Trust leakage: how small customer failures add up

What the term is for

A single awkward moment is not automatically a loss of trust. A late van, a wrong time on a listing, or a reply that answers the wrong question can each be explained away. Leakage is the cumulative reading a customer makes when those moments keep arriving, and when the business's own measures still say that each step went fine.

The neighbouring ideas are older than the label. What customer effort means after someone has already made contact is the work pushed back onto the customer: waiting, repeating, being passed on, trying again. A service failure, in ordinary language, is when the service does not match what the customer reasonably thought would happen. Double deviation is the sharper case, when the attempt to put a failure right becomes a second failure. Trust asymmetry is the observation that damaging evidence can be harder to undo than it was to cause. Repair is what a business does after confidence has already been dented, and it is a separate event from the original mistake.

Those are established lines of thinking. "Trust leakage" is only a way of holding them in one customer journey, including the losses that disappear when each team looks only at its own number.

A second miss changes the story

The first gap can still have a harmless explanation. The phone was busy. The email landed in the wrong place. The hours on the website had not been updated after a holiday. A customer can believe that once.

A second related miss makes the harmless explanation less comfortable. They were passed to someone who did not have the first conversation. The price they were told is not the price on the invoice. The "we'll come back to you tomorrow" never arrives. Why making customers chase you damages trust is that sequence looked at on its own. How inconsistent business information is one form of the same leak is the version that happens before anyone has even spoken to you: the public record disagrees with itself.

The recovery can then add a further miss. In 2016, Kenny Basso and Cristiane Pizzutti studied what they and other service researchers call a double deviation: a service failure followed by a recovery that also fails. In those experiments, the failed recovery intensified the damage to trust. The studies are experimental and sit in service settings, not in a panel of UK small firms, so they do not produce a rule for how much to refund or which words to use. They do support a narrower point. A second failure during the cleanup is not just another item on a list. It tells the customer that the business cannot sort out its own mistake. Whether trust can be repaired after that depends on what the customer thinks the failure revealed.

Paul Slovic's 1993 discussion of trust, written about risk and public institutions rather than local shops, is the usual reference for a related asymmetry. Damaging events can be harder to repair than their size first suggests. That is not a sum in which one bad visit cancels a fixed number of good ones. Severity, the customer's history with you, and what you do next all still matter. The useful warning is against the assumption that nine smooth jobs will automatically erase one breach by arithmetic.

A promised next step sits in the middle of this. Why a broken promise is a different kind of failure from a vague disappointment, because the business set the standard itself. This article's point is what happens when several of these misses accumulate, not the anatomy of any one of them.

When every internal number still looks fine

It is possible for the visible business metrics to look healthy while this is going on. More people can find the firm. The website can be converting more visitors into enquiries. The first reply can be getting faster. A ticket can be marked closed because it was passed on. None of those numbers is false. Each one is looking at its own slice.

The customer, meanwhile, may be meeting a stale opening time, a price that appears late, a handoff that makes them retell the story, and a callback that was promised and missed. Which?'s 2024 research on UK energy and broadband is the concrete evidence that this kind of journey is not trivial for the person on the receiving end. It is not a study of small local firms, and the detail of that research sits with the customer-effort article. What belongs here is the shape: problems that feel small from inside an organisation, such as not getting through or being passed between people, added up to a large amount of customer time and to customers giving up.

There is no UK study in this evidence that follows a particular small firm from one missed callback through to a measured drop in revenue. The mechanism is better supported than any claim about how many pounds one extra chase costs.

What the evidence can say about cost

There is no accepted single cost of lost trust. The money and the lost custom, where they exist, travel along different routes, and the evidence for each route is different.

Customers can abandon a process they have already started. Customers can avoid the next purchase. They can switch, buy less often, or become less willing to rely on the firm for a larger job. They can tell other people, including in a review. The firm can also spend more time recovering work that a cleaner handoff would have finished. Those are recognisable business consequences. Only some of them have been measured, and not for a general small business.

Which? estimated two consumer harms in UK energy and broadband, from a Yonder survey of 4,101 adults in May 2024, scaled up to the population. One was 27.3 million hours of customer time. The other was £298 million of financial harm to customers, including people who gave up trying to resolve a problem and people who did not make contact because an earlier experience had been poor. Neither figure is revenue those companies lost. Neither is a tariff for a local trades business or a professional firm. They show that friction in two essential services had a size the customer could feel. They do not show what a missed quote costs a joiner in Middlesbrough.

A different kind of evidence comes from reputation, which is not the same thing as trust. Why a review score and a willingness to rely on you are different is set out separately. Michael Luca's study is still the clearest example here of reputation information moving actual demand. Using restaurant revenue from the Washington State Department of Revenue, he found that a one-star difference in the rating Yelp displayed was associated with about 5 to 9 percent higher revenue for independent restaurants in Seattle, over roughly 2003 to 2009. Chain restaurants did not show that effect. It is one market, one platform, and one period. It is not a UK small-business multiplier. It does show that displayed ratings changed revenue for independent restaurants in that setting. Other people may meet the story later, as a review or as something a neighbour repeats.

What it cannot say

It cannot say that every inconvenience reduces trust by a set percentage. A delay can annoy someone without changing their view of whether you are competent or honest. It cannot add the Which? hours to the Which? pounds, or add either of them to Luca's restaurant result, and call the total the cost of distrust. It cannot treat a fast internal process as proof that the customer was looked after.

What it can say is more limited, and more useful day to day. Look for the promise gap: what the customer was told, what they then had to do, and which of your own numbers still looked fine while that was happening. If the gap is a chase, a broken commitment, or a cleanup that created a second problem, those are the next places to look, not another average of "trust" in the abstract.

If you want that read against a particular local business, how I help starts from the part of the journey that is actually thin, whether that is before the enquiry or after it.

References and further reading

1. Basso and Pizzutti (2016), Trust Recovery Following a Double Deviation, Journal of Service Research. Experimental research on a service failure followed by a failed recovery. Not a UK small-business panel, and not a compensation formula.
2. Slovic (1993), Perceived Risk, Trust, and Democracy, Risk Analysis. Trust asymmetry in a risk and public-institutions context. It does not support a fixed ratio of good experiences to bad ones.
3. Which? (26 September 2024), energy and broadband customer service. UK survey of 4,101 adults. The £298 million is estimated consumer financial harm, and the 27.3 million hours are estimated customer time. Neither is company revenue.
4. Luca, Reviews, Reputation, and Revenue: The Case of Yelp.com, Harvard Business School Working Paper 12-016. Seattle independent restaurants, about 2003 to 2009. The 5 to 9 percent figure is not a UK small-business rule.