Skip to main content

Should a brand say sorry or promise a fix when an angry complaint starts spreading on social media? What a Journal of Marketing Communications study found

We put a furious Facebook post about a brand in front of 564 people in Portugal, followed by the company’s response. A good response made them think better of the brand, and that left them less keen to spread the post. Saying sorry gave this the most force overall; for a service slip or a deleted comment, describing the fix did as well.

Langaro et al. tested brand replies to an angry Facebook complaint with 564 users: an apology curbed sharing best, but a fix matched it for service failures.
Date

7 November 2026

Marketing in 3 · 2024 in research · No. 14 · Journal issue published 2 April 2024

Silhouetted heads at night in front of a large burst of fire and sparks, one person holding up a phone that is live-streaming it
Every raised phone helps the blaze travel. What the brand says next can change how far it goes. Photo: Chitto Cancio on Unsplash.

The short answer

Apologise by default, but some complaints are met just as well with a fix. In our online experiment, 564 Facebook users in Portugal read an angry post on a brand's page, then the company's answer. Those who came away with a better opinion of the brand were less inclined to share the complaint, and that brake worked hardest after an apology. When the complaint was about poor service or a deleted comment, explaining the fix worked just as well.

Evidence level
  1. One study
  2. Several studies
  3. Meta-analysis

Key takeaways

  • People put the blame squarely on the brand, rating its responsibility at 5.39 out of 7: the kind of preventable crisis in which a brand is expected to own the problem.
  • Each extra point in people's opinion of the brand after its reply went with a 0.18-point drop in their readiness to share the complaint, with age and gender held constant.
  • The wish to pass the complaint on still averaged 3.69 after the reply, a little under the scale's midpoint of 4: a reply works as a brake, not an off switch.
  • In 2 of the 3 kinds of complaint we tested, a customer service failure and a deleted critical comment, the apology's advantage vanished and a corrective reply performed on a par with it.

What we did

Daniela Langaro, Sandra Loureiro and Helder Neves of the Business Research Unit at ISCTE, the University Institute of Lisbon, and I, then based at RMIT University in Australia, started from a familiar scene. One customer posts an angry comment on a brand's Facebook page, others pile in, and within hours the brand is facing what researchers call an online firestorm. Crisis theory says that when the brand is to blame it should accept responsibility. It does not say whether saying sorry or promising a fix is the better way to do that, or why either one would stop people from passing the complaint on.

So we built a small crisis of our own. We invented a retailer, Lusoretail, so that nobody arrived with an opinion of it, and mocked up posts that looked like Facebook. Each person saw one complaint, rated the brand, read one of two replies from the company, rated the brand again and then said whether they would forward the complaint. The complaints covered the three sparks that most often set off firestorms: something unethical, a failure in the core service, and clumsy handling of the brand's own communication.

Participants
564 adults in Portugal who use Facebook, aged 18 to over 65; 58.3% men; 57.6% on Facebook several times a day
Recruitment
Survey link posted in the Facebook networks of students and private and public sector workers; no incentive; people asked to pass it on
Complaints
One of three, at random: dumping waste in the city's river, two weeks without reaching customer service, or deleting a critic's comment
Replies
One of two, at random: an apology accepting responsibility, or corrective action now plus steps to stop a repeat; 282 people saw each
Measured
Brand attitude after the complaint and again after the reply (five seven-point items), then whether they would pass the post on (two items)
Checks
Questionnaire translated into Portuguese and back again; every participant correctly identified the reply they had read

The reply changes minds, and minds decide what gets shared

The first question was whether the reply moved anyone at all. It did. Straight after the complaint, people's view of Lusoretail sat deep in the negative half of the scale. Once the brand had answered, it climbed by well over a full point, for every type of complaint.

The second question matters most in a firestorm. Did feeling better about the brand make people less willing to spread the complaint? Yes. The more favourably someone rated the brand after its answer, the weaker their wish to forward the post. So the reply does its work through the reader's attitude to the brand, and that attitude then acts as the brake on sharing. For anyone writing a reply, this shifts the target. It is not enough for the reply to be correct. It has to leave readers liking the brand a little more.

Opinion of the brand, before and after its reply (rated 1 to 7)
Opinion of the brand, before and after its reply (rated 1 to 7)1234567After the complaint: 2.21 After the complaint2.21After the brand's reply: 3.54 After the brand's reply3.54
Opinion of the brand, before and after its reply (rated 1 to 7)
Group Value ()
After the complaint 2.21
After the brand's reply 3.54

Brand attitude (five seven-point items), 564 participants, both replies pooled. The rise was significant and held for all three complaint types.

