What a Customer Apology Must Contain
A recovery after a service failure has three parts, the outcome, the procedure and the treatment, and they do not stand in for one another. The research says money does least of the work, and an apology only works when it comes early, with empathy and an explanation.
Answer to the question
How should a business apologise to a customer?
Early, with empathy for what the failure cost the customer, and with an explanation attached, then fix the problem quickly and repay in kind up to about the value of the loss. Research on service recovery finds compensation is a poor substitute for a fast, apologetic response, and an apology with nothing useful attached does little.
Key takeaways
- Stephen Tax and colleagues found in 1998 that customers judge a complaint's handling on three things: the outcome, the procedure and how they are treated.
- In Jochen Wirtz and Anna Mattila's 2004 experiment, compensation did not raise satisfaction when the response was fast and apologetic, and could not repair a slow one with no apology.
- Katja Gelbrich and colleagues tested eleven compensation levels in 2015 and found satisfaction climbed fastest between about 60 and 120 per cent of the loss.
- A 2013 experiment led by Holger Roschk found empathy was the part of an apology with the strongest effect on satisfaction, and a late apology lowered it.
- Prior goodwill helps only so far: Tax's 1998 study found good past experience softened poor complaint handling only to a limited extent.
When you complain to a company, the reply has three parts: what you are given, how the matter is handled and how you are treated. A customer apology belongs to the third.
The treatment is where it often goes badly. In 1998 Stephen Tax and two colleagues studied customers who had recently complained to a service firm and found that a majority were dissatisfied with how the complaint had been handled. The same paper set out the structure the field still uses. Customers judge a recovery on three things: the outcome they receive, the procedure used to reach it, and the way they are treated along the way. Researchers call these distributive, procedural and interactional justice, and the parts of a recovery sort into them cleanly.
Compensation is outcome, speed and the ease of complaining are procedure, and the customer apology, the explanation and the attention a customer gets are treatment. None of the three stands in for another.
Wayne Simmons and Tom DeWitt give the practitioner's version in The Customer Excellence Enterprise (2025). They call it experience recovery: fixing the problem and restoring trust at once, drawing on what the book calls the goodwill bank, where good experiences are deposits, failures are debits, and the balance decides whether a lapse is forgiven. The bank has no unit and the book offers none. DeWitt's own research gets closer to a measure. In a hospitality scenario study of 459 people, published with two colleagues in 2008, perceived fairness after a recovery predicted trust, and trust carried the whole of fairness's effect on loyalty.
Does the remedy match what was lost?
A recovery gives back what the failure took, and it should give it back in kind: a refund answers an overcharge, and a free dessert does not answer a waiter who never came back to the table. That was the premise of a resource exchange model tested on restaurant and hotel customers in 1999 by Amy Smith and two colleagues, and the scenario experiments bore it out, though most of what follows rests on customers responding to described or staged failures, which cannot carry the anger of a real one. Customers preferred recovery resources that matched the type of failure, in amounts in proportion to how bad it was.
The matching idea held up fifteen years later, in a meta-analysis and a fresh experiment by Holger Roschk and Katja Gelbrich. The recovery effect was generally strongest when compensation was the same kind of resource as the loss: immediate money for a monetary loss, an exchange for a flawed product, doing the service again for a failed service, and attention for a lack of attention. The one exception ran the friendly way. A customer who had been ignored could also be won back by the other kinds of compensation. Money paid at once generally did more than money promised later, though how much more depended on the failure.
Compensation has a ceiling
How much to give was tested directly, with eleven levels of compensation running from nothing up to twice the value of the loss, in two experiments published in 2015 by Gelbrich and two colleagues. Where customers had rejected the flawed service, satisfaction climbed fastest somewhere between 60 and 120 per cent of the loss, and each further step bought less. Where customers had accepted the flawed service and lived with it, the first money offered did the most work. Paying double is a poor use of a budget, and the sizes are those of the experiments, not forecasts for any one firm.
Jochen Wirtz and Anna Mattila varied compensation, speed and the customer apology together in a 2004 experiment, and the money mattered least at the extremes. When the response was immediate and came with an apology, compensation did not raise satisfaction. When the response was slow and came without one, compensation failed to repair it. It helped only in the mixed cases, a slow response with an apology or a fast one without. Their summary was that "compensation is a poor substitute for a good recovery process".
Speed against ease of complaining
Making a complaint easy has a steadier record than answering it fast. Moshe Davidow's 2003 review of the complaint-handling research split a firm's response into six dimensions: timeliness, facilitation, redress, apology, credibility and attentiveness. Facilitation, meaning the policies and channels that make a complaint easy to make, was the most consistent: eight of the nine studies he reviewed found it raised satisfaction. Timeliness was less clear-cut. Davidow found the results mixed and proposed that speed "may only be a critical factor after an unreasonable delay".
That fits the Wirtz and Mattila result, where speed did its best work alongside the customer apology. Speed behaves more like a threshold than a dial, and past it the customer stops counting minutes and starts judging the people.
Mean the apology, then explain
For loyalty, treatment, of which the customer apology is the plainest part, weighs about as much as the payout, and slightly more. A 2011 meta-analysis of complaint handling by Gelbrich and Roschk found that the fairness of the outcome drove satisfaction with the single incident, while the cumulative satisfaction that feeds loyalty depended slightly more on interpersonal treatment than on what the customer was given. That weighting was strongest where the failure did not involve money and in service industries.
Compensation is a poor substitute for a good recovery process.
