Test with informed participants, then put every product through an approval gate

Pre-test with people who know they are testing

Doctrine separates making a product from releasing it: phase five exists so that nothing goes out until someone with authority, who did not write it, has approved it. Before approval, find out whether the products work, and do it with people who know they are helping you test. Recruit participants from the audience, tell them who you are, who is paying, and what the test is for, pay them for their time, and let them stop at any point. A dozen interviews in which people read a product aloud and tell you what it means will catch most problems.

If you compare versions in a live setting, every version must already be truthful and attributed, and the comparison must be one you would describe to the audience without discomfort. Experiments that vary how much pressure or distress people are exposed to, without their knowledge, are not testing. They are the harm.

Choose the metrics before you look

Write down what you will measure, and what result would make you change or drop a product, before you see any data. Measure understanding first: can people say what the product asks, where, when, and at what cost? Then measure what they intend to do, whether they felt pressured or talked down to, whether they could tell who it came from, and how they now feel about people who disagree.

Look for backfire in subgroups. In a randomised trial of vaccine messages, Nyhan and colleagues found that corrective content reduced a misperception and yet lowered intention to vaccinate among the parents with the least favourable attitudes. A test that reports only the average, or only the metric that came out well, is a test designed to approve.

Red-team the whole package

Give everything to people whose job is to break it. A red team needs at least one sceptical member of the audience, one person from a group that could be hurt by the campaign, and one person who understands the law. Ask them four questions. What is the most misleading reading of this? Which claim is weakest? Who could this harm, including people outside the audience? What would a fair critic say on seeing the full plan, the targeting, and the budget? Record every objection and what you did about it.

The approval gate

Approval is a decision by a named person who did not create the products. Two tests make the decision concrete. The TARES test (Baker and Martinson) asks whether the message is truthful, whether the persuader is authentic, whether the audience is respected, whether the appeal is equitable given what the audience knows and can check, and whether the campaign is socially responsible. The front-page test asks whether you would be content to see the complete plan (messages, targeting, messengers, payments, and funders) reported accurately in the newspaper your audience reads. A product passes both or it does not go out.

Legal review

Have counsel check the attribution and paid-for-by lines, the disclosure wording for anyone compensated, consent for every contact list and for text messaging, image and quote releases, and any rules specific to your subject. If the campaign touches an election or a ballot question, campaign-finance registration, reporting, and disclaimer requirements apply and vary by jurisdiction; get that advice in writing before launch. Legal clearance is a floor. It does not replace the two tests above.