Measuring what matters: attribution for political campaigns beyond impressions.

Most political media reports we see still lead with the same number: impressions delivered. It's the easiest metric to produce and the one that means the least. Nobody on a campaign team has ever lost sleep over an impression that didn't happen. They lose sleep over donations that didn't come in, voters who didn't show up, and volunteers who didn't sign up.

The point of modern measurement is to connect paid media to those outcomes — the ones that move the campaign — and to do it inside the same week the spend happened, not three months after Election Day.

The four outcomes that actually matter

Different campaigns weight them differently, but the universe is small: donations (every campaign), event RSVPs and attendance (rallies, town halls, fundraisers), volunteer and field conversions (door knocks signed up, shifts filled, voter contacts made), and vote outcomes (early vote pull-through, day-of turnout among targeted households).

An attribution model that doesn't tie back to at least one of those four isn't a political measurement framework. It's a media report.

Closing the loop with the voter file

Because addressable CTV serves at the household level — and the voter file knows which voters live in which household — the loop closes naturally. The same households that received CTV impressions can be matched back to early-vote rolls, donor lists, RSVP data, and volunteer signups after the fact.

The output isn't "we delivered 12 million impressions." The output is: of the 80,000 persuadable households we reached on CTV, 31% turned out for early voting, compared to 19% in a matched control group of unexposed households. That's a number a campaign manager can act on.

Nobody on a campaign team loses sleep over an impression that didn't happen. They lose sleep over donations that didn't come in.

Match controls or it doesn't count

The single most common attribution mistake in political media is reporting on the exposed group without a matched control. "63% of households who saw the ad voted" tells you nothing without "47% of comparable households who didn't see the ad also voted." The lift number is the only number that matters, and it requires designing a holdout into the buy from day one — not bolting one on after the flight.

On every campaign Adloop runs, we hold out a randomized portion of the addressable universe from the start. That's the control. Everything we report on lift — turnout, donations, RSVPs — is measured against that holdout. It's not optional. It's the difference between a measurement framework and a marketing slide.

Weekly cadence, not post-mortem

The reason most campaigns don't measure properly isn't that they don't believe in measurement. It's that the measurement they get arrives too late to do anything with it. A modern CTV program produces directional reads inside a week — sometimes inside 72 hours during the closing stretch.

That cadence changes how a campaign operates. Creative variants get pruned mid-flight instead of mid-cycle. Underperforming geographies get rebudgeted before the next wave airs. Persuasion universes get tightened or expanded as new data comes in. Measurement stops being a report and starts being an operating discipline — which is the only role it was ever supposed to play.