Skip to main content

Does offering compensation to customers who complain on social media invite more complaints? What a Journal of Research in Interactive Marketing study found

When Facebook let brands answer comments directly, airlines that replied to complaints in public drew more negativity than airlines that did not reply. Most of it followed replies that put money back or a perk on display.

Zhang and Zhang find airlines that offered refunds or freebies in public Facebook replies drew more, and harsher, complaints, while apologies added nothing.
Date

7 November 2026

Marketing in 3 · 2025 in research · No. 32 · Paper published 7 August 2025

A smiling woman in a headset and sheer white blouse turns from her keyboard towards the camera, a row of colleagues in headsets working beside her in warm window light
Whoever answers a complaint in public is also talking to every follower who reads the thread. Photo: BaljkanN4 on Unsplash.

The short answer

On Facebook, it did. When the platform added a reply button under comments, airlines that began answering complaints on their brand pages drew more negative comments, and harsher ones, than airlines that never replied. Most of that rise followed replies offering refunds, upgrades or freebies, which seemed to show other followers what a complaint could earn. Replies that apologised added nothing measurable either way.

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

Key takeaways

  • Before the reply button arrived, only 2 of 133 large brand pages answered individual comments; afterwards 78% of the airlines answered complaints.
  • Brands that answered complaints openly picked up about 3.5 extra negative comments per post, on pages that averaged 3.6.
  • A reply mentioning a refund, an upgrade or a freebie was linked to about 4 more negative comments per post than a plain reply.
  • Saying sorry made no measurable difference of its own on any of the 4 outcomes tracked, from how many complaints arrived to how harsh they sounded.

What the researchers did

Yuchi Zhang of Santa Clara University and Kunpeng Zhang of the University of Maryland tackled a dilemma most social media managers meet sooner or later. A customer complains on your brand page. Do you answer in public? Earlier work on review sites pointed both ways, and little of it asked whether the wording of the answer matters.

Their test came from a change Facebook made without warning in March 2013, when it put a reply button under every comment. Until then comments simply piled up in order, and answering one person was clumsy. Overnight, a brand could respond right under a complaint, in full view of its followers. Some airlines took up the new button and others ignored it, so the pair could compare the two groups before and after the switch.

The main analysis follows 67 airline brands from a year before the change to a year after. A sentiment model built to read emoticons and internet slang scored every comment. Replies were sorted by their words into offers of a refund, upgrade or freebie, apologies built on "sorry" or "apologize", and plain acknowledgements, with a topic model as a second way of coding them. Because brands chose whether to reply, the analysis corrected for that choice using each firm's headcount and its time on Facebook. It also controlled for the monthly airline complaint counts kept by the US Department of Transportation, a channel where firms cannot answer back.

Data
Every public post, comment, reply and like on the brand pages, downloaded through Facebook's developer interface
Main sample
67 airline brands, about 1.9 million comments, 24 months of activity
Natural experiment
Facebook's unannounced launch of direct replies to comments in March 2013
Comparison
Brands that began replying to complaints against brands that did not, two months either side of the launch
Measured
The share, number and tone of negative comments per post, plus all comments
Also covered
Hotel and car brands, reported only in an online appendix

What happened when brands replied

The raw figures already tell the story. On the pages of airlines that started replying, negative comments jumped in the first month after the button appeared and stayed above their old level. Pages without replies barely changed.

The model, focused on a short window around the launch, backs this up. Replying brands saw a larger share of their comments turn negative, and the complaints themselves grew sharper: their average sentiment score fell by almost one point (0.99), on a scale where these comments typically sat around minus 2.4. All comments taken together did not rise by a reliable amount. People were not simply talking more. More of what they said was complaint.

Estimated extra comments per post once airlines replied to complaints
Estimated extra comments per post once airlines replied to complaints01234Negative comments: 3.46 Negative comments3.46All comments (not significant): 0.98 All comments (not significant)0.98
Estimated extra comments per post once airlines replied to complaints
Group Value ()
Negative comments 3.46
All comments (not significant) 0.98

Table 2: 67 airline brand pages, two months before and after the March 2013 reply button (268 firm-months), replying brands against non-replying brands.

Why an audience makes it worse

The authors' explanation is an incentive. A reply is written for one customer but read by everyone who follows the page, and an answer that settles a grievance signals that speaking up gets results. If that is right, the effect should be bigger where more people are watching.

That is what the data show. The rise in negative comments was stronger for brands with larger followings, and stronger again where complaints had drawn more likes from other users. Both are rough proxies for how many eyes were on the exchange.

Refunds feed the cycle, apologies do not add to it

What the reply said mattered most. When a brand's replies offered something of value, whether money back, an upgrade or a free extra, negative comments rose further, made up a bigger share of the conversation and turned harsher in tone. The keyword coding and the topic-model coding agreed, and under the second the extra came to about six and a half negative comments per post (6.47).

Plain acknowledgements still came with a few extra negative comments on each post, but with no rise in their share and no change in tone. Apologies made no separate difference on any measure. The pair trace this to an economic motive: an offer posted where everyone can see it tells each follower what a complaint might be worth.

A word of care on how this is summarised. The paper describes apologies as neutralising further complaints. The tables support something narrower. An apology did not add to the rise that came with replying. Calming a page down would be a different, stronger claim.

Extra negative comments per post, by what the reply offered
Extra negative comments per post, by what the reply offered012345Plain reply: 3.12 Plain reply3.12Added when compensation offered: 4.24 Added when compensation offered4.24
Extra negative comments per post, by what the reply offered
Group Value ()
Plain reply 3.12
Added when compensation offered 4.24

Table 4 (keyword coding), same pages and window. Apologies came out at minus 0.63, not significant, so they have no bar.

