Why review an error that is already fixed
Most marketing errors are fixed quickly and forgotten just as quickly. The ad with the expired discount codeA code a customer enters at checkout to receive an offer, also used to identify sales from a creator, partner or channel that links cannot capture. is paused, the landing page with the wrong price is corrected, the email that went to the wrong segment is followed by a careful correction. The team moves on, relieved, and nothing about the way the work is done has changed. Some months later a different person, on a different campaign, makes a similar mistake through the same gap.
A review closes that gap. It does not repeat the fix, and it is not a search for whoever made the slip. Its question is narrower and more useful: which part of the process let this error reach customers, and what change would stop it next time? The answer is usually a missing field, a check that was skipped under pressure, a step nobody owned or a handover where information was lost. Those are things a team can change; individual care, on its own, is not.
The response to the error itself is owned by the channel it happened in, and each has its own page. Use them first; this review begins when they end.
| Where the error happened | Who handles the immediate response | Where it is covered |
|---|---|---|
| A marketing email already sent | The email owner | Fixing a mistake in a sent email |
| Output produced with AI tools | The editor and the tool's owner | Handling an AI mistake |
| Conversions or events stopped recording | The analytics owner | Tracking outage response SOP |
| Public criticism on social channels | The social and communications leads | Social media crisis plan |
| An automated journey sending the wrong thing | The workflow owner | Automation governance |
| An ad, a page or any other live asset | The campaign owner | The launch check in the campaign workflow guide |
For an ad, a page or a post, the live ad and page error response SOP runs the immediate response, from containing the wrong version to the log entry this review starts from.
Reviewing a whole campaign against its brief is a different exercise again. The campaign post-mortem template closes a campaign; the review here looks at one error and the route it took.
Before you start
- Confirm the error is contained. Whatever was wrong has been stopped, corrected or withdrawn, through the channel's own procedure. A review held while the error is still live competes with the fix for attention.
- Name a facilitator who was not involved. Someone from operations, or a lead from another team, keeps the conversation on facts and stops the chain of causes ending at a person.
- Gather access to the records. The work item and its history, the brief, the approval record, the platform's change history, the launch check result, messages about the work and any customer contacts it caused.
- Read the five whys method. The five whys framework sets the two rules this review depends on: every answer rests on evidence, and the chain stops at a process, rule or check rather than a person.
- Know who owns data protection. If the error touched personal data or contacted people who had opted out, that person needs to hear about it before the review, not from its write-up.
- Open the error log. If the team has none yet, the final step creates it; the section after the steps lists what it records.
The steps
Decide whether the error needs a review
Not every slip deserves a meeting. Write down the triggers once and apply them every time: customers saw the error, money was spent or lost because of it, personal data was involved, it reached a regulated claim or price, a check that should have caught it did not, or the same kind of error has happened before. An error that meets none of these is logged in a line and closed. One that meets any of them gets a review, however small it looks, because the size of this error says little about the size of the next one through the same gap.
Collect the facts while they are fresh
Before anyone meets, collect what the records say. Screenshots of what customers saw, the version history of the asset, the platform's change log, the approval record on the work item, the launch check and who ran it, the messages exchanged about the work and any complaints or replies. Ask the people involved to write down, briefly and separately, what they did and what they expected to happen. Written accounts collected early are more accurate than recollections gathered in a room weeks later, and they spare people from defending themselves in public.
Build a timeline from records, not memory
Lay out what happened in order: when the request arrived, when the brief was signed, when the asset was built, reviewed, approved and checked, when it went live, when the error was first visible to customers, when it was noticed, by whom and how, and when it was contained. Mark each entry with its source. Gaps in the timeline are findings in their own right: an approval that has no record, or a change made after sign-off that appears in the platform's history but nowhere else, often points straight at the cause.
List the checks that should have caught it
Against the timeline, list every point at which the error could have been stopped: the intake fields, the brief, the build procedure, each review the approval matrix required, the launch check, any monitoring after launch. For each, record whether the check exists, whether it was run and whether it could have seen this error. The approvals and compliance guide shows which reviews the matrix should have triggered; the campaign launch QA checklist shows what a launch check normally covers. Three answers are common, and each leads to a different fix: there was no check, there was a check that was skipped, or there was a check that could not have seen this kind of error.
