What a risk register is for
A plan describes what the business intends to do. A risk register describes what could stop it, and what will happen if it does. It is not a list of worries. Each entry earns its place by being specific enough to watch for and to prepare for.
The register sits between two neighbouring tools. The quarterly planning session names risks and dependencies as one step in agreeing a quarter's work, which captures the risks to particular initiatives. The scenario planning framework prepares for a few large uncertainties outside the business's control, such as a fall in market demand, by building plausible futures and agreeing responses to each. The register is the standing document that collects the risks to the whole marketing plan, inside and outside the business, and keeps them current between those sessions.
Its value comes from three things the plan does not provide:
- Warning signs agreed in advance. Most incidents announce themselves before they become serious, to someone who knows what to look for.
- One owner per risk. When something goes wrong, nobody has to work out whose problem it is.
- Responses prepared before they are needed. The decisions that are hard to make under pressure, such as what to pause, whom to tell and what to switch on instead, are made calmly in advance.
Some platforms' own rules can be read as a register of their own. The guide to Google's spam policies ends with a risk review that checks a team's content and link practices against them; the marketing register can point to reviews like that rather than repeat them. The register belongs with the brief, the calendar and the budget in the planning section: those documents say what will happen, and the register says what the team will do if it does not. The academy checkpoint on risk registers, crises and reforecasts tests what belongs on a register, from warning signs to prepared responses, and what follows when a risk arrives as a crisis.
Before you start
- The current plan. Objectives, channel mix, budget, calendar and the main campaigns. The register is built from what this plan depends on, not from a generic list.
- Recent incidents. Anything that disrupted marketing in the past year or two: tracking breaks, disapprovals, stock-outs, agency changes, complaints. Past incidents are the best evidence of what is likely.
- The people who see risks first. Someone from analytics, from operations or fulfilment, from finance and, where agencies run significant work, from each agency. They see warning signs marketing does not.
- The existing procedures. Incident procedures, approval matrices and handover checklists the business already has, so the register can point to them rather than reinvent them.
- A simple scoring scale, agreed before anyone scores, so likelihood and impact mean the same thing to everyone.
The steps
List risks by category
Work through the categories below and, for each, ask what in this plan could go wrong. Write specific risks, not headings: "the purchase tag stops firing after a site release" rather than "tracking issues". Most plans produce a manageable number of risks once duplicates are removed, and categories make gaps visible: a register with ten platform risks and no people risks has probably missed something.
Category Examples of specific risks Where the site covers it Measurement A tag stops firing after a release; a consent banner change alters what is recorded; a platform changes a definition Tracking outage SOP, measurement change log Platform Ads disapproved; an account restricted or suspended; a ranking or price change on a channel the plan depends on Ad policies and disapprovals, spam policies Concentration Most new customers depend on one channel, platform, partner or affiliate Reducing dependence on a single channel Supply and capacity A best-seller runs out; the team cannot serve the demand a campaign creates Planning marketing around capacity and stock Claims and approvals A claim goes out without substantiation; a price or endorsement is not reviewed Approvals and compliance reviews Data and consent People contacted without consent records to support it; a new data use nobody checked The approvals guide's consent and data checks Reputation Ads appear beside harmful content; an AI-assisted output reaches customers with an error Brand safety and viewability, handling an AI mistake People The only specialist in a channel leaves; an agency resigns or is replaced; a supplier fails The management and agency guides For the people row, the guide to reducing key-person risk finds the tasks and accounts that depend on one person before an absence or a departure turns them into an incident. Where the supplier that fails is a marketing tool, the guide to preparing for a marketing tool outage lists the jobs that stop and the fallback to prepare for each, which can become that row's prepared response.
Describe each as cause, event and effect
Rewrite every risk in three parts: because of a cause, an event may happen, leading to an effect on the plan. "Because site releases are not checked against analytics, the purchase tag may stop firing, leading to lost conversion data and automated bidding cutting spend on campaigns that are still selling." The structure does two jobs. It separates risks that share an effect but have different causes, which need different responses. And it shows where to act: on the cause to prevent the event, or on the effect to limit the damage. A risk that cannot be written this way is usually too vague to manage.
