How do I measure a billboard, YouTube or radio buy before I fund it?

How does a contractor measure a reach channel like billboards, YouTube or radio?

Set the measurement before the buy: a unique offer or URL per broadcast, a geographic holdout for billboards, a dedicated CRM source and a branded-search baseline for video. Otherwise the CRM’s last-touch rule credits the revenue to Google and repeat, which is what happened to $136,380 of our streaming spend from 2023 to 2025.

What went wrong with our streaming buy?

We put $63,030 into OTT (streaming TV) in 2023, $9,000 in 2024 and $64,350 in 2025 at the eight-figure Toledo exteriors company where I direct marketing, and the CRM attributed $0 to it in each of those years. That does not mean it produced nothing. It means we bought a reach channel with no way to see it. Reach channels do not show up as a source a homeowner names on the phone, so whatever it did landed under “Google” and “repeat” and stayed there.

What was missing was not the channel. It was three things that cost almost nothing and had to be in place before the flight: a way for the CRM to see the channel at all, a second read that does not depend on the call center, and a baseline to compare against. Every reach channel a vendor will pitch this year needs the same three, and each one has a different version.

What does a vendor’s “33%” actually mean?

The last video pitch we took quoted a cost of marketing figure for their clients. Before a quoted rate means anything there are two questions, and neither is hostile. First, what is it measured on: their client average across which markets, with whose attribution, over what period. A number computed on last-touch in a market where the brand is already known is a different number from the one a new market will produce. Second, what is the reachable in-market audience at the planned frequency, because that sets the spend ceiling arithmetically. If the platform can reach 40,000 households in our zip codes and the plan is to hit them six times a month, there is a dollar figure above which the next dollar buys a seventh impression of the same household, and the vendor should be able to state it.

The only precedent that exists for a video buy at our company is broadcast TV, which has run between 32% and 41% cost of marketing every year since 2022 as the CRM attributes it. A vendor quoting a third is quoting exactly what we already do. That is not a reason to say no. It is a reason to measure.

How to measure broadcast or live sports

A unique offer per broadcast, or a unique URL, or both, so the call center has something to write down that is not “TV.” A “how did you hear about us” script that lists the game or the station by name. Homeowners will say “the ballgame” when they will not say “broadcast.” And branded-search impressions from Search Console pulled weekly, so the spike around a game window is visible whether or not anyone mentions it on the phone.

How to measure billboards

A geographic holdout. Put the boards in three or four of the priority zip codes and none in the matched rest, then compare zip-level net revenue against the prior year and against the holdout zips. The nearest measured outdoor channel we have is yard signs, at 8.2% cost of marketing on $30,419 of spend in 2024, and that is a different mechanism entirely, a sign on the lawn of a finished job in front of neighbors who just watched the crew. A board on the interstate has no such neighbor. Without the holdout there is no way to separate the board from the season.

How to measure video

Its own source in the CRM, from day one, so that a homeowner who says “I saw the video” is not booked as social. Conversion tracking on the ad platform, which for YouTube already exists inside the Google Ads account most shops run. And the branded-search line as the second read, the same as broadcast, because video that is working shows up as people typing the company name.

Where this is wrong

Holdouts need enough volume to read. A shop doing forty jobs a month across twelve zip codes will not get a clean signal from three of them, and in that case the branded-search line is the only read available, which has to be pulled before the flight starts to mean anything. And in a market where branded search volume is a few hundred a month, the noise will swamp a modest buy, which is a reason to test one channel at a time rather than three at once.

What to do

• Tonight: pull branded-search impressions by week for the last twelve months from Search Console. That is the baseline, and it does not exist until it is pulled.
• Before the buy: create the CRM source, write the intake question, pick the holdout zips, and get the vendor’s two answers in writing.
• Ongoing: read all three lines together at ninety days. If none of them moved, the channel either did not work or was too small to see, and either way the next dollar is not justified yet.

Sources: Lead Perfection “Cost of Marketing by Source,” Toledo, 2022 to 2025, first party. Vendor quote paraphrased; no vendor named. Audience-ceiling arithmetic is the author’s method.

Related notes