Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

Marketing Technology › MarTech Tool Categories

Preparing for a Marketing Tool Outage

How to prepare for and handle a marketing tool outage: the jobs that stop, a fallback for each, the first hour, pausing what depends on it and catching up.
Guideintermediate15 min readUpdated

Why a supplier's outage becomes a marketing problem

Marketing that runs on rented software has handed part of the operation to somebody else's systems. The tool categories hub answers that by naming an owner before signing, with administrative access held by the organisation rather than an agency or one person's account. The day the supplier's service is unavailable tests something else: whether anyone has decided what the business does while the tool is gone.

Neighbouring pages stop short of this case. The tracking outage response SOP handles conversions or events that stop recording while sales carry on, and it leaves site downtime to the technical team's own process. The marketing incident management playbook covers any incident where something wrong has already got through to customers, spend, personal data or the figures. The error review tutorial keeps a routing table pointing each kind of error at the page responsible for its first response, and a supplier's service being unavailable is not among its rows. The marketing risk register tutorial lists a supplier failing among its people risks without a response written for it.

In a tool outage nothing is wrong with what the business sent, built or recorded; the problem is absence. Forms produce no leads, reminders are not sent, and opt-outs arrive by other routes. Each hour leaves work to catch up, and the catch-up is where avoidable harm can happen: a reminder that goes out twice, a campaign that restarts before the form works, a person who opted out by phone and is mailed the next morning.

Which jobs stop when each tool is down

Start with jobs, not tools: the fallbacks have to replace what the business stops being able to do. Go through the stack record with each owner and write down every job that runs through each tool, including the ones it does on behalf of other tools.

The register built in the tutorial on documenting marketing integrations finds the hidden jobs, since it records every connection, its direction, its owner and the check that shows it failing. Every row that starts or ends at the missing tool is a job that stops too: the form tool that no longer passes enquiries to the CRM, the CRM that no longer feeds the email platform's segments, the booking tool that no longer starts the reminder workflow. A connection missing from the integration register is a job that can stop unnoticed, because nothing on the list points to it.

Tool that is downJobs that stopJobs that stop elsewhere because of it
Email platformCampaign sends, automated journeys, transactional messages it sends, unsubscribe processingSegments other tools push into it; reports that read its results
Form toolEnquiries, sign-ups and the consent wording captured with themRecords it passes to the CRM; workflows those records would start
Booking widgetNew bookings, changes and cancellationsReminder and follow-up messages that read from it; capacity shown on the site
CRMLooking up customers, assigning leads, recording outcomesSyncs to sending tools; audiences and conversion uploads built from it
Chat toolLive answers and captured contact detailsHand-offs to the service team; enquiries logged from chats

Then mark each job by how long the business can live without it. A newsletter can wait until the tool returns. An enquiry that cannot be submitted goes elsewhere, and a booking that cannot be made leaves a place empty. And some jobs carry a duty, the clearest being opt-outs, which have to be honoured whether or not the tool that normally processes them is working. The jobs in the last two groups are the ones that need a fallback.

A fallback for each job that cannot wait

A fallback is the simplest way to keep a job going without the tool, written down and tested in advance. It has to capture what the tool would have captured, in a form that can be entered later without guesswork, and it has to be reachable without logging in to the tool that is down.

Job that cannot waitA fallback prepared in advanceWhat the fallback must capture
Taking enquiriesA second, simple enquiry route the site can switch to, such as a plain contact form handled by a different tool or a published phone lineThe person's details, what they asked and the consent wording they saw
Taking bookingsA phone or in-person booking sheet with the same fields the booking tool usesSession, time, contact details, any marketing permission and its wording
Honouring opt-outsA log for opt-outs received by any route while the tool is downWho, when, by which route, in the person's words, and the scope
Answering customersA short service note on the site and a briefed service teamWhat is unavailable, what customers can do instead, when they will hear more
Sending essential messagesA route through another tool the business already uses, agreed in advanceOnly the message that cannot wait, to only the people who need it

The consent wording matters more than it first appears. The guide to backing up marketing tool data and settings treats consent evidence, meaning source, date and wording, as one of the records a business cannot afford to lose, because without it nobody can prove what a person agreed to. A phone sheet without the wording creates that gap. The same guide's copies are the fallback for records: if the CRM is down and the service team needs to look up a customer, a recent copy held outside the tool, restricted to the people who need it, is what lets them answer.

A fallback that collects details collects personal data. Keep sheets where only the named people can reach them, enter them into the proper tool once it returns, and agree with whoever owns data protection what happens to them afterwards.

Write each fallback into the risk register as the prepared response to that tool failing, with an owner, a deputy and the warning sign that triggers it, so it does not live in one person's head. Try each one in quiet conditions, such as a practice booking taken on the sheet and entered afterwards; an untried fallback can fail on a detail nobody thought to check.

The first hour: confirming the cause and telling the right people

The tracking outage SOP runs its incidents through confirming, triaging, assessing the impact, protecting campaigns, fixing, verifying and annotating. A supplier outage follows the same shape, except that the fix belongs to the supplier, so the business's part is everything around it.

