Algeria AgencyDigital marketing
The week of an incident: what to do hour by hour
An incident is handled in time, not by channel. The first hour, the holding line, the third day, and what actually changes afterwards.
The article accompanying this one describes the places people write about you and what effort each demands. It is a map, and it is useful as long as nothing is happening.
This page is about the moment something does: a complaint that catches, a video, a screenshot going round, an unhappy customer writing on three surfaces at once. At that point the map is not enough, because the problem is no longer "where" but "when".
The main difficulty is misplaced, and that is what this article corrects first. It is almost never in what to write; it is in estimating the real size of the event, which is systematically overstated by the people living it from inside.
Then the sequence: the first hour, what you publish while you verify, who decides, the public reply, the third day when you have to stop, and the following week, where the only lasting thing is settled.
An incident is measured before it is handled
The first reflex, and the most expensive, is to answer the felt size rather than the real one. Inside the business fifteen people look at the same screenshot within the same hour, each forwards it, and the event occupies a hundred per cent of internal attention. From outside it may be four comments.
The measurement takes ten minutes and is the only thing to do before any writing. Count the distinct people who have written something — not views, not reactions, not internal forwards. The figure is almost always an order of magnitude smaller than the meeting assumed.
That difference is not a detail of perception, it decides the response. Four people are handled with four individual replies, within an hour, and it is over. The same four handled with a public statement inform thousands of people that there was a problem they had not heard about.
There is one exception to this rule and it has to be checked explicitly: personal safety or a legal obligation. A dangerous product, exposed data, an error putting somebody at risk — those are handled publicly whatever the number of people talking, and they are rare.
Write the figure and the time down somewhere. It is what will let you know tomorrow and the day after whether the event is growing or dying — a question nobody can answer from memory, because memory records intensity rather than volume.
The first hour: read before writing
The first hour is not for replying, it is for understanding, and it is the part internal pressure compresses most. Three things have to be established before a single public sentence is written.
First: what actually happened? Not what is being alleged, but what took place, established with the person who was there. A public reply built on the version received by telephone contradicts itself twenty-four hours later, and it is the contradiction that makes incidents last, not the original fault.
Second: is it isolated or systemic? An unhappy customer is an unhappy customer; five people saying the same thing in the same week describe an operational problem, and the answer then lies somewhere else entirely. The distinction is made by looking at the last two weeks of complaints, not by judging the tone of the message.
Third: where have people got to? Read every message, including comments under comments, before replying to one. The subject almost always drifts within hours — it starts on a product and continues on how you replied — and answering the original message once the conversation has moved produces an off-topic reply.
During that hour one thing only is published, and it is the subject of the next section. Complete silence while you verify is correct for an hour and becomes a problem beyond it.
The holding line: one sentence while you verify
You have to publish something quickly and you have to assert nothing, and those two requirements are reconciled by a sentence prepared in advance that acknowledges without ruling on substance.
It contains three elements and nothing else: we have seen it, we are checking, we will come back with an answer by such a time. The third is what makes it useful — without a stated deadline the sentence reads as a way of buying time, which it is, but without the return.
It never contains an apology or a justification. An apology before verification is either premature or false, and a justification read before the facts are established turns systematically against whoever wrote it. "We are sorry you felt that way" belongs in this category and does more damage than silence.
It is published where it started, and only there. Repeating it on three other surfaces informs three audiences who did not know — the same error of scale section 1 describes, committed this time in the very act meant to limit the damage.
Prepare it now, not on the day. A three-line sentence written calmly and approved in advance is available in fifteen seconds when needed; the same sentence written under pressure takes forty minutes and a meeting, which is exactly how long the absence of a reply is noticed for.
Who decides, and who does not write
An incident immediately reveals whether roles have been assigned, and it is too late to assign them then. Three roles are enough and they must be named in advance.
The one who decides: one person, reachable, who rules on substance and on any commercial gesture. Not a committee. An incident waiting for a two o’clock meeting has already changed in nature by two o’clock.
The one who writes: also one person, for a reason that is practical rather than hierarchical. Two people replying in parallel on the same surface produce two different formulations, and the gap between them becomes the new subject.
The one who checks the facts: somebody from operations, not communications. It is the most frequently missing role, and its absence is the direct cause of the denial that has to be withdrawn the next day.
One rule completes the arrangement and must be stated explicitly to the whole team, once, in calm conditions: nobody else speaks, including from their personal account, including in the business’s defence. An employee replying to an unhappy customer from their own account turns a business problem into a conflict between two people, which is the hardest form to stop.
The public reply: once, in the right place
Once the facts are established, the reply is published once, where the incident began, under the most visible message. Not as a separate post, which appears to people who saw nothing.
