What this playbook covers and when to run it
The playbook covers every incident in which something wrong has already reached customers, spend, personal data or the numbers the business decides with. It runs from preparation, before anything has gone wrong, through the report, containment and correction, telling the people who need to know, repairing reports and bidding, the review and the learning that joins one incident to the next. It does not teach any of those steps. The SOP library holds the procedures for the commonest incidents, the tutorial on reviewing a marketing error holds the routing table that sends each kind of error to the page owning its immediate response, and each phase below links the page that owns the method.
Run it when errors are fixed by whoever happens to notice and nobody else hears; when customer service learns about a wrong offer from customers; when the same kind of error has reached customers more than once; when nobody is sure who can pause an ad account out of hours; or when an error log exists that nobody reads. Run it also before a busy season, because the incidents that matter most arrive when the team has least time to improvise.
Like the other playbooks in the playbooks library, it is a starting shape. A small team may hold several roles in one person; a large team with agencies spreads them across operations, service, analytics and communications. The phases stay the same; the number of people and the frequency change.
Roles and responsibilities
- Incident lead. The owner of the channel, campaign or asset where the incident happened. Confirms it, runs the owning procedure, keeps one entry in the error log for the whole incident even when it spans channels, and closes it.
- Marketing operations owner. Owns the programme: the contact sheet, the error log, the written review triggers, the monthly read and the follow-up of every action. Facilitates reviews of incidents they were not involved in.
- Approver and commercial owner. The approver who signed off the original work approves the correction; the commercial owner decides what customers who acted on the error receive, as the live ad and page error response SOP sets out.
- Customer service lead. Hears about the incident, the facts and what customers will be offered before customers start asking, and reports back what customers say.
- Data protection owner. Whoever owns data protection in the business. Decides any duty that follows when an incident touches Personal DataAny information relating to an identified or identifiable living person, such as a name, email address, location data or an online identifier. or people who had opted out; the incident lead never decides it alone.
- Communications or social lead. Owns holding statementsA brief first public response in a crisis that acknowledges the issue, tells affected people what to do now and says when more will follow. and the public response, and decides with the named crisis owner when an incident moves to the crisis plan.
- Analytics owner. Confirms tracking incidents, sizes the effect on reports and bidding, and owns any correction to numbers that have already gone out.
A RACI matrix is a useful way to write these roles down against each kind of incident, because the commonest failure is an incident where several people act and nobody decides. Name a deputy for every role that may be needed out of hours, and make the escalation routeA short written route naming who front-line staff contact, how and how fast, when a problem is beyond what they can decide alone. part of the contact sheet rather than something remembered. The marketing incident contact sheet template lays the sheet out with one column per kind of incident: the lead and a deputy who can act, who holds access to each account, who approves, who decides and who is told, in order.
Cadence, inputs and outputs
The rhythm below is a starting point. Adjust it to how often incidents happen in the team's channels and what they put at risk, not to the calendar you would like to have.
| Cadence | Owner | Inputs | Outputs |
|---|---|---|---|
| Per incident | Incident lead, with the roles the procedure names | The report and the evidence, the contact sheet, the owning procedure | The error contained and corrected, customer service and the data protection owner told where needed, one log entry |
| After an incident that meets the triggers | Marketing operations owner as facilitator, incident lead | The log entry, the records and timeline, the written triggers | A blameless write-up with causes, owned and dated actions, and the risk it belonged to updated |
| Monthly | Marketing operations owner, team leads | The error log, open actions, the risk register | Actions confirmed done or chased, patterns named, fixes sent to the operations backlog |
| Quarterly | Marketing operations owner, communications lead, channel owners | The contact sheet, access lists, holding statements, the procedures | Owners, deputies and access confirmed, statements and procedures updated, one scenario rehearsed |
| Before a peak or risky period | Marketing operations owner, channel owners | The trading or campaign calendar, the change queue | A change freeze with its exceptions route, and the contact sheet checked for the period |
Each cadence feeds the next: log entries give the monthly read its material, the monthly read decides which patterns deserve a review and which fixes go to the backlog, and the quarterly refresh puts the lessons into the contact sheet, statements and procedures the next incident will use. Tracking incidents also belong to the measurement governance playbook, which runs outage response as one of its phases.
