What buying signals look like by company size

If you track buying signals, you have probably noticed that some accounts never do anything. Weeks pass, the record sits there, and nothing arrives. The natural reading is that the tool is missing something.

We went looking for the answer in our own data, because we are in a position to check both halves of it: what the world published, and whether we were able to see it. This post is what we found, including the parts that argue against having a signal feed at all for some segments. The short version, if you sell to small companies: the silence is mostly real, and the last section is about what to do with that.

Two measurements, both taken on 11 August 2026. The first counted signals across 277 monitored companies. The second went outside our own system and checked, by hand and by probe, whether the quiet companies were quiet in reality.

The finding

Signal density tracks company size, and the gap is not subtle. Here is the full split, with the number of companies in each bucket, because two of these buckets are too small to carry an opinion.

1–1012 companies

91.7%

11–5056 companies

96.4%

51–1007 companies

42.9%

101–50032 companies

62.5%

500+28 companies

21.4%

Size unknown142 companies

72.5%

Outlined bars are the two buckets too small to carry an estimate — seven and twelve companies. They are drawn because leaving them out would be its own kind of editing, not because they measure anything.

Read the bottom two rows against the top two. A company with more than 500 employees produced a median of 17 signals in the window and was silent about a fifth of the time; before you quote that 17 anywhere, read the method section on why the window is not a unit of time, so it cannot be converted into a monthly rate. A company with 11 to 50 employees was silent 96.4% of the time, and its median was zero. Not "low": zero. The 75th percentile was zero too, meaning three quarters of that bucket produced nothing at all.

Across every company we monitored whose size was known and under 100 employees, 68 of 75 produced no signal of any kind in 90 days. That is 90.7%.

Two cautions before you take that table anywhere. The 51–100 row covers seven companies, so its 42.9% is three of them; the 1–10 row is twelve companies. Those two rows are included for completeness, not as estimates of anything. The rows with enough companies to show a tendency are 11–50 and 500+, and the second is only 28.

Why this happens

The interesting part is not that big companies produce more. It is that they produce more through a completely different channel, and the split explains the silence better than the totals do.

11–50 employees (56 companies)500+ employees (28 companies)

Press coverage

1.8%

78.6%

Funding

0%

46.4%

Leadership change

0%

32.1%

Job postings

0%

17.9%

Three of the four channels are flat zero for the smaller bucket. Those bars are drawn as a hairline so a measured zero does not look like a missing row.

Large companies are carried almost entirely by the press. Nearly four in five had news coverage in the window; close to half had funding activity; a third had a leadership change reported. None of that requires the company to do anything unusual. It happens because other people write about companies of that size as a matter of routine.

For companies under 50 people, that channel is essentially closed. Press coverage reached 1.8% of that bucket: one company out of 56. Funding, leadership changes, job-posting activity: zero across the bucket. The one remaining channel that does not depend on a journalist is hiring, and that is where the second measurement comes in.

There is a related trap in the same data, worth naming because it distorts how people judge these tools. In the 500+ bucket, job postings covered only 5 companies out of 28, but those 5 produced 878 signals. Volume and coverage are different quantities. A feed can look busy because a handful of large employers are posting constantly, while most of the list has produced nothing.

This next part is interpretation, not measurement. We did not measure why the press writes about large companies and not small ones, and nothing in our data establishes a cause. The reading that fits the numbers is the ordinary one: coverage follows size and public visibility, so a signal source built on published events inherits whatever bias the publishers have. A tool that watches what publishes itself will always see loud companies best. That is a property of the source, not a defect in any particular product, and it is worth holding onto when you compare vendors: they are all drinking from the same well.

Is the silence real, or are we deaf?

This is the question that decides whether the first table means anything. A tool that cannot see an event reports the same nothing as an event that never happened, and the two are indistinguishable from the outside. So we checked.

Of the companies under 100 employees with zero hiring signals, every one had already been probed for a public job board, and none was found. The inverse also held: where a board was identified, we read it. Coverage there was 4 out of 4, and the share of companies where we found a board and failed to read it was 0%.

That still leaves the possibility that these companies were hiring somewhere we never looked. So the follow-up went outside the system entirely: take the 70 quiet companies, visit their sites, find the careers page, and look.

