Playbook replay
Crisis comms framework
A severity ladder, who speaks at each level, a first-hour template, and the three sentences that make a recall worse.
Run for Lumira, a demo brand we invented for this page. No customer data. Nothing here runs against a live account.
Every session here was run once and saved, against a demo brand or public data. On your own account Naise works from your brand voice, your campaigns and your connected channels, so it reasons about your data instead of a stand-in.
Triggers have to be objective. "Significant customer concern" is not a trigger, it is a debate. Counts and sources are triggers.
Building the ladder first. The hardest part of a recall is not the statement, it is that nobody agreed in advance what counts as level 2, so the first hour is spent arguing about whether this is serious.
Severity ladder
| Level | Trigger, objective | Who speaks | Response time |
|---|---|---|---|
| 1 | Under 5 reports, no medical language, no press contact | Customer Care, in channel | Same day |
| 2 | 5 or more reports in 48h, OR any report mentioning a doctor | Head of Customer Care, public post plus direct outreach | 4 hours |
| 3 | Any press enquiry, regulator contact, or a post over 10k views | Co-founder, statement plus named contact | 1 hour |
| 4 | Confirmed contamination, or any hospitalisation claim | Co-founder, with legal and regulator in the loop | Immediate, and stop sales first |
Level 4 has one rule above all others: stop selling before you finish drafting. Every recall that became a scandal did the statement first.
First-hour holding statement
We are aware of reports that [PRODUCT, BATCH] has caused [REACTION] for a small number of customers. We have [ACTION TAKEN, e.g. paused sales of this batch] while we [WHAT YOU ARE CHECKING]. If you have this batch, [WHAT THEY SHOULD DO]. We will update by [TIME, within 24h]. Anyone affected can reach us directly at [CHANNEL].
Every bracket is a fact, not a judgement. That is deliberate: it can be filled in by whoever is awake.
The three sentences never to say
- "We take safety extremely seriously." It is the sentence every brand uses at exactly this moment, so it reads as a template rather than a response. It also says nothing, which is noticed.
- "There is no evidence of any issue." Absence of evidence, one hour in, is absence of investigation. If it later turns out there was an issue, this is the line that gets replayed.
- "We are the only ones affected by this supplier problem." Naming a supplier or spreading blame before facts are settled turns one crisis into two.
Channel order, first 24 hours
Direct to affected customers first, then the channel where the complaint appeared, then owned social, then press if level 3 or above. Never press before customers. Someone finding out their product was recalled by reading it in a newspaper is a second story.
Voice note. Your brand voice rules stay in force here, with one exception: rule 3, no intensifier without a number, is doing real work in a crisis. "A small number" should become the actual number as soon as you have it.
Where this stops. This is a framework written in calm conditions. It is not legal advice, and the level 4 path in particular needs your lawyer and your regulator obligations mapped onto it before anyone relies on it.
Naise created an artifact
Four severity levels with objective triggers, a fill-in-the-blanks first-hour statement, and the lines that make it worse.
Severity ladder
| Level | Trigger | Who speaks |
|---|---|---|
| 1 | Under 5 reports | Customer Care, in channel |
| 2 | 5+ in 48h, or any doctor mentioned | Head of Customer Care |
| 3 | Press, regulator, or a post over 10k views | Co-founder |
| 4 | Confirmed contamination | Co-founder, stop sales FIRST |
Never say
- "We take safety extremely seriously."
- "There is no evidence of any issue."
- "We are the only ones affected by this supplier problem."
Not legal advice. Level 4 needs your lawyer's path mapped onto it.
That was a sample. Run it on your own brand.
Naise reads your brand, your channels, and your market, then does this against your real data.
Run this playbook free