What counts as an incident and what is just a poor placement
Not every unwelcome placement is a brand safetyControls that keep ads away from content that would harm the advertiser's reputation. incident. Programmatic buying runs across more domains than anyone can review, and the brand safety guide is clear that an occasional difficult adjacency on a reputable site is a risk a brand can choose to accept. The test is the brand's own written tolerance: the brand suitability policy says what is excluded, what is allowed only on named publishers and what is allowed under the default settings.
| What happened | How to treat it | Who leads |
|---|---|---|
| The ad ran beside content below the safety floor | Incident: stop at once, trace, report | Trader, with the policy owner told the same hour |
| The ad ran beside content the policy excludes | Incident: the settings did not match the policy | Trader, with the policy owner |
| The ad ran beside content the policy allows, and someone objected | Not a settings failure: reply, record, and ask the owner whether the policy should change | Policy owner |
| The ad ran on a made-for-advertising or unfamiliar domain | Poor placement: exclusion and list review in the normal routine | Trader |
| The ad itself was wrong: price, claim, landing page | Not this procedure: the live ad and page error response | Campaign owner |
| The placement is part of wider public criticism of the brand | Incident, with communications led under the business's crisis arrangements | Policy owner and communications lead |
The rest of this tutorial assumes the first two rows, though most steps apply to the third. The difference matters because an incident that turns out to be inside the policy should change the policy, not the settings. The academy exercise on handling a brand safety incident for a car seat maker practises deciding whether a screenshot shows an incident or a poor placement before choosing the narrowest stop.
Before you start
- The screenshot or complaint, with the sender's account of when and where they saw the ad, the device and, if they have it, the page address.
- Access to Campaign Manager 360 and Display & Video 360, and to the verification vendor's reports if one is used.
- The written suitability policy and the exception log, so the placement can be judged against what the brand decided rather than against the mood of the day.
- The insertion order, deal terms and agency agreement, open at the clauses on delivery, invalid traffic and credits.
- Agreed wording: the short replies for customers, colleagues and press that the communications lead approved in advance.
- The change log where settings changes are recorded.
The steps
Capture the evidence before anything changes
Save the screenshot with its original file and the message it came in, and ask the sender for the page address, the time and the device if they did not give them. Visit the page yourself and capture it, noting that the ad you see may differ because programmatic ads are chosen per impression. Record who reported it, when, and who has been told so far. Do this before touching any setting, because the changes made in the next hour will make the situation harder to reconstruct, and the record is what the team, the agency and any vendor will argue from later.
Find the impression in the ad server and verification records
The ad server records every impression it served, as the ad serving basics guide explains, and a verification vendor's tag records the page, the context and whether the traffic looked human, independently of the demand-side platform (DSPSoftware advertisers use to buy programmatic inventory.) and the exchange. Look for the domain or app and the date in both, and in Campaign Manager 360's own content verification where it is enabled. In Display & Video 360, the DV360 reporting guide points to the verification report type and a site and app report at URL level as the reports to run when someone asks where ads appeared. Confirm which creative ran and on which campaign. A screenshot that matches no record may be a different advertiser's ad, an old capture or an edited image, and is worth confirming before acting.
Trace the route through line item, deal or open auction, exchange and seller
Find the insertion order and line item that bought the impression, then whether it came through a deal or the open auction, which exchange sent it and which seller offered it. Reports split delivery by exchange and inventory source, and the programmatic ecosystem guide explains how ads.txt,
sellers.jsonand the SupplyChain object show who stood behind each seller. Note which settings the line item had on the day: content labels, sensitive categories, channels, authorised-seller targeting, vendor segments. The route decides who else needs to hear about it: a deal points to the publisher, a reseller path to the exchange and the supply-side platform (SSPSoftware publishers use to sell inventory.), and an open-auction line item with loose settings points back to the buyer.Stop delivery at the narrowest level that works
Stop the specific risk first. The DV360 inventory and deals guide lists the controls that act before the bid: an exclusion channel for the domain or app, a sensitive category or content label setting, a deal paused or removed from the line item. Choose the narrowest one that removes the risk the route revealed, apply it where it will catch the same route on every affected line item, and record the change. Avoid blanket keyword lists and whole-category blocks unless the route shows the problem really is that broad; the brand safety guide explains how they remove quality journalism along with the problem. Post-bid blocking in a verification tag can stop the creative rendering on failing pages while the cause is found, but it is a reputational control rather than a financial one.