One. Reading every one of the five by hand: one had three roles listed openly, one was a genuine careers board printing "Results (0)", and three could not be judged because the listing was drawn by JavaScript or absent.

The honest answer is a range, not a number. Confirmed hiring we were blind to: 1 of 70, or 1.4%. If all three unreadable pages turn out to have roles: 4 of 70, or 5.7%. The absolute ceiling, counting all nine unclassifiable cases as hiring: 13 of 70, or 18.6%. Two things we could not measure (boards rendered entirely in the browser, and roles that exist only on the large job networks) can only push that number up, never down. Even at the ceiling, most of the silence is the world being quiet.

One more result deserves its own line, because it kills the obvious fix. Of the 34 careers surfaces we found, exactly zero sat on a job board with a public listing feed of the kind that can be read anonymously. The quiet companies were not hiding behind a naming mismatch we could have guessed our way past. Most were on staffing-industry systems that expose nothing publicly, or on hand-built pages. Better guessing would have returned nothing.

What this means if you track buying signals

Four things follow, and only the first is about tooling.

Calibrate by segment size before you judge anything. If your ideal customer has 30 employees, silence is the normal state of your list, not a failure. Our 11–50 bucket was silent 96.4% of the time over three months. Any expectation built on a weekly rhythm of events is going to be wrong for that segment, and the disappointment will get blamed on whichever tool is closest.

Judge a tool by how it handles nothing. This is the part we changed in our own product after this research. An empty account should tell you what was checked and when, in the spirit of "news checked 4 hours ago, hiring systems: checked, none found", so that "nothing happened" is distinguishable from "nothing was looked at". A feed that pads silence with weak items is optimising for the appearance of activity. Look at what a tool shows you on its worst day, not its demo.

For small-company segments, treat this as a patience game. If a rare event is the thing that matters, the value is in being early when one does fire, not in daily volume. That flips the metric: the useful question is not how many items arrived this week, but whether the ones that did arrive reached you while they still meant something. We have written separately about why timing beats volume in B2B sales, and this data is the uncomfortable version of the same argument: for some segments, timing is not just the better lever, it is the only one available.

Be skeptical of density forecasts. If a vendor tells you how many signals a segment will produce per month, ask what sample the estimate rests on and over how long. We tried to build exactly that forecast from our own data and could not. Split by industry and size, our companies fell into 68 groups. Exactly one had 30 or more companies in it, and that one was the group where both the industry and the size were unknown, which carries no predictive information at all. Five groups had ten or more. The largest meaningful group had 19 companies. That is not a modelling problem we could solve with a better model; the data to answer the question does not exist yet, in our set or, we suspect, in most.

If you want to know which signal types are worth watching in the first place, our guide to B2B buying signals covers what each one means and how long it stays useful. And if you are weighing whether monitoring a bounded set of accounts is worth it for your segment, the numbers above are the honest input: our pricing page sets out what depth of monitoring each plan covers, and this research is the reason we would rather you check your segment against that table than find out later.

Method, and what this does not show

The first measurement covered 277 companies we monitor, read-only, on 11 August 2026. Company size was known for 135 of them (49%) and industry for 161 (58%), so every split above describes about half the set. Signal counts were taken over a 90-day window.

That window needs a caveat that undercuts our own headline, so here it is plainly. The median company had been in our system for 5 days at the time of measurement, and the oldest for 21 days. For everything except news, "90 days" is really an observation window of three weeks or less. News is the exception because 82% of news signals were backfilled with dates preceding our records. So the table shows how much is findable about a company, not a rate per unit of time, and nobody should convert these figures into signals per month.

What this does not establish, stated in full: any causal claim about why the pattern exists; a density forecast for any segment; a rate over time; anything about the 142 companies whose size we do not know; and anything about the market at large. This is one product's set at one moment, and the companies in it came from the searches we ran, which is its own bias: 25 of the 277 companies sit in a single industry, staffing and recruiting, which was 96% silent and skews toward small firms. A different set would give different numbers. What would not change, we think, is the shape: published events cluster where publishing attention already is.

We would rather show the measurement with its holes than a cleaner number we cannot defend.