Every service company fails a customer sooner or later. What it does next is service recovery. The research puts the priorities in order. Avoid the failure first, then fix it fast and fairly.

What is service recovery?

Service recovery is the set of steps a company takes to fix a failure and win the customer back. A 1990 Harvard Business Review article brought the idea to managers. Its authors were Christopher Hart, James Heskett and W. Earl Sasser Jr. They wrote that errors are inevitable in service, but dissatisfied customers are not. A good recovery, they said, can create more goodwill than if things had gone smoothly (HBR, 1990).

The article opened with a flight to Club Med-Cancun that left 6 hours late, made 2 unexpected stops and ran out of food. The resort’s general manager took half his staff to the airport. They set up snacks and drinks, played lively music and gave each guest a personal greeting. A chauffeured ride and a banquet at the resort followed (HBR, 1990).

Recovery is judged on three counts. Stephen Tax and his co-authors found that customers weigh the outcome they get, the procedure used to reach it and the way they were treated. Most of the complaining customers they studied were unhappy with how their last complaint was handled (Tax et al., 1998).

Does a good recovery make customers more loyal?

The 1990 claim became known as the service recovery paradox. In it, a customer whose problem was well solved rates the firm as high as, or higher than, a customer who never had a problem. Later studies tested the idea, and the results are mixed. The table sums up six of them.

StudySettingWhat it found
Hart, Heskett and Sasser, 1990Article on service firmsA good recovery can create more goodwill than if things had gone smoothly
Smith and Bolton, 1998Two experiments in which real customers faced a hypothetical failureExcellent recoveries can lift satisfaction and repatronage, but treating failures as chances to impress carries substantial risks
McCollough, Berry and Yadav, 2000Two scenario experimentsSatisfaction was lower after a failure and recovery, even with high recovery performance, than after error-free service
Maxham and Netemeyer, 200220-month field study of complaining customersA paradox can follow one failure but not two, and two bad recoveries in a row cause “double deviations”
de Matos et al., 2007Meta-analysisPositive and significant for satisfaction, not significant for repurchase intent, word of mouth or corporate image
Michel and Meuter, 2008More than 11,000 bank customer interviewsThe paradox is rare, and the differences are significant but not large

Read together, the studies say a great recovery can lift satisfaction. It does not reliably lift repurchase, and it does not beat error-free service. McCollough and colleagues found that firms fare better by avoiding failure than by recovering well (McCollough et al., 2000). The meta-analysis found no effect on repurchase intent (de Matos et al., 2007). The bank study calls the paradox rare (Michel and Meuter, 2008).

Two failures change the picture. In a 20-month field study, a good recovery produced a paradox after one failure but not after two (Maxham and Netemeyer, 2002). Smith and Bolton add a warning. Treating a failure as a chance to impress customers can carry substantial risks (Smith and Bolton, 1998).

What makes a service recovery work?

Four things show up in the evidence. The first is fairness, meaning the outcome, the procedure and the treatment all count. A refund with a rude reply fails on one of the three (Tax et al., 1998).

The second is speed of news. Start by telling the customer before the customer asks. Delta promises news within 30 minutes of learning of a delay, a canceled flight or a diversion (Delta plan). JetBlue sets the same 30-minute window (JetBlue plan).

The third is ease. After the CrowdStrike incident of July 2024, Delta let customers cancel on its site or app and get an automatic refund for the unflown part of the trip. By its count, 70% of the refunds processed since July 19 were completed that way (Delta).

The fourth is front-line power. At the Ritz-Carlton, every employee, at all levels, can spend a set amount per guest per incident. The hotel says the full amount rarely gets used (Ritz-Carlton Leadership Center). A rule that the front line cannot act on means a slow reply.

Some of this is now public. The US Department of Transportation runs a dashboard of the commitments airlines make when a delay or a canceled flight is their own doing (DOT). It lets travelers compare airlines side by side.

How do loyalty programs use service recovery?

Miles are a ready way to say sorry. Delta used them after the July 2024 CrowdStrike incident. Its email offered affected customers a choice. They could take SkyMiles or a credit voucher (Delta).

The CEO named meals, hotels and rides as well, paid by voucher or reimbursement. He also called the miles and vouchers a further gesture of apology (Delta CEO update).

Complaints show up in the ACSI too. For airlines in 2026, fewer business travelers complained: 33%, against 38% a year earlier. Their score rose from 73 to 77 (ACSI).

Leisure travelers followed. Their complaint share slipped from 17% to 16%, and their score rose from 64 to 72. In car rentals, the ACSI says complaint handling has not improved much. It calls the touchpoint potentially valuable but underused.

Treat repeat failures as a separate group. Since the paradox faded after a second failure, a member who was failed twice needs a fix first, and a gift second.

How do you price an apology in points?

For illustration, a program credits 1,000 points to each of 4,000 members hit by a partner outage. Assume a cost of 1 cent a point. Each apology costs $10, so the offer costs $40,000. A retained member is worth $120 a year in margin. The offer breaks even if it keeps 333 members who would have left, or about 8 of every 100. Run it against a holdout group to see the real lift. Our entry on cost per point covers the math.

Scale the credit to the member’s value and to how bad the failure was. Give the front line a fixed limit, like the Ritz-Carlton. Review the limit each quarter.

Track recovery as its own segment. Flag every member who had a failure and a recovery. Then watch their spend for 90 days against members who had no failure. Split the group by the credit they received. The gap shows what the apology bought.

How GRAVTY handles service recovery

GRAVTY’s airline page describes a loop. It senses, reasons, recommends and acts. A move is proposed with its economics, then runs inside policy (airline loyalty). A goodwill credit fits that loop. A rule fires on a defined failure, and the credit follows the program’s own limits.

Patented Visual Rules let a business team change program logic with no IT ticket (GRAVTY technology). The airline page also lists agents for Trust and Risk. They detect and auto-heal anomalies.

Frequently asked questions

Where does the idea of service recovery come from?

It reached managers through a 1990 Harvard Business Review article, The Profitable Art of Service Recovery, by Christopher Hart, James Heskett and W. Earl Sasser Jr. (HBR, 1990). Academic tests of its central claim followed through the 1990s and 2000s.

Does compensation alone fix a complaint?

No. Customers judge the outcome they get, the procedure used and the way they were treated (Tax et al., 1998). Compensation covers only the first. A fast, fair process and a respectful reply cover the other two.

Does a second failure hurt more than the first?

A second failure makes recovery work less well. In a 20-month field study, a good recovery produced a paradox after one failure. It did not after two. Two bad recoveries in a row caused “double deviations” (Maxham and Netemeyer, 2002).

Who should be allowed to fix a problem?

The person who meets the customer. At the Ritz-Carlton, every employee at all levels can spend a set amount per guest per incident (Ritz-Carlton Leadership Center). Set a limit, then let staff act inside it.

Should an apology be paid in points or cash?

Offer a choice if you can. Delta’s apology email gave affected customers SkyMiles or an electronic transportation credit voucher (Delta). Points cost the program their cost per point, not their face value, so price both options before you send.