolution, and movement as a queryable event stream with join, leave and title-change types. Those are the two categories that are observed rather than inferred, which is also why they are the two that can be checked. The database and live resolution products cover both, and the free job change detector will run a movement check on a single person if you want to see the shape of it.
Five mistakes
Treating all four as equally reliable inputs to one score. If a model weights an inferred signal the same as an observed one, it is more confident than its inputs justify. Weight by verifiability.
Buying intent before defining fit. High intent from the wrong kind of company is still a bad lead, and you will spend the quarter learning that.
Using industry as a precise filter. It is a judgment call. Use it to exclude obviously wrong segments, not to define narrow ones.
Never refreshing. Each type decays at a different rate, and refreshing all of them on one calendar means over-refreshing the stable fields and under-refreshing the volatile ones. The field-level model is in CRM data decay.
Assuming absence is evidence. No intent signal does not mean not in market. No detected movement does not mean nobody moved. Both mean nothing was detected, which is a different claim.
Frequently asked questions
What is the difference between firmographic and technographic data?
Firmographic data describes what a company is: industry, headcount, location, founded date. Technographic data describes what it runs: CRM, cloud, analytics, security tools. The first is mostly observed, the second is detected from evidence.
Is intent data better than firmographic data?
They answer different questions and neither replaces the other. Firmographic data tells you who qualifies, intent data suggests when a qualified account may be paying attention. Intent used without a fit filter points at companies that were never viable.
What are the main types of B2B data?
The three usually listed are firmographic, technographic and intent. A fourth, movement data covering job changes and hiring activity, is often omitted despite being the most directly verifiable of the four.
Which B2B data type is most reliable?
Firmographic and movement data are observed, so they are checkable. Technographic data is detected and carries false positives from residual or unused tooling. Intent data is inferred and is the hardest to validate independently.
How often should each type be refreshed?
Firmographic quarterly to annually, technographic monthly to quarterly, intent daily to weekly, and movement continuously, since its value is concentrated in a short window after the event.
Can I use these together?
Yes, and that is the normal pattern. Firmographic defines the universe, technographic refines fit, movement and intent indicate timing. Weight them by how verifiable each one is rather than treating them as equivalent.
The one-line version
Firmographic answers who, technographic answers what they run, intent suggests when, and movement data, which most comparisons omit entirely, records what actually changed and when.
The useful axis is not fit versus timing, it is observed versus inferred. Firmographic and movement data are checkable. Technographic data is detected and carries residue. Intent data is inferred twice over and should be weighted accordingly rather than trusted as a fact.
Before you buy any of them, work out how you would validate it. If you cannot design that test, you cannot tell whether it is working, which means you will keep paying for it regardless of whether it is.
If you want to check the movement side on real people before committing to anything, the free tools will run a single lookup with no signup, and for volume or a procurement conversation, talk to us directly.
