Every funder deck we've ever been shown from another agency leads with impressions. "Reached 42 million people." "Delivered 15 million ad views." "500,000 unique impressions." The number is always in the seven figures. The number is always impressive. The number is almost always meaningless.

An impression is a browser or a phone loading a thing for a fraction of a second, whether or not anyone actually looked at it. Half of impressions happen off-screen. A third happen to bots. The "42 million people" figure counts the same teenager scrolling past your ad forty times.

But the number is easy to generate — a modest programmatic budget will manufacture almost any impression count you want. That's the problem. If a metric can be bought without producing anything real, it stops being evidence. It becomes a receipt.

What we count instead

When we scope a campaign or a build, we agree three questions with the client before we start:

  1. What behaviour are we trying to move? Signing a petition. Making a callback. Enrolling in the programme. Turning up to the health camp.
  2. What population is it? Not "everyone in India." A specific set of people, geographically or demographically or by need.
  3. What's the counterfactual? What would that population have done without us? If we can't estimate this even roughly, we can't claim impact — we can only claim activity.

The metrics we report follow from those questions. They're usually small numbers. They're usually harder to game.

Rule of thumb: if the metric would improve because someone wandered onto the site by accident, it's not an impact metric. It's a traffic metric. Traffic isn't impact. It's the raw material impact might be made from.

Three numbers we like

Behaviour, not attention.

"312,000 petition signatures." "1.2M IVRS calls completed." "84 members co-designed the new programme." These are things people did. You can't fake a signature, complete a 9-minute voice flow, or sit for a 40-minute focus group without meaning to.

Retention, not reach.

"38% of callers came back for a second call." Return is harder to buy than first-touch. A single interaction proves a poster works. A repeat interaction proves the thing on the poster works.

Refusal, not accolade.

"9% of members told us they'd stopped attending because of the session timing." Uncomfortable numbers are often the most valuable ones. A dashboard that only carries success metrics has been designed to reassure funders, not to help you learn.

What we still say about reach

We do sometimes need a reach number — funders live in the world where those numbers exist and we can't wish that away. When we quote one, we caveat it. "Reached 1.5 million people via campaign channels — of whom 312,000 signed the petition." The petition number is the honest one. The reach number is context.

If a funder pushes for "impressions" as the primary number, we push back once. If they still want it, we deliver it and we tell you what it means and what it doesn't. But we've never lost a funder for pushing back. Usually the opposite — they're relieved someone finally said it out loud.