What the authors advise

Their advice is to keep generosity out of the public thread. Compensation, as they put it, cannot be handled in public without inviting a spiral, while a public apology is the safer route for a brand's reputation. They frame the task as a balance between being generous and being restrained, not as a case for refusing help.

Nor do they tell brands to stop replying. Replying brings complaints into the open where they can be fixed, and the authors note that it also wins followers, so the costs and benefits have to be weighed against sales and loyalty, which this data could not track. Sending offers by private message follows from their reasoning, yet private conversations were invisible to the analysis, so that route remains untested.

Marketing in 3 steps

Ready. Set. Market.

ReadyDecide and prepare
  • Sort your recent replies. Pull your last 50 public replies to complaints. Label each one as a plain acknowledgement, an apology, or an offer of money, credit, an upgrade or something free. Count each pile.
  • Set one rule for offers. Decide that refunds, vouchers and upgrades are only ever offered in a private message, never in the public thread. Write the rule down and brief everyone who posts for the brand.
SetBuild it
  • Write a short public apology. Draft two or three public replies that say sorry, name the problem in plain words and invite the person to message you. Keep each under 40 words and leave out anything they might receive.
  • Settle the case in private. In the direct message, confirm the details, explain what happens next and make any goodwill gesture there. End by asking whether the problem is now sorted.
MarketLaunch and measure
  • Answer the busiest threads first. Each morning, sort open complaints by likes and replies. Reach the top five within two hours, because those are the threads most followers will see.
  • Count what follows each reply. Once a week, count the negative comments that arrive in the 48 hours after each public reply, and note which kind of reply it was. Compare the counts month by month.

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

  • More threads to manage. Public replies on these pages came with extra negative comments, so set aside moderator time for the conversation a reply can start.
  • Visible offers do the damage. The sharpest rise in volume and tone came after replies showing money or perks.
  • Apologies are safe, not soothing. A sorry-led reply left things no worse, but do not count on it to quieten a page by itself.
  • Some upside remains. Replying surfaces complaints a brand can then fix, and responsive pages gained followers, by the authors' account.

Where it does not work

  • One platform, one moment. Everything comes from Facebook around its 2013 reply-button launch; other platforms and later years may behave differently.
  • Big brands only. The pages belong mostly to large companies. The authors expect a small local business might meet a gentler crowd.
  • No sales or satisfaction data. The analysis could not say whether replying helped or hurt sales, or whether the original complainers were happy with the outcome.
  • Private messages unseen. Direct conversations were outside the data, so whether moving offers out of view helps is not shown.
  • Coded from words. Reply types were read from terms such as refund, upgrade, freebie and sorry, plus a topic model, so they capture what replies said rather than what customers received.

Questions marketers ask

Should brands reply to complaints on Facebook?

Replying in public carries a risk. On 67 airline pages, brands that started answering complaints once Facebook added its reply button saw a heavier load of negative comments, and a harsher tone, than brands that did not respond (Zhang & Zhang, 2026). How the reply was worded explained most of the gap.

Is it a mistake to offer a refund in a public reply?

On this evidence it can be. Public offers of refunds, upgrades or free extras were followed by more complaints, a bigger negative share and sharper wording, most likely because other followers saw that complaining paid and posted their own.

Do apologies stop complaints from spreading?

That was not shown. Apology replies did not move later complaints in either direction beyond what any reply did. They did no harm beyond a plain reply, but the data do not show them reducing negativity.

Why would one reply affect other customers?

Because a public reply has an audience. The increase in complaints was larger for brands with bigger followings and for complaints with more likes, both signs that more people saw the exchange.

Does this apply beyond Facebook and airlines?

Not yet known. The data come from one platform, mostly large companies and a window around 2013. Hotel and car brands appear in an online appendix, but other platforms, smaller firms and recent years were not tested.

Also new at the time

Journal of Public Policy & Marketing · 2026

Consumer Evaluations of Generative Artificial Intelligence Disclosures in Responses to Health Care Reviews

Mousavi & Villanova

Health care providers are starting to let generative AI draft their answers to online patient reviews. Drawing on real reviews and several experiments, Mousavi and Villanova show that consumers generally could not tell an AI-written answer from one written by a person. Being told AI was involved changed that, often for the worse, and how much depended on who made the disclosure and how much people felt they knew about AI. If a machine helps with your review replies, who says so may matter as much as what it writes.

Australasian Marketing Journal · 2026

Push the Paw: A Field Experiment on Personalised Push Notifications and User Engagement

Kim, Kim, & Choi

A pet activity-tracking app ran its own field experiment on push notifications with 595 pet caregivers. Tailored messages drew more engagement than generic ones, and naming both the caregiver and the pet beat naming the pet alone. Adding the caregiver's own activity history also lifted engagement, though not significantly more than the names did. Two names in a notification can carry more weight than a cleverer line of copy.

Source

Zhang, Y., & Zhang, K. (2026). Feeding negativity on social media: how compensating customers creates vicious cycles of complaint. Journal of Research in Interactive Marketing, 20(3), 455-471. https://doi.org/10.1108/jrim-08-2024-0411

This brief is written for practitioners from the published article. Open access (CC BY 4.0): free to read at the publisher. Numbers are as reported by the authors; read the article for full methods.

Keep reading

More issues on Social media · 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