Holger Roschk led a 2013 experiment that broke the customer apology into three components, empathy, intensity and timing, and varied each. Empathy had the strongest effect on satisfaction, intensity came next and timing last, and a late apology lowered satisfaction. For failures in the process rather than the outcome, an apology did about as much as compensation.
In a telephone experiment led by Nira Munichor, callers were put on hold for the same 108 seconds with music, with music and repeated apologies, or with music and their place in the queue. Apologies did nothing the study could detect to stop people hanging up. Position information nearly halved the share who did.
A customer apology on its own is not a recovery, and the customer apology that works is attached to something the customer can use, which is also why the explanation belongs in it. Davidow calls this credibility, the firm's willingness to account for what went wrong, and in his own earlier study it had a significant effect on satisfaction and on intentions to buy again. The same study rated attentiveness, the quality of the conversation itself, the single most important dimension.
Read together, with most of these studies seen only at abstract level and only Davidow's review and DeWitt's 2008 paper in full, the research describes a recovery as a sequence, and the order matters:
- Apologise at once, with empathy for what the failure cost the customer.
- Explain what happened, plainly.
- Fix it quickly, through a channel that makes complaining easy.
- Repay in kind, in proportion to the loss, and stop near the full amount.
The goodwill bank deserves a place in that account, with a limit attached. DeWitt's earlier work, four studies published in 2003 with a co-author, found that an existing rapport between customer and employee raised satisfaction after a failure, raised intentions to return and reduced negative word of mouth. Tax and his colleagues found that prior good experience softened the damage of poor complaint handling only to a limited extent. Both hold. Banked goodwill buys patience, and it does not pay for a bad recovery. Whether a well-fixed mistake can leave a customer happier than no mistake at all, the service recovery paradox, is a different question, and this piece leaves it open.
The evidence favours a recovery built from treatment, a fast, empathetic, explained response from someone with the authority to act. That has to be paid for on every ordinary day, through trained and trusted staff, because a recovery depends on those people being in place before the failure arrives, and no per-incident refund buys them after it. The staff side of a service failure is examined in an article on [emotional contagion at the counter](https://thewhyweproject.com/articles/emotional-contagion-between-staff-and-customers).
Common questions
How should a business apologise to a customer?
A customer apology works when it is quick, carries empathy for what the failure cost, and comes with an explanation. In Holger Roschk's 2013 experiments, empathy had the strongest effect on satisfaction and a late apology lowered it.
Is compensation more important than an apology?
Not on its own. Wirtz and Mattila found in 2004 that compensation added nothing when the response was fast and apologetic, and could not rescue a slow response with no apology.
How much compensation should a company give after a mistake?
Enough to cover the loss, in the same kind of resource. In Katja Gelbrich's 2015 experiments, satisfaction rose fastest between about 60 and 120 per cent of the loss, and each further step bought less.
Sources
Tax, S. S., Brown, S. W., & Chandrashekaran, M. (1998). 'Customer Evaluations of Service Complaint Experiences: Implications for Relationship Marketing.' Journal of Marketing, 62(2), 60-76. doi:10.1177/002224299806200205
Smith, A. K., Bolton, R. N., & Wagner, J. (1999). 'A Model of Customer Satisfaction with Service Encounters Involving Failure and Recovery.' Journal of Marketing Research, 36(3), 356-372. doi:10.1177/002224379903600305
Davidow, M. (2003). 'Organizational Responses to Customer Complaints: What Works and What Doesn't.' Journal of Service Research, 5(3), 225-250. doi:10.1177/1094670502238917
DeWitt, T., & Brady, M. K. (2003). 'Rethinking Service Recovery Strategies: The Effect of Rapport on Consumer Responses to Service Failure.' Journal of Service Research, 6(2), 193-207. doi:10.1177/1094670503257048
Wirtz, J., & Mattila, A. S. (2004). 'Consumer responses to compensation, speed of recovery and apology after a service failure.' International Journal of Service Industry Management, 15(2), 150-166. doi:10.1108/09564230410532484
Munichor, N., & Rafaeli, A. (2007). 'Numbers or apologies? Customer reactions to telephone waiting time fillers.' Journal of Applied Psychology, 92(2), 511-518. doi:10.1037/0021-9010.92.2.511
DeWitt, T., Nguyen, D. T., & Marshall, R. (2008). 'Exploring Customer Loyalty Following Service Recovery: The Mediating Effects of Trust and Emotions.' Journal of Service Research, 10(3), 269-281. doi:10.1177/1094670507310767
Gelbrich, K., & Roschk, H. (2011). 'A Meta-Analysis of Organizational Complaint Handling and Customer Responses.' Journal of Service Research, 14(1), 24-43. doi:10.1177/1094670510387914
Roschk, H., & Kaiser, S. (2013). 'The nature of an apology: An experimental study on how to apologize after a service failure.' Marketing Letters, 24(3), 293-309. doi:10.1007/s11002-012-9218-x
Roschk, H., & Gelbrich, K. (2014). 'Identifying Appropriate Compensation Types for Service Failures: A Meta-Analytic and Experimental Analysis.' Journal of Service Research, 17(2), 195-211. doi:10.1177/1094670513507486
Gelbrich, K., Gäthke, J., & Grégoire, Y. (2015). 'How Much Compensation Should a Firm Offer for a Flawed Service? An Examination of the Nonlinear Effects of Compensation on Satisfaction.' Journal of Service Research, 18(1), 107-123. doi:10.1177/1094670514543149
Simmons, W., & DeWitt, T. (2025). The Customer Excellence Enterprise: A Playbook for Creating Customers for Life. Hoboken, NJ: John Wiley and Sons. ISBN 9781394253685