The playbook in phases
Prepare: owners, access and prepared responses
Most of what decides how an incident goes happens before it starts. The tutorial on building a marketing risk register gives each riskThe single table of what could knock a marketing plan off course, each risk with a warning sign, an owner and a response prepared in advance. an owner, a deputy and an early warning sign, and prepares a response in advance: a procedure, a contact list, a paused alternative, a statement. Use it to decide which kinds of incident the team must be ready for and which procedure each hands to. The guide to reducing key-person risk finds accounts and rights held in one name, so someone can always pause, revert or recover. For incidents that may become public, the social media crisis plan tutorial sets what counts as a crisis, severity levels built from the brand's own history, who decides and who speaks, and holding statements with a fast approval route. Keep the contact sheet, procedures and statements where every incident lead can reach them without logging in to anything.
Notice, confirm and route each report
Reports arrive from anywhere: a customer, a colleague, a partner, a platform notice, a check that failed. Treat every report as real until it is confirmed or ruled out, and confirm it as customers see it before anything changes. Then route it. The live ad and page error response SOP confirms and captures an error in a live ad, page or post and sends anything another procedure owns to that procedure. The social media inbox triage SOP sorts what arrives on the brand's accounts on every shift and escalates anything the crisis plan names. The tracking outage response SOP starts with who confirms that conversions or events have really stopped. Whatever the route, the incident gets one named lead and one entry in the error log from the first hour, even when it later turns out to span several channels.
Contain and correct through the owning procedure
Containment and correction belong to the procedure for the kind of incident; this playbook adds one rule: stop the wrong version at the narrowest level that works, then correct it through the approver. An account in the wrong hands follows the compromised marketing account response SOP. An email already sent follows the tutorial on fixing a mistake in a sent email, which decides whether a correction is needed before sending one. AI-assisted output follows the five moves in handling an AI mistake: contain, correct, tell, trace and log. An ad beside content the brand would never choose follows the tutorial on responding to a programmatic brand safety incident, and an automated journey sending the wrong thing follows the incident routine in the automation governance guide. When an image, claim or offer must come down everywhere it runs, the tutorial on withdrawing an asset from every channel finds every use and confirms each removal. Where customers acted on the error, the commercial owner decides what they receive before the correction goes out, so service and the correction say the same thing.
Tell customers, customer service and the business
Customer service hears first. The tutorial on briefing customer service before a campaign gives the shape of a one-page brief, with the facts, what customers will be offered, the questions to expect and the escalation route; use the same shape for an incident brief, written in the first hour. Where a known group of customers is affected, the guide to proactive service messages covers finding exactly who is affected and leading with what the problem means for them. Tell the data protection owner straight away when an incident touches personal data or reached people who had opted out: the error review tutorial summarises the GDPR's breach duties and leaves any decision with that owner, and notes that a message to people who had opted out usually means a Suppression ListA list of addresses or contacts excluded from every send, such as unsubscribes, hard bounces and spam complaints. did not reach every sending tool, against the UK regulator's direct marketing guidance that opt-outs are honoured. When the routine cannot cope, or the trouble is in the business rather than in marketing, hand over to the crisis plan or to the tutorial on leading marketing through a business crisis.
Repair the record: reports, data and bidding
An incident is not over while the numbers it touched are still wrong. When a figure has already gone out, the tutorial on correcting a number after a report has gone out confirms it against the source of truth, sizes it for every period it touches, tells the people who acted on it first and records the correction, rather than fixing it quietly in the next report. When tracking stopped or a sale distorted the data, the guide to bidding through sales and tracking outages explains the seasonality adjustments and data exclusions Google documents, so automated bidding does not learn the wrong lesson from the incident. Date every gap and restatement in the measurement change log, so anyone reading the trend later knows why it moved.
Review what let it through
Once the incident is contained, the marketing operations owner checks it against the written triggers. An incident that meets any of them is reviewed as the tutorial on reviewing a marketing error sets out, from a timeline built from records to causes traced past the person with the five whys and actions with one owner and a date each, with a facilitator who was not involved. One that meets none is logged in a line and closed.
