Engagement Scoring Is Useless If It Only Watches Email

September 29, 2026 · 4 min read · By Growth7

You send to four thousand contacts. Sixty people reply, a couple hundred click, and the rest do nothing visible. Next month you build the campaign again and send it to the same four thousand — same subject line logic, same cadence, same everything. The sixty who raised their hand get treated identically to the people who haven't opened anything since the list was imported.

That's the failure. Not the send, not the copy. The fact that nothing you learned in month one changed anything in month two.

Engagement scoring is supposed to fix this, and in most stacks it doesn't, for two reasons. The score is calculated from a single channel, so it's mostly noise. And nothing downstream is wired to read it, so it sits in a column nobody sorts by.

A Score Built From One Channel Is a Guess

Email opens have been unreliable for years — privacy proxies fire pixels for recipients who never looked. Clicks are better but sparse. If your engagement score is a weighted blend of opens and clicks, you're grading people on the one behavior that's hardest to observe honestly, while ignoring everything else they did.

Meanwhile the same person replied to an SMS two weeks ago. Or picked up an AI voice call and stayed on for ninety seconds instead of hanging up at hello. Or commented on a social post. Those are stronger signals than an open, and in a four-tool stack they're stranded in four different databases that never compare notes.

Growth7 runs email, SMS, AI voice and social against one audience, which means the score can read all of it. A contact who ignores every email but answers the phone isn't disengaged — they're channel-preferential, and a single-channel score will quietly bury them in a suppression list. The point of one shared audience isn't tidiness. It's that a score computed across it actually describes the person.

The tracking underneath matters too. Because opens, clicks and unsubscribes are tracked on infrastructure you control rather than a shared vendor domain, the signal feeding the score is yours and consistent, not filtered through whatever a third party decided to pass along.

The Score Has to Change What Happens Next

A number that only appears in a report is decoration. The question to ask of any scoring feature is: what decision does this make automatically?

Here's where it should show up:

Who gets the expensive channel. Voice calls and one-to-one follow-up are finite. Spend them on the top of the score, not the top of the alphabet.
Who gets contacted less. Low scores across every channel over a long window means backing off, not shouting louder.
Which new leads are worth working. Names sourced from Maps or LinkedIn arrive with no history; how they respond to the first touch is the first real data you have on them, and it should start accruing immediately.

The same logic applies to autopilot. Social and blog posting that runs on a schedule with human approval is only as good as its sense of who's out there. Scoring gives the approval step context — you're not just approving a post, you're approving it knowing whether this audience segment has been warm or silent for a month.

Ask the Score Questions in Plain Language

The practical barrier to using engagement data has always been that querying it requires knowing where it lives. With an MCP connection, an AI assistant can work against the real platform rather than a description of it — so "who in the Denver list has engaged on any channel in the last sixty days but hasn't been called" becomes a question you ask, not a segment you build by hand.

For operators running several brands, this only works if the separation is real. Scores are per-tenant. A contact who is hot for one brand you manage tells you nothing about their interest in another, and any system that blends them is producing a number worse than no number at all.

Start smaller than a scoring model. Pull the list of everyone who took any action on any channel last quarter, and check whether this quarter's campaign treats them differently from everyone else. If it doesn't, the score isn't the problem yet — the wiring is.