The evidence test: how to tell a real lead from a plausible one
Most lead lists are confident guesses. One question separates a prospect worth contacting from a row in a spreadsheet: what did you actually see?
A lead list is easy to produce and hard to trust. Filters return rows, rows look like progress, and nobody finds out the rows were wrong until a month of outreach has produced nothing.
There is one question that sorts them, and it is not about the company. It is about you.
What did you actually see?
Not what you inferred. Not what the filter implied. What you saw, in a place you could link to right now.
Inference is where lead lists go wrong
Most qualification is a chain of inferences that each sound reasonable and compound into nonsense.
The company is in construction. Construction firms deal with permits. Firms that deal with permits need help with filings. Therefore this firm needs help with filings.
Every step is plausible. The conclusion is a guess. And because the chain is invisible by the time it reaches a spreadsheet, nobody remembers it was a guess at all.
The evidence test breaks the chain. For every lead, you should be able to point at a specific artifact: a job posting, a thread, a page on their site, a filing, a public complaint. Something with a URL and a date.
No artifact, no lead. It is a company you have heard of.
Three tiers that are worth separating
Not all evidence is equal, and treating it as if it is produces the same problem in a subtler form.
Stated. Someone at the company said the thing out loud. A post asking for help, a job description naming the responsibility, a public complaint about the tool they use now. This is the strongest signal there is, because it removes the guess entirely.
Structural. Something about the company makes the problem near-certain even though nobody said it. They run twelve locations with no regional manager. They are hiring their first salesperson. The problem follows from the shape of the business.
Circumstantial. They match a pattern that correlates with buying. Right industry, right size, right region. This is where most lead databases stop, and on its own it is close to worthless.
Sort your list by tier and work top down. A day spent on ten "stated" leads beats a week on two hundred "circumstantial" ones, and you will find out faster whether your positioning is right, because the people you are talking to actually have the problem.
The date matters as much as the signal
A hiring post from last week and one from fourteen months ago are not the same signal, even though a filter treats them identically. The role is filled. The project shipped. The person who complained switched tools and forgot about it.
Anything older than about ninety days should be treated as circumstantial regardless of how strong it looked when it was fresh. Recency is not a tiebreaker, it is part of the evidence.
Write the reason down
The habit that makes all of this stick: for every lead you keep, write one sentence naming what you saw and why it matters. It is the same sentence an ideal customer profile is built out of. If the sentence is hard to write, the lead is weak. If the sentence turns out to be "they are in the right industry," you have found a row pretending to be a prospect.
That sentence is also your first line of outreach, which is a useful side effect. The work of qualifying and the work of writing turn out to be the same work.
What this costs you
Fewer leads. Considerably fewer.
That feels like going backwards, especially if you have been judging the week by how many rows you added. It is also the standard any AI outbound tool should be held to. But a pipeline of forty leads you can each justify converts better than four hundred you cannot, and it tells you something true about your market either way.
If you apply the evidence test and find almost nobody passes, that is not a failure of the method. That is the market telling you something about your positioning that the four hundred rows were hiding.