Lead harvest gives you the number you asked for

Asking for a number of leads now gets you that number. Duplicates and rejected records used to be counted against your total, so a request for five hundred could add three hundred and say it was finished.

What changed

A harvest pulls candidate records, checks each one, and adds the ones that pass. Some fail: a record can duplicate a contact you already have, or miss a required field, or fail validation.

The counting was done against candidates pulled rather than leads added. So a harvest that pulled five hundred and successfully added three hundred and twenty reported itself complete, having delivered two thirds of what was asked for.

Worse, the shortfall was invisible. The result said the harvest was done. Working out that it had under delivered meant counting the list yourself and comparing it to what you had asked for, which nobody does.

The count now refers to leads actually added. If you ask for five hundred and some candidates fail, more are pulled to make up the difference, and the harvest continues until the number is met or the source is genuinely exhausted.

Why it matters

A campaign built for a number of leads and given two thirds of it produces two thirds of the outcome, and the arithmetic afterwards is wrong in a way that is hard to spot. Reply rate is a fraction, and if the denominator is wrong then so is every conclusion drawn from it.

There is also a planning cost. Sizing outreach against a number that turns out to be notional makes every forecast built on it wrong by an amount that varies per harvest.

And a silent shortfall is corrosive in the way all silent failures are. Once you know a number is sometimes wrong, you stop trusting all of them, including the correct ones.

How to use it

Nothing changes in how you request leads. Ask for the number you want.

The result reports three figures: added, duplicate, and rejected. Added went to your list. Duplicate matched a contact you already have. Rejected failed validation.

Those three explain the whole harvest. If added is lower than you asked for, the reason is in the other two or in the source having run out.

Reading the numbers

A high duplicate count means you have harvested this segment before. Not a fault, and worth knowing, because it usually means the segment is smaller than you thought or that two lists overlap more than intended.

A high reject count means the source data was poor for this search. If it persists across searches, the criteria are pulling from a part of the source where records are thin rather than the search being wrong.

Both figures are more useful than the added count alone, because they explain it. A harvest that added three hundred out of five hundred with two hundred duplicates tells a completely different story from one with two hundred rejects.

When the source runs out

Sometimes the number genuinely is not available. A narrow enough search has a finite population and no amount of pulling produces more.

In that case the harvest stops and says so, reporting how many it found. This is different from stopping early, and the message distinguishes the two so you know whether to widen the search or investigate a problem.

The probe in AI Leads is the way to see this before spending anything. If a probe returns two hundred, no harvest is going to add five hundred.

Duplicates are not waste

A duplicate is a record you already have, which means the harvest correctly avoided creating a second copy of a person you are already talking to. That is the system working, not failing.

Sending the same opening email twice to one person is among the most damaging things outreach can do, because it announces that nobody is paying attention. Dedupe is what prevents that, and every duplicate reported is one instance prevented.

What changed is only the accounting. Duplicates are still skipped, they are just no longer counted as though they were delivered.

Credits and the count

You are charged for leads added, not for candidates examined. A harvest that pulled seven hundred candidates to add five hundred leads costs five hundred leads' worth.

This follows directly from counting the right thing. When the count was against candidates, the charging and the delivery could drift apart, and drifting in that direction is the kind of drift nobody wants to explain.

Duplicates and rejects are free, which is the correct arrangement: you should not pay for a record you already own or for one that failed validation.

Checking a harvest afterwards

Open the list and confirm the count matches. It should, and now it does, but the check takes seconds and the habit is worth keeping for the case where the source ran out.

Look at the duplicate figure across successive harvests of similar segments. A climbing duplicate rate is the earliest sign that a segment is nearly exhausted, well before a harvest actually runs short.

Where the numbers appear

The three figures show on the harvest result at the time it finishes, and they stay on the list afterwards, so a list you built last month still explains how it was assembled.

That history matters when a campaign underperforms. A list with a high reject rate at harvest time is a plausible explanation for weak results months later, and without the record it would be invisible.

Harvests started from Chat report the same three figures on the result card in the conversation.

Availability

Live now on all plans, applied to lead harvests from every source.