Ask why until the cause is a process, not a person
Take the most important missed check and ask why it missed, then why of that answer, using the five whys. Support each answer with the records collected earlier, and mark anything unsupported as a hypothesis to check before going on. When an answer names a person, such as "the specialist forgot", ask why the process depended on remembering. Stop when the answer is something the team can change and changing it would stop a repeat: a field that does not exist, a check that is optional under deadline pressure, a step with no owner, a value copied by hand from one system to another. Where several checks failed, run a chain for each; the five whys follows one path, and an error with several causes needs several chains.
Separate the causes from what made it worse
Some findings explain why the error happened; others explain why it lasted or spread. A missing end date on an ad is a cause; nobody watching the ad after launch is an aggravating factor; a busy week is context, not a cause. Sorting them keeps the actions honest. Causes need changes that stop the error being made or reaching customers. Aggravating factors need changes that catch it sooner. Context explains the day but changes nothing, and an action list built on context ("be more careful during busy periods") is a sign the review stopped too early.
Agree actions, each with an owner and a date
Each cause gets at least one action that changes the system: a new required field, a check added to the build procedure, a review row added to the matrix, a step moved earlier, an owner named for a step that had none. Procedures change through the tutorial on writing a marketing SOP, so the fix lands in the document people actually follow. Give every action one owner, a date and a way to show it is done. Where the error touched measurement, date it in the measurement change log so later readers of the numbers know why they moved. Where the review finds a risk that could recur in another form, add it to the marketing risk register.
Write it up briefly and share it without blame
The write-up fits on a page: what happened and what customers saw, the timeline, the checks and why each missed, the causes and aggravating factors, and the actions with owners and dates. Describe roles and steps, not individuals. Share it with the people involved first, then with the wider team, because the same gap probably exists in work other people do. A write-up nobody reads beyond the review group changes one person's habits; one the team reads changes the team's.
Log the error and check the actions were done
Add the error to the team's error log, then return to it on the actions' dates and confirm each one is in place. An action that has not happened is the commonest reason the same error returns. Over time, the log shows patterns no single review can: the same gate missing different errors, the same kind of error across channels, errors clustering around a particular handover. Those patterns are the input to the next change in the process, and they are worth reviewing on their own when the team plans its operations work. The marketing incident management playbook runs that read every month beside the risk register, with the incident leads, contact sheet and decisions that surround each review. When a review finds that the right person could not be reached or could not act, the fix usually lands in the marketing incident contact sheet template, which names a lead, a deputy and the order of telling for each kind of incident.
Errors that touch personal data need one more line of care. If the error exposed customer details or sent marketing to people who had opted out, the review records it and hands the decision about any duty to whoever owns data protection; it does not decide it. The GDPR entry summarises the regulation's breach duties: assess the breach, notify the supervisory authority unless the risk to people is unlikely, tell those affected where the risk to them is high, and document it. A message that reached people who had opted out usually means the suppression list did not reach every sending tool; the UK regulator's direct marketing guidance expects opt-outs to be honoured, and the data operations guide makes checking that suppression matches across tools an operations routine. Treat a gap there as a cause, not as bad luck. Keep Personal DataAny information relating to an identified or identifiable living person, such as a name, email address, location data or an online identifier. out of the review documents themselves; refer to records by reference.
What the error log records
The log is one list for every channel, kept where the team's other operational records live. It records enough to spot a pattern without retelling each story.
| Field | What it holds | Why it matters |
|---|---|---|
| Date found and date contained | When the error was noticed and when it stopped | Shows how long errors stay live, and whether that is improving |
| Channel and work type | Where the error appeared and what kind of work it was | Reveals errors clustering in one channel or work type |
| What customers saw | A short description, with a reference to the evidence | Keeps the record factual and checkable |
| Checks that missed it | The gates that existed and why each missed | Shows which gates repeatedly fail |
| Causes | The process causes found, not people | The input to process changes |
| Actions, owners and dates | Each change agreed and who owns it | Makes follow-through visible |
| Actions confirmed | Whether and when each action was checked as done | The commonest gap in any review process |
| Repeat of | A reference to any earlier error with the same cause | Turns separate incidents into a pattern |
Resist turning the log into a scoreboard. A team that reports more errors may simply be looking harder, and a falling count can mean errors are going unreported. If the team wants a KPIA metric tied directly to an objective. for this work, the share of reviews whose actions were confirmed done, and the number of repeats with a known cause, say more about whether the process is learning than the raw count does. When the repeat field keeps pointing at the same kind of error, the diagnosis for marketing mistakes that keep happening reads those entries together and traces them to the gate, procedure, matrix row, handover or peak that keeps letting them through.