Its structure has four elements in this order: what happened, in one sentence and without the passive; what is being done for the person concerned; what changes so it does not recur; and a way to reach you directly. The order matters — starting with context makes the rest read as an excuse.
On responsibility, one question settles it: did the thing complained of happen? If so, acknowledge it, simply, without explaining why. Explanations, even accurate ones, are read as mitigation, and they lengthen a reply that gains from being short.
If it did not happen, the reply is different and legitimate: state what did happen, factually, without impugning the person. "Here is what our records show for this order" is usable; "this customer is lying" never is, however certain you are.
And one thing is always done privately after the public reply: contact the person directly. The public reply addresses readers; the private conversation addresses the customer. Confusing the two audiences is what produces over-long public replies and customers who stay unhappy.
What is never written
Four formulations recur in nearly every badly handled incident, and they are recognisable in advance, which allows them to be banned once and for all.
"We are sorry you felt…": that is not an apology, it is a challenge to the reader’s perception, and it is identified as such immediately. If you apologise, apologise for what happened; otherwise do not apologise.
"Contact us by private message" as the only reply: the sentence is fine as an addition and destructive alone, because it reads as an attempt to move the conversation out of sight. Reply publicly on substance, then offer the private channel for detail.
Any mention of footfall, seniority, or the number of satisfied customers. "We serve two thousand customers a month" in reply to a complaint tells the reader their case is statistically negligible, which may be true and has never convinced anybody.
And any reply written in the hour after a tense meeting. The practical rule is this: the reply is drafted, then reread twenty minutes later by somebody who was not in the conversation. Those twenty minutes remove on average one sentence, and it is almost always the one that would have prolonged the affair.
When it moves to another surface
An incident changes nature when it leaves where it was born — when a screenshot of a comment becomes a post elsewhere, or when somebody makes a video of it. That moment calls for an explicit decision and is usually handled by reflex.
The general rule is not to follow. Replying on the new surface validates the transfer, informs its audience of the origin, and commits you to a conversation whose frame and tone you do not control. Most such pick-ups die within forty-eight hours if nobody feeds them.
Two exceptions justify following, and they are verified rather than felt. If the pick-up contains a factually false assertion that harms you lastingly — not a harsh opinion, a false fact. And if the destination surface is where your customers actually are, in which case saying nothing where they are is a default rather than a decision.
In both cases the reply is the same as the original, with no addition and no adaptation to the local tone. A reply rewritten to "speak the language" of the new surface reads as a change of version, and the comparison between the two becomes a subject in itself.
Finally, note what actually triggers pick-ups, because that can be prevented: it is almost never the original problem, it is a reply judged contemptuous. What circulates is rarely the complaint; it is the screenshot of your answer.
The third day: stop
There is a point at which continuing to reply sustains the incident instead of settling it, and it comes sooner than expected — generally after forty-eight to seventy-two hours.
The sign is clear and is read in the messages rather than in the count. When new comments carry no new information, when the same people are returning, and when the conversation is about your replies rather than the original fact, it is over and only your participation keeps it alive.
The exit is a final reply that closes without withdrawing: what was done, what changes, and how to reach you for any particular case. Then you stop replying in the thread, without deleting anything and without closing comments.
Deleting nothing is the hardest point to hold and the most important. A deleted thread is noticed, the deletion becomes the new subject, and screenshots exist already anyway. The only thing that gets removed is what is illegal or abusive, and that is removed by reporting it to the platform rather than hiding it.
Do keep replying privately to people who write directly, for as long as they do. The end of the public incident is not the end of the individual cases, and half of what will remain is settled there.
The people who defend you
In nearly every incident, satisfied customers spontaneously come to your defence, and the phenomenon is badly used because it is badly understood.
What must not be done is to organise it. Asking customers to write something produces messages that resemble each other, arrive together, and are identified within minutes — after which the incident changes subject and becomes about manipulation, which is far more serious than the original matter.
What should be done is simply to thank them, individually and soberly, without amplifying. Do not reshare those messages during the incident: a business promoting its defenders while being criticised puts itself in the position of winning a debate, and nobody wants to watch a business win a debate against a customer.
There is however something to do afterwards, and nobody does it: contact those people again a week later, privately, to thank them. They are your most attached customers and you have just identified them for free — it is the only net benefit of an incident, and it disappears if nobody writes their names down.
The following week
The incident ends and the business moves on, which guarantees it will recur. The following week is the useful part, and it fits in an hour.
Three questions, asked with operations people rather than communications. Did the original problem actually exist? If so, what changes concretely, and who owns it? And above all: how many times has this happened without anybody writing publicly?
The third is what earns the meeting. A public incident is a sample of a silent phenomenon, and the interesting question is never the person who spoke, it is the twenty who did not. The figure sits in the complaints received by telephone, which are almost never set beside the public event.