Why the apology pulled harder

Here the two replies parted ways. After an apology, brand attitude made a large difference: readers who had come round to the brand were far less keen to share the complaint than readers who had not. After the corrective message the line was almost flat: opinion of the brand hardly changed what readers planned to do with the post.

We think forgiveness is the reason. Saying sorry in public is a small act of humility, close to a symbolic punishment the brand accepts in front of everyone, and it invites readers to forgive. Once they do, they have little reason to keep the complaint moving. A promise to fix things restores the situation but asks for nothing emotional in return. We did not measure forgiveness itself, so this is our explanation, and a test for future work.

The chart carries a warning too. At the low end of the scale, readers who were still unconvinced after an apology leaned further towards sharing than equally unhappy readers of the corrective message. An apology raises the stakes: it pays off when it wins people over. We did not test that pattern on its own, so take it as something the figure shows rather than a firm result.

Intention to forward the complaint at low and high brand attitude, by reply type (rated 1 to 7)
Intention to forward the complaint at low and high brand attitude, by reply type (rated 1 to 7)1234567Apology, low brand attitude: 4.4 Apology, low brand attitude4.4Apology, high brand attitude: 3.4 Apology, high brand attitude3.4Fix, low brand attitude: 3.6 Fix, low brand attitude3.6Fix, high brand attitude: 3.4 Fix, high brand attitude3.4
Intention to forward the complaint at low and high brand attitude, by reply type (rated 1 to 7)
Group Value ()
Apology, low brand attitude 4.4
Apology, high brand attitude 3.4
Fix, low brand attitude 3.6
Fix, high brand attitude 3.4

Values read from the end points of Figure 1 of the article, at brand attitudes of about 1.8 and 5.1; 564 participants, 282 per reply. The slopes differed significantly between the two replies.

Service failures and deleted comments: either reply will do

We also looked inside each type of complaint. For the two about a brand failing to deliver, the customer who could not reach customer service for more than two weeks and the customer whose comment had vanished, the apology lost its edge. Saying sorry and describing the fix worked about equally well.

Earlier research points to an explanation. Problems like these are familiar: many readers have been left on hold or ignored by a brand themselves, or expect it to happen to them. What they want to hear is that the problem is being solved, and a corrective reply tells them exactly that. A pollution scandal is different. It is a moral judgement that fewer readers have lived through, which probably makes an admission and an apology matter more.

For a social team this is useful freedom. When the complaint is operational, the brand can pick whichever reply its operations and legal teams are able to stand behind.

Why I think replies matter even more now

We ran this on Facebook, with one complaint and one reply. Since then the places where firestorms start have multiplied, and a complaint can be screenshotted, stitched into a video or quoted in a thread within minutes. What I do not think has changed is the mechanism. Whether people pass a complaint on depends heavily on their opinion of the brand in that moment, and the reply is among the very few levers a brand holds over that opinion.

Three habits worry me in how brands answer today. The first is non-response: research we drew on found that 30% of negative posts on social media receive no response at all (Herhausen et al., 2019). The second is replies written to say as little as possible. Because the reply works through liking, a statement cleared by a legal template, or drafted by an AI tool in a flat corporate voice, can be safe on paper and still win nobody back. The third is newer: automatic filters that hide critical comments. Deleting a comment was one of the three sparks we tested. I suspect, though our data cannot show it, that a filter removing complaints without notice can start the very storm it was meant to prevent.

Marketing in 3 steps

Ready. Set. Market.

ReadyDecide and prepare
  • Answer every complaint that could spread. Reply in public to any angry post on your page that others are liking, sharing or commenting on. Aim to have your first reply up within two hours, evenings and weekends included.
  • Sort the complaint into one of three boxes. Before you write a word, decide which kind it is. Box one: something wrong the brand did, such as harming the environment. Box two: a failure in the service itself, such as customers who cannot get through. Box three: a mistake in how the brand treated people online, such as removing their comments.
SetBuild it
  • Say sorry for box one. For wrongdoing, apologise in plain words and accept responsibility. Keep it to two or three sentences and sign it with a real person's name.
  • Pick the reply you can deliver for boxes two and three. For service and comment-handling problems, either apologise or say exactly what is being fixed and how you will stop it happening again. Agree both versions with your operations and legal teams in advance, so you can post the one they will stand behind.
MarketLaunch and measure
  • Write the reply to be liked. Read the draft aloud to three colleagues who have not seen the complaint. If any of them finds it cold, defensive or full of jargon, rewrite it until all three call it warm and fair.
  • Watch where the post travels. For the next 48 hours, check every few hours how often the complaint is being shared and how many new angry comments arrive. Once a month, compare those counts after apologies and after fix-it replies, and adjust your scripts.

Get it by email

New issues every Monday and Thursday, in your inbox. Free, no ads, unsubscribe anytime.