A kitchen retailer reviews an expired-code ad (illustrative)
Feldrosk Kitchens reviews an expired-code ad
Feldrosk Kitchens, an invented kitchen retailer selling fitted kitchens online and through 6 showrooms, ran a paid social campaign offering 15% off with a discount code until the end of a spring sale. The ad kept running for 4 days after the code expired. Customers who tried the code at checkout were told it was invalid, and the showrooms received 23 complaints before a sales adviser told marketing.
The paid social specialist paused the ad within the hour of hearing, and the campaign owner agreed with the showroom managers to honour the discount for the customers who had complained. Because customers saw the error and it involved a price, it met the team's review triggers, and the operations lead facilitated.
The timeline showed that the sale's end date was in the signed brief, but the ad build sheet had no end-date field, so the ad was set to run until its budget was spent. The launch check confirmed the code worked on launch day and had no item about end dates. Nobody monitored the ad after launch, because the handover named no monitor for always-on social campaigns. The chain of whys for the build ended at "the build sheet has no field for the end of an offer"; the chain for the launch check ended at "the checklist checks offers when they start, not when they end".
The team separated the causes (no end-date field; no end-of-offer check) from the aggravating factor (no named monitor). It agreed 3 actions: an end-date field added to the build sheet and the paid social procedure, owned by the paid social lead; an end-of-offer item added to the launch check, owned by operations; and a named monitor required at handover for every campaign carrying an offer, owned by the head of marketing. The one-page write-up was shared with the whole team and the showroom managers. When the operations lead checked the log 5 weeks later, all 3 actions were in place, and the log showed one earlier error with the same cause: an email sent the previous autumn promoting a delivery offer that had already ended.
Verify it worked
- Check the timeline. Every entry should name its source, and any gap should be written down as a finding.
- Check the list of checks. Each should be marked as missing, skipped or unable to see this kind of error.
- Read the end of each chain of whys. It should stop at a field, a check or a step the team can change, never at a person.
- Compare causes with actions. Each cause needs an action that changes the system, with one owner, a date and a way to show it is done.
- Open the log and the write-up. The entry should be complete, and the write-up should describe roles and steps, not individuals.
If it goes wrong
- A chain of whys ends at a name, such as "the specialist forgot". The process depended on memory. Ask why again until the answer is something the team can change.
- The action list says "be more careful during busy periods". Context has been treated as a cause. Return to the causes and write actions that change a field, a check or an owner.
- The timeline has gaps or conflicting accounts. Go back to the records, such as the platform's change history. Mark unsupported points as hypotheses.
- The discussion turns defensive. Let the facilitator steer it back to the process, and keep names out of the causes.
- An action's date has passed and nothing has changed. Ask the owner what is blocking it, and record the answer in the log.
Next steps
- Read the tutorial on writing a marketing SOP to draft the revised procedure when an action adds a field, a check or an owner.
- Read the campaign workflow guide to see where intake, build and launch check sit, so a new check lands at the right stage.
- Read fixing a mistake in a sent email to compare the email owner's immediate response with your own.
Common mistakes
- Reviewing while the error is still live. The fix comes first, through the channel's own procedure; the review starts once customers are no longer seeing the error.
- Stopping at a person. "The specialist forgot" explains nothing the team can change. Ask why the process depended on remembering.
- Building the timeline from memory. Recollections drift and protect their owners. Use the records, and treat gaps in them as findings.
- Treating context as a cause. A busy week explains the day but changes nothing; actions built on it, such as being more careful, will not hold.
- Actions without owners or dates. They are intentions, and the same gap stays open. Every action needs one owner, a date and a way to show it is done.
- Changing people's habits instead of the procedure. A fix that lives only in the heads of the people in the room disappears with them; update the procedure, the template or the check.
- Deciding data protection questions in the review. Record the facts and pass the decision to whoever owns data protection, straight away.
- Never checking the actions were done. The review feels finished when the write-up is shared; it is finished when the changes are confirmed in place. The process hub explains how changes like these show up in cycle time and rework.
