Blog · Signal
A mention is not a buyer
Set a keyword alert for your category on X and you will get volume immediately. That volume feels like traction for about four days. Then you notice that almost none of it is a person who might pay you, and the alert becomes another notification you swipe away.
The problem is not the alert. It is that a keyword match and a buying signal are different things, and no amount of tuning your keywords will turn one into the other. Monitoring tools were built to count how often a brand name appears. That is a useful number for a consumer brand tracking reputation. It is close to worthless for a founder trying to find the next ten customers.
The eight false positives
Run a raw keyword search on any software category and this is what actually comes back, roughly in order of volume.
- Vendor marketing. Companies in your category posting about themselves, plus their agencies, resellers and affiliates. In crowded categories this alone is half the results.
- Jokes and memes. “Me pretending to update the CRM before the Monday call.” Sarcasm reliably defeats sentiment analysis, which will happily score this as negative brand sentiment and hand it to you as an opportunity.
- Job posts. “Hiring a Salesforce admin, remote, apply here.” Your competitor’s name, a company that clearly uses them, and zero chance of switching this quarter.
- News and funding. Round announcements, acquisitions, launches, and the two hundred accounts quoting them.
- Engagement bait. “What tool could you not live without? 👇” Hundreds of replies, one person farming impressions, no buyers.
- Thought leadership. Consultants and creators explaining the category to their audience. They talk about the problem professionally; they do not have it.
- Resurfaced threads. A quote-tweet of a question from eighteen months ago. The decision was made long before you arrived.
- Outright spam. Giveaways, engagement pods, and in some categories a wall of wallet-drainer bots.
What is left after those eight is usually somewhere between one and ten percent of the raw feed. That remainder is the only part worth reading, and finding it by hand is most of the daily work described in the guide to finding buyers on X.
Three questions that settle it
For anything that survives the first pass, three questions decide whether to spend a reply on it.
Is this a person, or an account?
A named human with a bio describing an actual job is a lead. A company account, a newsletter, a roundup bot, or a 200,000-follower creator who covers the category for a living is not — they are in the business of talking about this, not living it. Follower count is a weak proxy but a useful one: above roughly fifty thousand, you are usually looking at a publisher rather than a buyer.
Are they in it, or thinking about it?
Tense and specificity give it away. “We’re moving off this quarter, team of six, need something with an API” is committed and detailed. “One day I should really look at alternatives” is a mood. Concrete detail — a team size, a deadline, a named failure, a price — is the single best predictor that someone will actually reply to you.
Can you genuinely help?
This is the one people skip, and it is the one that costs the most. If their problem is adjacent to what you sell but not quite it, replying burns credibility in public. The strongest thing you can do in a category conversation is occasionally say “honestly that’s not us, look at Z” — people remember it, and it is the cheapest reputation you will ever buy.
Score in bands, not in decimals
Once you are scoring, resist false precision. If nothing about your behaviour changes between a 78 and an 86, the number is decoration. Three bands are enough:
- Answer today. A person, in the market, with a problem you actually solve. Drop what you are doing.
- Worth a look this week. Real person, real problem, unclear timing. Read the thread before deciding.
- Watch or ignore. On-topic but not a buyer. Follow them if they are interesting; do not reply.
A fourth band is worth keeping privately: never again. Accounts that turned out to be aggregators, bots or competitors. Excluding them permanently is the highest-leverage tuning you can do, and it compounds every week.
Why intent beats sentiment
Most listening tools score sentiment: positive, neutral, negative. It is the wrong axis for selling. A furious post about a competitor can be an excellent lead or a pointless rant, and sentiment cannot tell you which. A perfectly neutral post — “what are people using for X these days?” — is often the best lead of the week.
What matters is not how someone feels. It is whether they are about to spend money, and whether you are a plausible answer. Those are the two things worth scoring, and the reason a mention count and a lead count are different numbers entirely.
Where the noise goes
Two habits keep the signal clean over months rather than days.
First, write your exclusions down. Every noisy handle, every recurring bait phrase, every job-post pattern belongs in your query as a -term, permanently. A search that is not getting more precise every week is getting worse.
Second, mark the misses. When something looked good and turned out to be nothing, record why. Over a couple of months, that record is a far better description of your real buyer than any persona document.
The discarding is where the cost of this channel actually lives — the arithmetic is here. Askline does both of these automatically: it filters the eight categories above before you see anything, scores what is left by intent rather than sentiment, and learns from what you mark as not a buyer. Run it on your offer and see what a filtered feed looks like next to a raw keyword alert.
Next: why the first useful reply usually wins, then how to write one that does not read as an ad. If you want to sharpen the queries that produce this feed in the first place, start with the operator reference.
If you would rather ask the question than run a search, the same thing works from Claude through an MCP server.