Score likelihood and impact
Score each risk on likelihood and impact using the scale agreed beforehand, such as low, medium and high, with a written description of what each level means for this business. Impact is easiest to anchor in the plan's own terms: the effect on the objective's KPIA metric tied directly to an objective., the budget at risk, the customers affected or the time to recover. Likelihood rests on the business's own history and the judgement of the people closest to the risk; there is no published rate for how often most marketing incidents happen, so do not borrow one. Plot the scores on a grid, in the way the impact-effort matrix plots work, and give the high-likelihood, high-impact corner the most attention. Score the risk as it stands today, with the controls already in place, so the score changes when a prevention measure is added.
Name an early warning sign for each
For every risk that matters, decide what would show it is starting to happen, where that signal can be seen and what level should trigger action. A gap opening between recorded conversions and the business's own orders warns of a tracking break; a disapproval notice or a policy email warns of an account problem; stock cover falling below the lead time for reordering warns of a shortage; a specialist mentioning interviews warns of a departure. A warning sign works like a Guardrail MetricA metric an experiment must not worsen beyond an agreed bound, watched beside the primary metric that decides the test. in a test: a measure watched beside the main result, with a bound agreed in advance, whose breach prompts a decision. Where a risk has no warning sign at all, the prepared response matters even more.
Assign an owner
Give each risk one owner: the person best placed to watch its warning sign and to start the response. The owner does not have to fix everything personally, but they are accountable for noticing and acting. Owners often sit outside marketing, in analytics, operations or an agency. The RACI matrix helps where a response involves several people: one accountable owner, those responsible for the work, those consulted and those informed. Name a deputy for risks that could happen while the owner is away, because incidents rarely wait for convenient times.
Choose a response and prepare it
For each risk, choose one of four responses. Prevent: act on the cause so the event becomes less likely, such as checking analytics after every release. Reduce: limit the effect if it happens, such as keeping a second channel running so one restriction does not stop all sales. Accept: record the risk and do nothing more, which is right for risks that are unlikely and cheap to absorb. Prepare: write the procedure in advance, so that when the event happens the owner follows steps rather than improvising. Prepare means producing something: a short procedure, a contact list, a paused set of alternative campaigns, 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.. For many categories the procedure already exists elsewhere on the site. The tracking outage SOP runs a measurement incident from triage to report; the ad policies guide sets out how to respond to a disapproval by fixing the cause before resubmitting or appealing, and notes that repeated or serious breaches can lead to suspension, which is far harder to reverse; the AI mistake guide's five moves, contain, correct, tell, trace and log, apply to most public errors. For reputation risks that surface on social, the social media crisis plan tutorial prepares severity levels, owners and holding statements before they are needed. Where the risk is a crisis in the business itself, such as a long outage or a supply failure, the tutorial on leading marketing through a business crisis supplies the prepared response: a pause list by channel and the order of the restart. For data and consent, the approvals guide's check is whether the work uses personal data or contacts people in a way the organisation's consent records support, which protects the First-Party DataData a business collects directly from its own customers, visitors and users. the business has collected; for the UK, the Information Commissioner's Office's direct marketing guidance explains how the rules apply to marketing. Nothing here is legal advice.
Record everything in one table
Keep the register as a single table in one place that the whole team can read, with one row per risk. The columns that matter are: an identifier, the category, the cause-event-effect description, likelihood, impact, the warning sign and where it is seen, the owner and deputy, the response type, a link to the prepared procedure, the date last reviewed and the status (open, watching, triggered, closed). Resist splitting the register across documents or tools: a register that has to be assembled before it can be read will not be read when it matters. Link the register from the plan, and link each prepared procedure from its row.
Review monthly and at each quarterly plan
Review the register once a month in a short slot, ideally inside an existing meeting. For each high-scoring risk: has the warning sign moved, is the prepared response still current, is the owner still the right person? Then go deeper at each quarterly planning session, where the plan changes most. New campaigns, channels, partners and launches bring new risks; the quarterly session's own step on risks and dependencies feeds straight into the register. After any incident, review the risk it belonged to within days: did the warning sign appear, did the response work, what should change? Record measurement-related changes in the measurement change log as well, so later readers of the data can see what happened and when. The marketing incident management playbook holds this monthly review in the same slot as its read of the error logOne list across every channel of the errors that reached customers: when found and contained, what customers saw, the checks that missed it, causes and actions., so every incident reopens the risk it belonged to.
Close, merge and add risks as the plan changes
A register only stays useful if it stays short and current. Close risks that no longer apply, such as a campaign that has ended or a dependency that has been removed, and keep the closed rows in a separate tab for reference. Merge risks that turn out to share a cause and a response. Add risks when the plan changes: a new channel, a new agency, a new market, a new type of claim, a new use of customer data. If the register keeps growing without ever shrinking, the scoring has stopped discriminating, and it is time to rescore everything against the scale and accept more of the small risks explicitly.