Confirm that the cause is the supplier. First rule out the business's own causes. A release that changed the page the form sits on, an edited embed, credentials belonging to someone who has left, a lapsed payment method or an access change can all look like an outage from outside. The integration register shows whose credentials each connection uses; the data flow and integrations guide asks that monitoring alerts reach the owner of each flow and name the failing check along with the last time the flow was known to work, which is the evidence this step needs. Then reproduce the failure the way a customer would meet it, with a test submission or a test booking, and read whatever notices the supplier publishes. This guide does not describe what any supplier's notices contain.

Name one lead. The incident playbook makes the owner of the affected channel, campaign or asset the lead, which for a tool outage means the tool's owner. One person leads; several people acting at once with nobody deciding is the failure the playbook warns against.

Tell people in the right order. The playbook's rule is that customer service hears first, before customers start asking. The marketing incident contact sheet template sets out, per kind of incident, who is told and in what sequence, plus an 'Agencies and suppliers' row naming, for each supplier, who to ring there, what that person can act on and what remains the business's own job. A supplier outage deserves its own column there, and the person named for each supplier should be someone still at the business. The template treats an incident type that has no page responsible for it as a gap to close in advance, and this guide can be that page.

Decide what customers are told. A short holding statement on the site and in the service team's replies says what is unavailable, what to do instead and when the next update will come. Keep blame out of it: the glossary entry warns that pointing at a supplier or a platform looks like deflection.

Start the record. Note the time the failure was first seen, the evidence that confirmed it and every decision from then on. Where customers met the failure, it gets one entry in the team's Error Log, as the playbook asks for every incident from the first hour. If the supplier's notices mention anything about personal data, pass that to whoever owns data protection at once; that decision is theirs, not the incident lead's.

Pausing what depends on the missing tool

Once the outage is confirmed, pause anything that sends someone to the broken step or depends on data the missing tool would have supplied. The candidates come from the jobs list. Ads and emails that lead to a form that cannot be submitted. Automated journeys that read from the booking tool or the CRM, which may send the wrong thing or nothing at all while the source is silent. Scheduled sends from other tools that link to the broken page. Syncs into or out of the missing tool, which the integration register lists row by row. Write each paused item down with its owner and the time, so the restart is complete rather than remembered.

Pause; do not delete. The crisis tutorial for marketing managers makes the point in its pause list by channel, where deleting campaigns, lists, workflows or accounts sits in a column of steps kept for the owner alone because each is a one-way door, and a pause does nearly all of what a deletion would while remaining reversible. Where the outage is not one tool but the business's own service, that tutorial and its pause list take over from this guide.

When the tool that fails reports conversions rather than producing them, the case is a Tracking Outage and belongs to the tracking outage SOP, with bidding protected as the Smart Bidding guide for sales and outages describes, including the data exclusions Google documents for some strategies. When the form or booking step itself is down, enquiries genuinely stop, a case that guide's table does not describe; pausing the campaigns that lead there is the paid media owner's call, and its warning against changing targets during an unusual period still holds.

Catching up without double sends or lost opt-outs

The outage ends when the service works again; the incident does not. Catching up has its own order.

Suppression first. Before any message goes out, every opt-out received during the outage is recorded. The SOP for marketing opt-out requests treats a request by any route, a phone call, a reply or a word at a desk, as an opt-out, records it a single time, in the system of record and on the suppression list, and confirms that every sending tool received it. The glossary entry on the Suppression List notes that direct marketing guidance from the UK regulator wants opt-outs acted on promptly, and an outage does not suspend that. Work through the fallback opt-out log first, and confirm each entry is in place before any send or workflow restarts.

Then enter what the fallbacks captured. Bookings, enquiries and contact details taken on sheets go into the proper tool with the consent wording that was recorded with them. Entered by hand, they are ordinary records; loaded in bulk, they can match and change existing records, and that change cannot be walked back, so take a copy of the affected records before the load, as the backup guide asks before any bulk change.

Then check what is waiting to send. Before an automated journey or a scheduled send is switched back on, look at who it will reach and with what. A reminder for a past session and a confirmation to someone already confirmed by phone are the double sends to catch. The guide to testing across the stack counts a message sent to customers as a one-way change, because a sent message cannot be recalled, and data sent to ad platforms as one-way too, since the platform keeps it and its bidding adjusts to it. The framework of one-way and two-way doors supplies the test for this moment: a two-way step can be taken and corrected, while a one-way step is checked before anyone takes it. Restarting a paused ad is a two-way door; releasing a held send to a full list is not.

Then reconnect and watch. Switch the syncs back on one at a time and confirm each one delivers fresh data within its expected window before relying on it. Restart paused campaigns from the written list, testing the step each one leads to first.

Then tell the customers who were affected. Where a known group of customers could not book, could not get an answer or received something wrong, the guide to proactive service messages sets out finding exactly who was affected, leading with what the problem meant for them and keeping the message free of promotion. If the business wants to make amends, that guide frames it as a remedy, set out plainly for the people who were affected, rather than an offer attached to the notice.

Recording the outage for the next renewal

Record the outage in three places.