Your address is used only to send Marketing in 3. You will get one email to confirm.

What to expect

  • A partial recovery. A good reply improves how onlookers see the brand, but in our test their view was still on the negative side afterwards. The complaint had already done its damage.
  • Two scripts, not one. Teams that keep an apology and a corrective reply ready can match the reply to the complaint rather than reaching for one formula every time.
  • More work behind the scenes. Choosing case by case means training community managers and settling with legal and operations, before any crisis, what can be said.
  • Success is hard to see. A reply that works shows up as a complaint that stops travelling: fewer shares and fewer pile-on comments, not a wave of praise.

Where it does not work

  • An invented brand. Real brands bring a crisis history and a reputation, which shape how a complaint is read. We kept both out on purpose.
  • Keen Facebook users in one country. Volunteers recruited on Facebook in Portugal, most of them on it daily. Other platforms and cultures may react differently.
  • Intentions, one post, one moment. We measured willingness to forward, not actual shares, and each person saw a single complaint, not the pile-up of a real firestorm.
  • No comparison without a reply. Every participant read a reply, which leaves open how much a reply adds compared with not replying.
  • Complaints not perfectly matched. The three were written to be equally angry but differed in wording. We did not report a separate test for the pollution complaint, so the case for apologising after an ethical lapse rests on the overall result.

Questions marketers ask

Should a brand apologise or explain the fix when a complaint goes viral?

As a default, apologise. When we tested replies on Facebook users, warmer feelings towards the brand cut the urge to share the complaint more sharply after an apology than after a corrective message, although for complaints about customer service or removed comments the two worked about as well (Langaro et al., 2024).

What is an online firestorm?

A sudden flood of angry posts about a brand, usually starting from one complaint that others join and share. They tend to follow something people believe the brand could have prevented: an ethical lapse, a failure in its main service, or poor handling of its own communication.

Does replying to complaints on social media actually help?

Yes, and not only with the person complaining. The readers who decide whether a firestorm spreads are onlookers. Their view of the brand improved once it answered, and a better view went with less wish to pass the post on.

Is it risky to delete negative comments on a brand page?

It can be. Earlier research counts clumsy or unfair handling of communication among the most common triggers of a firestorm, which is why one of our test complaints was about a deleted comment. In our test, answering that complaint, with an apology or a fix, improved how onlookers saw the brand.

What should a reply to an angry post sound like?

Likeable. The reply works by changing onlookers' opinion of the brand, so its wording and structure should come across as pleasant and agreeable rather than defensive. A short, warm, human answer is a better bet than a legal statement.

Also new at the time

Journal of Marketing Research · 2024

A Topic-Based Segmentation Model for Identifying Segment-Level Drivers of Star Ratings from Unstructured Text Reviews

Kim et al.

A new machine learning model reads the text of online reviews and works out which words really push star ratings up or down, while grouping businesses whose customers care about different things. Run on Yelp reviews of 5,241 restaurants in Arizona, it found three such groups, each with three to five topics that mattered most, and it predicted ratings better than six rival models. Reading reviews group by group, instead of watching the average star score, shows which complaint is costing which kind of business.

Journal of Consumer Psychology · 2025

The mobile giving gap: The negative impact of smartphones on donation behavior

Hock et al.

People were less likely to give to charity on a smartphone than on a computer, and the gap showed up in all three studies. One asked for real money, another traced the drop to people thinking less about others while on their phones, and Google display ads run in the field showed the same pattern. Appeals centred on the people being helped narrowed it. A donation page that converts on a laptop may need a different pitch on a phone.

Source

Langaro, D., Loureiro, S., Schivinski, B., & Neves, H. (2024). In the eye of the (fire)storm: better safe or sorry? Crisis communication strategies for managing virality of online negative brand-related content. Journal of Marketing Communications, 30(3), 301-317. https://doi.org/10.1080/13527266.2022.2109056

This brief is written for practitioners from the authors' accepted manuscript of the article, which matches the published text apart from layout and page numbers. Numbers are as reported by the authors; read the article for full methods.

Keep reading

More issues on Social media · UGC · Service

2026 · No. 2 ·

How to answer customers who praise you online

Replying to a 5-star review by downplaying the praise made 60.8% of people choose that business over one that repeated the compliment. Wording that works, and what to avoid.

2024 · No. 46 ·

The bad reviews that repeat themselves cost the most

Tran et al. find bad reviews dwelling on one problem cut hotel sales; a heartfelt apology softens them, while compensation suits posts that keep citing causes.

Bruno Schivinski

Bruno Schivinski is Associate Professor of Marketing at Gdańsk University of Technology. He studies how consumers engage with brands in digital and social media.

About Bruno · All Marketing in 3 issues · Subscribe