The offline campaign measurement case studies worth studying twice

What to take away

Four cases, and what each one is

Be Clear on Cancer (Public Health England, 2010 to 2013). This national bowel-cancer symptom campaign ran in phases across press, radio, and outdoor advertising. Its message told people to watch for blood in their stool. The UK government case study sets out the media plan, the audience, and the measurement window.

Awareness was tracked before, during, and after each burst, so the design hands you a baseline instead of a single peak number.

ONS trust-barrier evaluation (UK Office for National Statistics). The ONS tested whether trust-barrier communication for social surveys changed response. Randomization was not feasible, so the team used a theory-based contribution approach and said so in the report. Read it for how to write up a limitation instead of hiding it.

CDC Tips From Former Smokers (2012 onward). A paid mass-media campaign with a national buy and a published evaluation. CDC reported an estimated 1.6 million quit attempts during the 2012 wave, and calls to the 1-800-QUIT-NOW quitline more than doubled. Its scale and its public-health outcome measure are the two things a small advertiser cannot reproduce, which is exactly why it belongs on the list.

Canada Post and Nielsen mail study (2016). A neuro-measurement study on how consumers process addressed mail against digital channels. It used eye-tracking and brain-response measures rather than sales records, so it reports attention and recall. Treat it as evidence about attention, and nothing more.

Three more with usable write-ups

Green and Gerber voter turnout experiments (1998 onward). Donald Green and Alan Gerber assigned households at random to receive mail, a commercial phone call, or a face-to-face visit, then checked the voter file for turnout. The random assignment is the point: the untreated group answers the counterfactual your average case study cannot.

Their published results put face-to-face canvassing at several percentage points of extra turnout, and commercial phone calls at close to zero. Mail effects were small and moved with message and timing. Personal contact drove the effect more than the budget did.

Truth campaign (American Legacy Foundation, 2000 onward). The Truth youth anti-smoking campaign bought national paid media and tracked exposure with the Legacy Media Tracking Survey. Published analyses found that youth in markets with heavier campaign exposure were less likely to progress to established smoking. The design is worth copying because it measured dose of exposure, not just awareness.

Royal Mail MarketReach mail research (2015 onward). MarketReach's Private Life of Mail project used eye-tracking, biometrics, and household interviews to study how UK consumers handle physical mail. It measured attention, emotional response, and where mail was kept at home, not sales. Use it for how mail gets processed, not for what it returns.

Read the method before the number

Each of these cases states its audience, period, media, and evaluation design. That is what makes them usable. A result without those fields is a headline you cannot check. For a fuller treatment of the designs and their trade-offs, see the complete guide to offline campaign measurement for 2027.

The ONS case is the useful model here. When randomization is off the table, the honest move is to name the design you used and the explanations you could not rule out.

Audit the evidence before you cite it

The FTC advertising substantiation policy requires a reasonable basis for objective claims before they run. If you repeat a case's performance figure in your own material, that standard applies to you, not to the original publisher. Confirm the current position with the FTC or a qualified advertising lawyer.

The FTC guidance on soliciting and paying for reviews covers fake feedback, selective requests, conditioned incentives, and hidden relationships. Apply the same checks when a vendor hands you a case study to quote.

Separate four things: what the provider claims, what the customer says, what the records show, and what an independent party found. If the method is missing, file the item as a lead to test, not a verified result.

Worked example: transferring one control

Suppose a mail campaign reports a lift in response. Your market differs in list quality, offer, season, and price.

Write the arithmetic in your own variables: response rate (R), pieces mailed (N), cost per piece (C), and value per response (V). Net = (R x N x V) minus (N x C).

Substitute your figures, then run one cell of the campaign as a test before committing the budget.

That is the transferable part: the design control, not the result. If you want to know which parts of the process deserve that effort, start with how to measure offline campaign measurement: the parts worth your time.

Common questions

Can a government case predict my results?

No. It can show you a method and an operating lesson. Audience, budget, authority, and incentives will differ from yours.

What is the strongest thing to copy?

A design control or a research question that held up under scrutiny. Then test it under your own conditions.

What if the case gives no method?

Treat it as a lead. Ask the publisher for the design, and if it is not available, do not cite the number. The recurring sticking points are collected in what offline campaign measurement questions keep coming up.

Where does the current guidance live?

Advertising substantiation and review rules sit with the FTC in the US. Postage rates and mail specifications sit with Canada Post. Check both before you commit spend.