Every Contact Is a Guess Until It Behaves

You probably have three lists right now. One is a spreadsheet a contractor scraped for you, sorted by city. One is an export from a CRM you stopped maintaining two years ago. One is a pile of badge scans from a trade show. At some point you loaded all three into an email tool, sent the same announcement to everyone, got a handful of replies, a wave of unsubscribes, and no way to tell which list produced either. Next quarter, you'll probably do it again.

The problem isn't that the lists were bad. It's that a sourced contact is treated as an asset — a thing you own, a number you can report — when it is actually a hypothesis waiting to be tested.

A Sourced Lead Is a Hypothesis, Not an Asset

When you pull businesses from Maps by category and radius, or people from LinkedIn by title and industry, what you get is fit. This company is the right type, in the right place, and the person has the right job. That is genuinely useful, and it beats guessing. But fit is a claim about whether someone should care, not evidence that they do.

The hypothesis only resolves when the contact does something: opens, clicks, replies, stays on an AI voice call past the first sentence, engages with a social post, or unsubscribes. Every one of those is a data point about interest, and none of them exist at the moment of sourcing.

This is the practical case for sourcing inside the same platform that sends. When Growth7 pulls leads from Maps or LinkedIn, those records land in the same audience your campaigns run against, so they start accumulating behavioral evidence from the first touch. When sourcing happens in a separate tool and gets exported as a CSV, you sever the record from its own history — and you inherit the same blind spreadsheet you started with.

Scoring Is Only Honest If It Can See Every Channel

Most engagement scores are really email scores wearing a costume. They're built from opens and clicks in one tool, because that's the only behavior that tool can see. So a contact who ignored three newsletters but tapped an SMS link and sat through a voice call gets labeled cold — and the next campaign politely stops talking to your most interested prospect.

A score is trustworthy in proportion to what it can observe. If email, SMS, AI voice and social all run against one audience, engagement scoring has something real to work with: not "did this person open an email" but "how has this person responded to us, in any form, over time." It also matters that the open, click and unsubscribe tracking is self-hosted rather than borrowed, because a score assembled from someone else's reporting is only as durable as your contract with them.

Once the score is honest, it changes ordinary decisions:

• Who is worth an AI voice call, and who should stay on email for now
• When to stop contacting someone before they unsubscribe for you
• Which sourced segment — this city, this job title, this business category — actually converted, so you know what to source more of
• What a salesperson opens with, because they can see what the person already engaged with

None of that requires a data science project. It requires the behavior and the sourcing to be in the same place.

Operators: Don't Let One Brand's Baseline Define Another's

If you run marketing for several businesses, there's a second trap. Engagement looks different for a restaurant than for a B2B manufacturer with a nine-month cycle. A score that averages both is a score that misleads both. Multi-tenant separation isn't only a privacy feature — it's what lets each brand develop its own baseline for what "warm" means, while you still work from one login and, if you're between meetings, one iPhone app.

The agencies that get this right stop selling list volume and start selling something harder to copy: a growing record of who in each market has actually responded, and to what. That record can't be bought. It only accumulates, in one place, over time.

Sourcing is cheap and getting cheaper. Attention is the scarce thing. The best list you'll ever have isn't the one you bought — it's the one you've been watching.