How do I tell a real buying signal from press-release noise?
Score it against five tests and act only above a threshold you set in advance. A real signal is dated, sourced to a primary record, creates a consequence you can state in one sentence, has an owner you can name by role, and is specific to your category rather than to everyone selling anything. Press-release noise usually passes the first two tests and fails the last three, which is exactly why it fools people: it has a date and a link, so it looks like evidence.
The test that does the most work is the consequence test. Ask what this event makes operationally harder, for whom, in the next thirty to ninety days. If the honest answer is nothing in particular, you are holding news. That is not a judgment about the company or the announcement, it is a statement that the event does not create work in your area, which means there is no reason for anyone there to take your call this month.
The five tests
Each test is scored zero, one, or two. Zero means the test fails outright. One means partial or inferred. Two means clearly satisfied from the evidence in front of you. Total is out of ten, and the point of scoring rather than judging is that it forces you to be explicit about which part is weak, which is what makes the score arguable and therefore useful.
- Dated: does the evidence carry a real event date, not a publication date or the day you found it?
- Sourced: is there a primary record, a filing, a company newsroom post, an official register, rather than an aggregator summarizing an aggregator?
- Consequence: can you write one sentence naming what this makes harder, in the next thirty to ninety days?
- Owner: can you name the role that absorbs that consequence, without having looked up a single person yet?
- Specific to you: would this event score differently for a vendor in a different category? If every vendor would score it the same, it is a general news item.
The rubric
| Row | Test | 0 points | 1 point | 2 points |
|---|---|---|---|---|
| 01 | Dated | No date, or the date is when you found it | Approximate period, such as a quarter | Exact event date from the record |
| 02 | Sourced | Aggregator or unattributed post | Trade coverage citing the company | Primary record: filing, register, or company newsroom |
| 03 | Consequence | Cannot name what gets harder | Plausible but generic, would apply to most companies | One specific sentence about this company in the next 30 to 90 days |
| 04 | Owner | No role identifiable | A department, but not a role | A specific role that absorbs the consequence |
| 05 | Specific to you | Every vendor would score it the same | Relevant to your broad category | Maps directly to the problem you solve |
Set the threshold before you score anything, so you are not adjusting it to justify an account you already like. A reasonable starting point is eight or above for a human-written first touch this week, six or seven for a watch with an alert set, and five or below for nothing at all. Then hold the line for a quarter and look at what came back, because the threshold is the number you should be tuning with outcome data, not the tests themselves.
What high-scoring noise looks like
The dangerous items are the ones that score well on the easy tests. An award announcement is dated and sourced to the company newsroom, so it takes four points before you have thought about it, and then scores zero on all three tests that matter. A partnership announcement with no deliverable does the same. A funding rumor fails the source test but reads like a scoop, which is worse, because people forward it.
- Awards, rankings, and best-places-to-work lists: dated and sourced, no consequence, no owner outside marketing.
- Rebrands and site relaunches: real work, but it is finished, and the budget went with it.
- Partnership announcements with no product, revenue, or delivery commitment named in the release.
- Thought-leadership posts and executive commentary: tells you what someone thinks, not what the company must now do.
- Anniversaries, headcount milestones, and office openings without a stated capacity or hiring plan attached.
The other failure mode: real events you cannot use
There is a second category worth naming, which is genuine, consequential events that are simply not yours. A manufacturer opening a second plant is a major event with a clear consequence and a clear owner, and if you sell legal intake software it should score two, two, two, two, zero. The specificity test is what catches it, and it is the test most teams leave out because it feels like it is about you rather than about the account.
Leaving it out is how signal feeds turn into news feeds. Once a team is receiving every real event at every account, the volume forces reps back to skimming, and skimming restores exactly the gut-feel selection the system was meant to replace.
This is what the category calls signal quality and false positives.
Vendors talk about signal quality, precision, and false positive rate, and those words describe the thing this rubric measures. A false positive in this context is not a factually wrong event, it is a factually correct event that should not have reached a rep, and it is the more common failure by a wide margin.
There is no industry standard for what counts as a signal, which is why comparing vendors on signal count is meaningless: a product that counts every press mention will always report more signals than one that discards developments it cannot attach a consequence to. Ask what gets thrown away, not how much gets delivered.
Questions people ask next
Is a press mention ever a signal?
Yes, when the coverage contains an operational fact rather than a promotional one. A profile that states the company doubled its client count this year names a consequence. An article celebrating an award does not. Score the content, not the format.
Who should set the threshold, the rep or the manager?
The manager sets it and the team argues with it monthly using outcome data. If reps set their own thresholds you lose the ability to compare, and if it never moves you are not learning from what came back.
Does an intent surge score on this rubric?
Usually poorly. It rarely carries an event date, the source is inferred rather than primary, and it names no consequence and no owner. That does not make it useless, it makes it a tiebreak rather than a reason.
Where to check this yourself
Primary records and published research, not vendor blog posts. Every link was checked on September 1, 2026.
- SEC EDGAR full-text searchsec.gov
The primary-record test in practice: search the filing text yourself rather than trusting a summary.
- SEC list of filing formssec.gov
Which form carries which event, so you can tell a reportable corporate event from a press item.
- Federal Registerfederalregister.gov
Rules and effective dates, an example of a primary register that is dated, sourced, and carries a real deadline.
- Electronic Code of Federal Regulations, WARN Act regulations, 20 CFR Part 639ecfr.gov
An example of a public rule that generates dated, verifiable company events rather than press coverage.
- Crunchbase data documentationdata.crunchbase.com
Shows how a commercial database sources and structures events, useful when judging whether a feed is a primary or secondary source.
Intakra applies this filter before anything reaches you: a development is dropped unless it can be tied to a dated public source, an operational consequence, and the role that absorbs it.
Free scan, no signup, no card