Learn across incidents and close the gaps
Once a month, as a starting rhythm, read the error log beside the risk register. Confirm each action is done on its date, because an action that never happened is the commonest reason an error returns. Look for what no single review can see: the same gate missing different errors, the same kind of error across channels, errors clustering around one handover. Reopen the risk each incident belonged to and ask whether the warning sign appeared and the prepared response worked. When the read shows the same kind of error reaching customers again, the diagnosis for marketing mistakes that keep happening sorts the repeats by the check that missed them and works back to the cause. Send fixes that change the process to the one backlog that the guide to prioritising operations work sorts by effect, risk and effort, so they compete for time rather than waiting for a quiet week. Change procedures through the tutorial on writing a marketing SOP, so the fix lands in the document people follow. Before a peak or any period when an incident would cost most, plan a Change FreezeA set period around a peak when changes that are hard to reverse are stopped while routine, reversible work continues through an agreed route. with the tutorial on planning a change freeze, which stops risky changes while routine work carries on; for an online store's sale, the peak season trading playbook already places the freeze beside the service brief.
Decisions during an incident and who takes them
The procedures say what to do; the commonest delay in a live incident is not knowing who decides. Agree these owners in advance, write them on the contact sheet and change them only at the quarterly refresh.
| Decision | Who takes it | Who is consulted | Where the method lives |
|---|---|---|---|
| Is this an incident, and which procedure owns it? | Incident lead | Whoever reported it | The routing table in the error review tutorial |
| What customers who acted on the error receive | Approver and commercial owner | Customer service lead | Live ad and page error response SOP |
| Whether a correction or notice goes to customers | Incident lead | Customer service and communications leads | The owning procedure, or proactive service messages |
| What follows when personal data or opted-out people are involved | Data protection owner | Incident lead | The data protection section of the error review tutorial |
| Whether the incident moves to the crisis plan | Communications or social lead, with the named crisis owner | Incident lead | Social media crisis plan |
| Whether reported figures are restated | Analytics owner | The report's owner | Correcting a number after a report has gone out |
| Whether the incident gets a review | Marketing operations owner, against the written triggers | Incident lead | The error review tutorial |
An incident is closed only when the correction is confirmed everywhere the error ran and the log entry says so. A decision taken by someone not named in the table is recorded in the log with who took it and why, because that is the gap the quarterly refresh needs to see.
A flooring retailer's first quarter of incident management
Zimbrelto Flooring, an invented online flooring retailer with a marketing team of four and an agency running paid search, starts the playbook after three errors in six weeks: a shopping ad that kept last month's sale price for 2 days, a sample-request email sent to all 18,000 subscribers instead of a segment of 2,400, and a checkout release that stopped the purchase tag for 9 hours. Each was fixed by whoever noticed, and customer service heard about the price from customers. The marketing operations owner names an incident lead and a deputy for each channel, adds a second administrator to 2 ad accounts held only by the agency, writes 4 review triggers and puts the contact sheet, procedures and 2 holding statements in one folder. In the next quarter 7 incidents are logged and service is briefed within the first hour in all but the first. Three meet the triggers and are reviewed; two came through the same gap, an edit made after sign-off, so late edits now go back to the approver. At the monthly read 6 of 8 actions are confirmed done, and the quarterly rehearsal of an account takeover finds one deputy who could not log in.
Measuring the playbook
The playbook works when incidents are contained quickly, the right people hear in time and the same gap does not let the next one through. Operating signals show whether the rhythm holds: one lead and one log entry for every incident from the first hour, customer service briefed before customers ask, every incident that meets the triggers reviewed, the monthly read held and every action confirmed done on its date. Track how long incidents stay live, from first visible to contained, against the team's own history rather than any published figure.
Outcome signals show whether the team is learning: repeats of an error with a known cause, the share of reviews whose actions were all confirmed done, and incidents first reported by customers rather than by the team's own checks. Do not make the count of incidents a target; a team that reports more may simply be looking harder, as the error review tutorial says of its log. Report these signals in the existing monthly and quarterly packs, beside the results the incidents touched; the reporting cadence playbook sets who sees them and when.
Common mistakes
- No lead, several fixers. Three people fixing in parallel produce three partial corrections and no log entry. Name one incident lead from the first hour.
- Service hears last. Customers who ask before the service team knows get different answers from different agents. Brief service before the correction goes out.
- Contacts who cannot act. Deputies without access, or access held by a former agency, fail out of hours. Test the contact sheet at the quarterly refresh.
- Fixing the report quietly. A number corrected silently costs more trust than the error. Restate it and tell the people who acted on it.
- A log nobody reads. Entries never read together never show a pattern. Read the log monthly beside the risk register.
- Treating every incident as a crisis. Escalating routine errors wears out the crisis route. Use the crisis plan's own levels and keep everything below them in the normal procedures.