Check too what your reply produced, by rereading the thread cold. It is the only moment you can read your own replies without the pressure of the moment, and it is where you see the one that should have been cut before publication.
Finally, update the holding line and the list of roles if the week revealed something missing. It is a one-page document; adjusting it takes ten minutes once a year, and it is the difference between an incident managed and an incident endured the next time.
What is not an incident
A large share of what triggers a crisis meeting does not warrant one, and recognising them saves most of the energy spent in this area.
An isolated negative review is not an incident. It is answered once, calmly, and the answer addresses the next reader rather than its author. Treating it as a crisis gives a message three people would have seen an importance it did not have.
A competitor criticising you is not one either. The response is not to respond: a public discussion between two businesses interests the two businesses and nobody else, and the audience that sees it remembers that there is a conflict, not who was right.
A person who returns regularly, over months, is not an incident but a lasting situation, and it calls for an entirely different treatment — repetition changes everything, and it is the subject of the neighbouring article on video surfaces, where the case is common.
Finally, an operational problem being discussed publicly is not a communications problem. If deliveries are two weeks late, no written reply settles anything, and the only useful thing communications can do is tell the truth about the delay and about when normal service returns.
What we do, and what we refuse
What we do is bounded. We prepare the holding line and the list of roles in calm conditions, we produce the count of distinct people on the day, we draft the replies and have them read before publication, and we run the following week’s review with operations.
We refuse to draft a public statement for a volume that does not warrant one. It is the most frequent request in the first two hours, and it comes from a size estimate made from inside. A statement informs thousands of people of a problem four had seen, and it cannot be undone.
We also refuse to reply in your name during an incident. We write, you publish, and that apparent slowness is a protection: a reply published in your name by a supplier commits the business on facts the supplier did not verify themselves.
And we do not promise that anything will disappear. Published content is not removed for being unfair, only for being illegal or against a platform’s rules, and a supplier promising otherwise is selling something that does not exist. There is also preparation only you can do and it is worth more than all the rest: decide today, in writing, who decides.
Frequently asked questions
Should we reply immediately?
Publish a holding line within the hour — we have seen it, we are checking, we will come back by this time — and nothing else. The first hour is for establishing what actually happened with the person who was there. A substantive reply built on a version received by telephone contradicts itself the next day, and it is the contradiction that makes an incident last, not the original fault.
How do we know whether it is serious?
Count the distinct people who have written something — not views, not reactions, not internal forwards. The figure is almost always an order of magnitude smaller than the meeting assumes, because fifteen people looking at the same screenshot produce a sense of saturation. Four people are handled with four replies, not a statement.
Should we publish a statement?
Almost never. A statement informs thousands of a problem a few had seen, and it cannot be undone. The two cases that justify one are personal safety and a legal obligation, and they are verified rather than felt. Reply where it began, once, under the most visible message.
Can we delete the comments?
No, except illegal or abusive content, which is reported to the platform rather than hidden. A deleted thread is noticed, the deletion becomes the new subject, and screenshots already exist. Do not close comments either: it reads as an admission and moves the conversation somewhere you can no longer see it.
Somebody has picked it up elsewhere. Should we reply there?
As a rule no: replying validates the transfer and informs an audience that did not know, and most pick-ups die within forty-eight hours if nobody feeds them. Two verifiable exceptions — a lasting factually false assertion, or a surface where your customers actually are. And note what triggers pick-ups: it is almost always a reply judged contemptuous, not the original problem.
When is it over?
When new messages carry no information, the same people are returning, and the conversation is about your replies rather than the original fact — generally after two to three days. Close with a final reply, stop replying in the thread, and continue privately with those who write directly.
Where we come in
A holding sentence written on a calm day is ready in fifteen seconds when it is needed. What is still missing is the count that says whether this is an incident or a bad day.
- We set the numeric threshold that triggers the procedure, before it is used.
- We stay reachable through the week, at the hours when it is decided.
- We write the replies, you publish them, including at night.
- We run the third-day review, when everybody is tired.
If you do not have three people free in the same week, do not start: a procedure held by one person breaks on the second day.
Read next
Online reputation: five places, and you watch one
What gets written about you is spread across five surfaces. Most businesses monitor one, and it is almost never the one that decides.Your Google listing: what anybody can change about you
Your business listing is a public record. Strangers can change its opening hours, answer questions on it, and add photographs to it.The arithmetic of a rating: why volume is the only lever
A one-star review is a catastrophe at twelve reviews and a footnote at two hundred. What that implies, and what to do instead of panicking.
Let us talk about your project
A free audit, no commitment: we look at your online presence and tell you what is holding it back.