Check whether the impression was billed and what the contract allows
The brand safety guide notes that a blocked impression may still have been paid for, depending on the exchange and the contract, and an unblocked one almost certainly was. Find the cost in the platform's reports and in the ad server's count. If the verification records show invalid trafficAd impressions and clicks not caused by genuine human interest: bots, fraud and accidents., which the Media Rating Council's standards define and the media auditing and verification guide places in invoice reconciliation, raise it under the insertion order's terms. If the placement broke a deal's terms or the agency agreement, raise that too. If it was inside the settings the buyer chose, it was a delivery the buyer asked for, and the remedy is the fix rather than a refund.
Tell the people who need to know in words agreed in advance
Tell the policy owner first, then the agency or trader if they did not find it, then whoever handles customer contact. Reply to the person who reported it, using a holding statementA brief first public response in a crisis that acknowledges the issue, tells affected people what to do now and says when more will follow. the communications lead agreed beforehand, in the sense the holding statement entry gives it: thank them, say the ad has been removed from that page and that the cause is being found, and promise nothing about the platform or the publisher. The guide to responding to negative or inaccurate coverage sets out how to handle a journalist or a public post about it, and when a situation belongs to a crisis plan rather than to a reply. Keep the internal note factual: what ran, where, how it was bought and what has been stopped.
Find the cause in the settings and their history
Ask why until the answer is something the team can change, using the five whys. Why did the ad appear there? Because the line item allowed the domain. Why? Because the suitability settings excluded nothing relevant. Why? Because the policy said nothing about this kind of page, or the policy did but the setting was loosened for a deal and never restored, or a new line item was built without the advertiser defaults. Check the change log and the exception log for when the relevant setting last changed and who changed it. The cause usually lies in one of three places: the policy, the list or the setting.
Fix the cause in the policy, the list or the setting
A setting that did not match the policy is corrected, preferably at advertiser level so new line items inherit it. A list that admitted the domain is revised, and the tutorial on building a programmatic inclusion list shows how to record the reason for each entry so the next review catches the same problem. A policy that was silent goes back to its owner, who decides the stance; the trader then translates it into settings. Then reconsider the emergency stop from the fourth step: if the permanent fix covers the risk, a broader temporary block can be lifted, so the incident does not leave the buy permanently narrower than the policy intends.
Write the incident record and review it with the team
Write a short record: what was reported, by whom and when, what the records showed, the route, what was stopped, the cost and any claim made, who was told and with which words, the cause and the fix. Store it beside the change log, review it with the trader, the agency and the policy owner, and add anything the review learns to the programmatic campaign management playbook's weekly routine, so the same pattern is caught in a placement report before anyone takes a screenshot of it.
A river cruise operator handles a screenshot from a customer (illustrative)
Kaplurvex River Cruises answers a customer's screenshot
Kaplurvex River Cruises, an invented operator with 7 river ships, receives an email from a past customer at 9:40 on a Tuesday: a screenshot of its "Spring sailings from £899" banner beside a news report of a boating accident.
| Time | Step | What happened |
|---|---|---|
| 9:50 | Evidence | Screenshot saved; the customer replies with the page address and says she saw it on her phone the evening before |
| 10:20 | Records | The ad server shows 3 impressions on that page the previous evening; the verification tag measured them as human and on a news domain |
| 10:40 | Route | Open-auction prospecting line item; the news domain is on the inclusion channel because the policy allows hard news on trusted publishers |
| 11:00 | Stop | Sensitive-category exclusion for tragedy added to the 4 cruise line items only; the news domain stays on the list |
| 11:15 | Bill | The 3 impressions cost a few pence; inside the settings the brand chose, so no claim |
| 11:30 | Tell | Marketing director informed; the customer receives the agreed reply |
| 14:00 | Cause | The policy allowed hard news but said nothing about accidents beside travel advertising |
| 14:30 | Fix | The marketing director adds a line: accident and disaster reports excluded on all travel line items; the trader makes it an advertiser default |
| Next day | Record | Incident record written and reviewed with the agency |
The rest of the news publishers keep delivering, and the following month's site report shows the cruise ads still running on the same titles beside travel, food and culture pages.
Common mistakes
- Changing settings before capturing evidence. The record of what happened is lost in the rush to fix it.
- Blocking a whole category or keyword list. It removes good inventory along with the problem; stop the route.
- Pausing everything. It stops the good with the bad and tells you nothing about the cause.
- Assuming a refund. A placement inside the settings the buyer chose was a delivery the buyer asked for.
- Replying in improvised words. Promises about platforms and publishers are hard to keep; use agreed wording.
- Fixing the symptom only. Without finding the cause in the policy, the list or the setting, it recurs elsewhere.
- Leaving the emergency block in place. Temporary stops that outlive the fix narrow the buy for no reason.
- Handling an error in the ad as a placement problem. A wrong price or broken page has its own procedure.