A pet supplies retailer's register (illustrative)
Four rows from a pet supplies retailer's risk register
Burrowlark Pet Supplies, an invented online pet supplies retailer, builds a register of 12 risks in a two-hour session with its marketing manager, its developer, its warehouse lead and its paid media specialist. Scores use a 1 to 3 scale for likelihood and impact, with impact anchored to weekly revenue at risk: 1 is under £2,000, 2 is £2,000 to £10,000, 3 is over £10,000. Four rows from the register:
| Risk (cause, event, effect) | L | I | Warning sign | Owner | Response and preparation |
|---|---|---|---|---|---|
| Because releases are not checked against analytics, the purchase tag may stop firing, so bidding cuts spend on campaigns that still sell | 3 | 3 | Recorded purchases fall more than 15% below shop orders on any day | Developer, deputy marketing manager | Prevent: test purchase after every release. Prepare: tracking outage procedure, data exclusion steps |
| Because ads and landing pages are edited without a pre-launch check, repeated disapprovals may lead to the main ad account being restricted, so search and Shopping stop | 2 | 3 | More than 2 disapprovals in a month; any policy email | Paid media specialist | Prevent: pre-launch policy review. Reduce: email and organic plans ready to scale. Prepare: a record of fixes for any appeal |
| Because one supplier makes the best-selling dog food, a shortage may leave it out of stock, so the main campaign promotes an unavailable product | 2 | 3 | Stock cover below 3 weeks, the supplier's lead time | Warehouse lead | Prepare: feed availability rule; alternative brand campaign built and paused; customer email drafted |
| Because only one person runs paid media, their departure may leave campaigns unmanaged, so spend drifts and changes stop | 2 | 2 | Notice given; access held only by them | Marketing manager | Reduce: second admin on every account, documented settings. Prepare: a freelance contact briefed |
Four months later a checkout release stops the purchase tag at 11pm. The daily check flags the gap at 9am, the developer follows the procedure, the data exclusion covers the window and the tag is fixed by noon. The same month the dog food supplier warns of a delay; with 2 weeks of cover left, the warehouse lead triggers the response, the paused alternative-brand campaign goes live and the drafted email goes to regular buyers of the food. The register's review that month closes one risk, adds a new one for a marketplace launch and rescores the supplier risk upwards.
Verify it worked
- Read any row aloud. It gives a cause, an event and an effect, not a heading such as "tracking".
- Check the scores. Each risk has likelihood and impact from the agreed scale, scored with today's controls in place.
- Pick a risk that matters. It has a warning sign, a place where it is seen, a trigger level, one owner and a deputy.
- Open a response link. It leads to a written procedure or a built alternative, and the owner can say what they would do first.
- Confirm the register is a single table linked from the plan, with a last-reviewed date and a status on every row.
If it goes wrong
- Nearly every risk scores high. The scoring has stopped discriminating. Rescore against the written scale and accept more small risks explicitly.
- A risk has no warning sign. It will only be noticed after it happens. Name what someone would see first, and rely harder on the prepared response.
- Nobody acts when a warning sign moves. The owner is a team, or has no deputy. Name one person and a deputy.
- The response reads "we would pause the campaigns". That is an intention. Write the procedure, build the alternative and link both from the row.
- The register has gone stale. Reviews slipped, or incidents never reopened their risks. Review it inside an existing meeting and re-check the risk after any incident.
Next steps
- Link the tracking outage SOP from every measurement row, so the owner has steps to follow.
- Read reducing dependence on a single channel if the concentration rows score high and need a reduce response.
- Use the scenario planning framework for large outside uncertainties that fit better as scenarios than as rows.
Common mistakes
- Risks written as headings. "Tracking" or "competition" cannot be watched or prepared for; write cause, event and effect.
- Borrowed probabilities. Score likelihood from the business's own history and judgement, not from an imported figure.
- No warning signs. A risk that is only noticed when it has happened has no early response.
- Owners who are teams. One named person per risk, with a deputy.
- Responses that are intentions. "We would pause the campaigns" is not preparation; a written procedure and a built alternative are.
- A register separate from the plan. Review it monthly and at each quarterly plan, and link it from the plan.
- Only marketing in the room. Analytics, operations, finance and agencies see the warning signs first.
- A register that never shrinks. Close and merge risks as the plan changes, and accept small ones explicitly.
