A lead is a person, not a touch

If you count conversion events, you are counting a thing your business does not care about.

Your business cares about people: how many new humans made contact this month, what they wanted, and what it cost to reach them. Conversion events are a proxy for that, and like most proxies they are fine until you look closely.

Look closely and about a third of them are not new people.

Where the third comes from

This is not a theoretical inflation figure. It is what deduplication removes from raw contact records in live data, and it comes from two specific behaviours that anybody who has ever run a phone line will recognise immediately.

Repeat callers

Somebody calls. It rings out, or it goes to voicemail, or they get put on hold and give up. Twenty minutes later they call back.

Your call tracking platform records two calls, because two calls happened. Your reporting counts two leads, because it was handed two records. One person wanted one thing.

The worse version: they call, do not get through, and call back from a different number the following morning. Now the deduplication has to be smarter than matching on caller ID, and most reporting does not attempt it at all.

Channel switchers

This one is more damaging because it also corrupts your channel attribution.

Somebody finds you through paid search and fills in your contact form. Nothing happens fast enough for them, so they find your phone number and call. The call is recorded as direct, or as organic, or as whatever the last-touch model decides.

You now have two leads in your report, credited to two different channels, for one person and one job. Your paid channel gets a conversion. Your direct channel gets a conversion. Your actual new customer count is one.

This is common enough that if your reporting does not handle it, your channel mix is not describing your channel mix.

Why the tracking is not broken

It is tempting to treat this as a data quality problem to be fixed at the source. It is not.

Every one of those events genuinely happened. The call tracking correctly recorded two calls. The form platform correctly recorded a submission. Each system did its job accurately within its own boundary.

The problem is that nothing owns the question “how many distinct people were these.” That question sits above every individual platform, and by default nothing answers it, so the numbers get summed instead.

Summing is the failure. Not the collection.

What a defensible definition looks like

Here is the definition worth writing down, and worth holding your reporting to:

A new lead is a distinct person whose first contact with the business falls inside the reporting window.

Three parts of that sentence are doing work.

Distinct person. Not a session, not a submission, not a call. Matched across channels using whatever identity evidence exists: phone number, email address, name plus company, the content of what they asked for.

First contact. If somebody first contacted you in January and contacted you again in March, they are not a new lead in March. They are an existing relationship. Counting them again inflates March and hides the fact that you did not add anybody new.

Inside the window. Which means a lead belongs to the period they arrived in, not the period they converted in. Otherwise your monthly numbers move for reasons that have nothing to do with what you did that month.

The uncomfortable consequence

Apply this properly and your lead numbers go down by about a third. Every one of them. Retroactively.

That is a genuinely awkward conversation, and it is worth naming why it is awkward: somebody has been reporting the larger number, in good faith, possibly for years. Correcting it looks like an accusation even when it is not.

The way through is to publish the definition, apply it to the full history rather than just from today forward, and show both series side by side for a period. The shape of the trend usually survives intact. It is the level that moves. Once people can see that the line still goes the same direction, the conversation stops being about blame and starts being about the denominator, which is where it should have been.

Where credit goes

Once you are counting people rather than touches, one more decision follows immediately: which channel gets credit for a person who touched three of them?

The answer that describes reality is the first touch. What actually brought this person into your world.

Last-touch attribution, which is the default nearly everywhere, credits whatever they clicked immediately before converting. For anyone who took more than one session to decide, that is very often branded search: they already knew your name, they typed it in, they clicked. Crediting the acquisition to that click tells you almost nothing, and it systematically starves whatever created the demand in the first place.

First touch is not a perfect model. Nothing is. But it errs toward explaining where demand came from, and that is the question you are actually asking.

The excluded records belong on the page

One more thing, because it is adjacent and it is skipped almost universally.

Every business with a public phone number and a contact form receives inbound sales solicitations. SEO agencies, staffing firms, software vendors, all arriving through exactly the same channels as your real enquiries.

They must be excluded from lead counts. That part is obvious.

What is less obvious is that the excluded count should be displayed. If you filter 40 records out of a 130-record month and never mention it, you are asking to be trusted about a filter nobody can inspect. Show the number. It is auditable, it takes one line, and it converts “trust me” into “check me.”

What this actually costs you to fix

Nothing, if the underlying records exist. This is not a data collection problem, it is a counting problem, and the raw material is usually already sitting in your call tracking platform and your form submissions.

What it requires is a place where those records meet, a matching rule you are willing to publish, and the discipline to apply it in both directions rather than only when the number is inconvenient.

That is what the lead record is for. Every call and form in one table, with the deduplicated rows and the excluded rows visible rather than silently dropped, because a filter you cannot see is a filter you have to take on faith.