The pre-launch compliance sign-off, step by step

It is 4:50pm on a Thursday. A regional brokerage wants a listing campaign live at 9am: a Meta carousel, a portal-syndicated listing, an email to 12,000 subscribers. Three people have already looked at the creative. None of them was checking the same thing, and none of them wrote down what they checked. The headline promises guaranteed appreciation, and the hero image is a rendering that does not match the delivered finishes.

The rest of this module explains why that campaign is a problem. This lesson is about the forty minutes that catch it before 9am, and about the harder version: the same campaign, live since Monday, and a screenshot of it already circulating.

Order the gates by cost of rework

A sign-off is not one review. It is a sequence, and the sequence matters because each gate can invalidate the work downstream of it. Run them in the order that kills bad work early.

Gate 0, at brief stage, before anyone books a photographer. Two questions: can the central promise be evidenced at all (the standard the claims lesson sets), and is the audience construction one you are allowed to buy (the targeting restrictions the fair-treatment lesson covers). A campaign whose whole idea is "guaranteed 8% yield, targeted to people like our best investors" dies here, at the cost of a meeting. Discovered at Gate 4, it costs a shoot, a media plan and a launch date.

Gate 1: the evidence pack is attached, not promised. Every substantive claim in the copy carries a document reference: the survey for the floor area, the dated comparables for the price framing, the source for the school or transport claim. If the reviewer has to ask the agent where a number came from, the campaign is not ready for review.

Gate 2: the disclosure artefacts, checked against the jurisdiction of the property rather than the jurisdiction of the office. The list is the one the disclosures lesson gives you. What sign-off adds is the trap: a template built for one market and reused in another, or an agent who changed firms last month and whose creative still carries the previous registration. Both pass a copy read. Neither passes a check against the file.

Gate 3: platform configuration, reviewed separately from creative. Copy can be immaculate while the audience is wrong. Meta's housing flow and Google's restrictions on personalised advertising for housing sit in the campaign settings, not in the ad. Someone has to open the settings and look. Note that Meta and Google both sell the inventory and write the policy, so a green tick from either is an inventory decision, not a legal opinion. Their approval does not transfer liability to them.

Gate 4: the asset that ships is the asset that was approved. This is the gate people skip, and it fails more often than any of the others.

Who signs, and what they are signing

Four signatures, each attached to a sentence rather than a vibe:

• The listing agent signs that the property facts match the file as at today's date.
• The compliance reviewer signs the fair-treatment screen and any financing or return claim.
• The marketing lead signs the platform configuration, the channel mix and the destination page.
• A final approver signs the release, and holds the only authority to delay a launch.

If nobody can hold a launch, you do not have a sign-off, you have a notification. The test is simple and worth tracking: how many launches did the final approver delay last quarter? A hold rate of zero over three months means the gate is decorative, or that everything upstream is genuinely excellent. It is almost never the second one.

Set an operational cut-off rather than negotiating each time. Creative submitted after 3pm launches the following day. That single rule removes most of the 4:50pm scenarios, because the pressure to wave something through comes from the clock, not from the content.

Where it is logged

One record per creative version, with: the version identifier or file hash, the platform preview link (not a PDF of what the ad was meant to look like), a screenshot of the audience and placement settings, the evidence document references, the four named signers with timestamps, and an expiry date for any time-limited claim such as a price, an incentive or an availability status.

Retention matters more than people expect. A campaign can stop running long before the consequences arrive, and in the UK an Advertising Standards Authority ruling is published on asa.org.uk and stays there, indexed under the advertiser's name, long after the ad itself is gone. The ASA cannot fine, but it can ask platforms to remove paid ads and can list persistently non-compliant advertisers publicly. The reputational half-life of a ruling is much longer than the campaign that caused it. Separately, repeated policy violations on Google can escalate from disapproval of one ad to suspension of the account, which takes your compliant campaigns down with the bad one.

The log is what converts a complaint into a correction. A dated, signed record showing what was approved, by whom, on what evidence, changes the conversation from "was this deliberate" to "here is where the process failed and here is the fix".

🎬 [VIDEO: "Fair Housing Advertising: What You Can and Can't Say" - youtube.com - a HUD-aligned training walkthrough of compliant vs. non-compliant real estate ad language, useful for training non-specialist marketing staff]

When it is already live and wrong

The sequence changes when the ad is in market. Work in this order.

1. Screenshot before you touch anything. Export the ad, the delivery report and the audience definition. Once an ad is deleted, your own export may be the only copy of it you control.
2. Stop delivery, at the right level. Pausing a single ad set inside a campaign that uses a shared campaign budget can push that budget into the remaining ad sets, so you can accidentally spend more on the surviving creative. Pause the campaign, then unpick it.
3. Rebuild rather than quietly edit. Editing a live ad rewrites the thing you may later need to produce as evidence. It also has a delivery cost: material changes to a Meta ad set push it back into the learning phase, which needs roughly 50 optimisation events in a week to exit, so the corrected version performs worse than the offending one did. Duplicate, correct, log, relaunch.
4. Decide whether anyone needs telling. A wrong claim that could have changed someone's decision (a payment figure, a yield, a floor area) is different from a missing registration number. The first may need a correction to the people who saw it. That decision belongs to the final approver, within a stated window, not to whoever noticed.

Then the parts that are easy to forget:

• Email cannot be recalled. The remedy is a correction to the same list, logged, and a note of the send volume.
• Print and outdoor have lead times, so the cost of the error is the media buy, not the artwork.
• Killing a portal feed at source does not clear cached copies or third-party republication. Someone checks each portal by hand and records the date it came down.
• Agents who reposted the approved creative to personal accounts are still your exposure. The kill list is a list of people as well as placements.

Failure modes worth naming

Version drift. The approved file and the uploaded file diverge, usually because a last-minute size crop or a copy tweak happened after sign-off. This is why the log stores a preview link or a hash rather than a description.

Expired truth. The creative was accurate on the day it was approved. Then the price moved, the unit went under offer, or the incentive ended. Nothing in the ad is wrong except that it is out of date, which is still a misleading ad. Time-limited claims need an expiry date and a scheduled recheck, not goodwill.

A compliant ad pointing at a non-compliant destination. The disclosures live on the landing pagelanding pageA standalone web page built for a single campaign goal, designed to maximise conversions by removing distractions and focusing visitors on one action.View full definition →, the landing page gets rebuilt mid-campaign, and the artefacts vanish. Or worse, it 404s while the ad keeps spending. Sign-off covers the click-through.

Single-point compliance. One person holds all the rules in their head, and they are on leave when the developer wants a launch. Two trained reviewers per market is the minimum, and the second one should have signed at least one live campaign in the last quarter.

Key takeaways

• Run the gates in cost-of-rework order: promise and audience at brief stage, then evidence, then disclosure artefacts, then platform configuration, then a check that the shipped asset is the approved asset.
• Four signatures, each tied to a specific statement, and one named approver with the authority to hold a launch. Track the hold rate; a zero hold rate over a quarter means the gate is not real.
• Log the version identifier, the platform preview link, the settings screenshot, the evidence references and an expiry date for time-limited claims. The record is what turns enforcement into correction.
• Platform approval from Meta or Google is an inventory decision by a seller of ad space, not a compliance clearance.
• When a live campaign is wrong: screenshot, pause at campaign level, rebuild rather than edit (editing resets Meta's learning phase, which needs about 50 optimisation events a week to clear), then decide on notification. Email cannot be recalled and portal copies need checking by hand.