The numbers come first, because anyone reading the trend later needs to know why it dipped. Add the window, with its start, its end and the reports it affects, to the measurement change log, which treats outages as one of the changes that move the numbers and records them as windows rather than single dates.

The risk register comes next. Reopen the row for that supplier failing and ask the questions the register tutorial puts after every incident: did the warning sign show itself in time, did the prepared response work, and what should change? Rewrite any fallback that failed while the details are fresh.

The renewal is the third. The tutorial on reviewing a tool before it renews lists outages that affected campaigns among the incidents to gather before the notice date, alongside support requests and the workarounds the team built. Write down what the outage cost in terms the review can use: enquiries or bookings lost where they can be estimated from the business's own history, hours spent on fallbacks and catch-up, campaigns paused, and how the supplier's support handled it. Those notes turn a vague sense of unreliability into evidence.

Availability itself belongs in the contract from the start. The guide to choosing and evaluating tools lists service levels, both for support and for how available the tool is, among the terms worth negotiating, where the use case depends on them. What any supplier offers is for its own contract; the business decides, before signing, how much will depend on the tool and reads the terms with that in mind.

A pottery studio's booking tool goes down (illustrative)

Illustrative example — not a real result

Gwaskothe Pottery keeps taking bookings through a Saturday outage

Gwaskothe Pottery, an invented studio running wheel-throwing classes at 2 sites, takes bookings through an online booking tool and sends reminders from its email platform. Its risk register has a row for the booking tool failing, owned by the studio manager, with a prepared enquiry form and a phone sheet copying the booking tool's fields.

At 09:40 on the Saturday before a school holiday week, staff notice that the booking page shows an error. The studio manager makes a test booking, which fails, checks that nothing on the studio's side changed since Friday, and reads the supplier's published notice confirming a problem with its service. By 10:00 customer service has a short brief, a holding statement is on the site, and the booking button points to the prepared enquiry form.

The manager pauses the class ads that link to the booking page and the reminder workflow that reads from the booking tool, and writes both on the pause list. Staff take 37 bookings by phone on the sheet, with the permission wording read out, and log 2 callers who ask to stop the newsletter.

The service returns at 15:15. The 2 opt-outs go into the suppression list first and are checked in the email platform. The 37 bookings are entered by hand, and the reminder workflow is checked before it restarts: 9 of the phone bookings had already received a confirmation from staff, so the workflow's automated confirmation is skipped for them. The class ads restart after a test booking succeeds.

The window from 09:40 to 15:15 goes into the change log. The manager estimates, from previous holiday Saturdays, that around 20 bookings were lost, and records that figure with the 11 staff hours spent on the fallbacks. Both go into the booking tool's renewal file as a negotiation item.

Common mistakes

  • Planning by tool instead of by job. Only a list of jobs shows which fallbacks to build.
  • Blaming the supplier before checking your own side. An edited embed, expired credentials or a lapsed payment looks like an outage from the outside. Rule them out first.
  • Fallbacks that lose the consent wording. A phone sheet with names and numbers but no record of what people agreed to leaves permissions nobody can show.
  • Deleting instead of pausing. A deleted workflow has to be rebuilt. Pause, and list what was paused.
  • Restarting sends before suppression is current. Opt-outs taken by phone or reply during the outage must be in place before the first message goes out.
  • Releasing held messages without looking. Reminders for past sessions and confirmations already given by hand become double sends. Check who each held message will reach.
  • Treating a stopped form like a tracking outage. When conversions really stop, the problem is lost business, not missing data; pause what leads to the broken step and leave targets alone.
  • Forgetting the outage at renewal. Without a record of the window and the cost, the review has nothing to use.

Frequently asked questions

How do we know the outage is on the supplier's side and not ours?

Rule out the business's own causes first. Check whether anything changed on your side just before the failure: a site release, an edited form or embed, credentials that belonged to someone who has left, a payment method that lapsed or access that was removed. Then reproduce the failure the way a customer would meet it and read whatever notices the supplier publishes, under its own terms. Only when your side is clear and the failure repeats should the incident be treated as a supplier outage, because the fix and the people involved are different.

Should we pause all our campaigns while a tool is down?

No. Pause what sends people to the broken step or depends on the missing tool, and leave the rest running. Ads that lead to a form that cannot be submitted, reminders that read from a booking tool that is not answering and sends that point at a page that no longer works are the candidates. A pause can be undone quickly, whereas a deleted campaign or workflow cannot, so pause rather than remove, and keep a list of what was paused so the restart is complete.

What should we ask the supplier for after an outage?

A written account of what happened and when, in whatever form its terms provide, kept with the incident record. Beyond that, the useful questions are for the business itself: what the outage cost in lost enquiries or orders and in the hours spent on workarounds, whether the fallbacks worked, and whether the tool's availability terms match how much the business now depends on it. Those answers belong in the tool's renewal review, and what the supplier owes under its contract is a matter for its terms and whoever manages the contract.

Terms used on this page

Frameworks

Where this page is used

Sources

  1. Direct marketing guidance — Information Commissioner's Office (UK)

Related