The Quick Wins Companies Leave on the Table

A new customer finds the couch she wants in about ninety seconds. Good photos, clear price, free returns. She adds it to her cart, types in her address, and the checkout form can't find her street. Not a typo, just a newer development the lookup tool hasn't caught up with yet. She types it manually, gets a red error anyway, tries again, and finally gets through by leaving off the apartment number and adding it to the "delivery notes" field instead, hoping someone actually reads that.

At the last screen she remembers a 15% code from an email last week. It works, technically, but the total doesn't move: turns out it excludes anything already on sale, which the couch was, in small gray text only available in the terms and conditions of the website rather than on the coupon itself. She checks out anyway, a little annoyed, telling herself it's not worth the fight over forty dollars. Nine days later the tracking page says "delivered." It wasn't. She checks the porch, the side gate, asks a neighbor. Three days after that, a real person on the phone finally confirms it went to the wrong building entirely and is somewhere on its way back to the warehouse. She asks for a refund instead of a replacement, because at this point she just wants it over with.

The refund confirmation email lands in her spam folder, because nobody on the IT team ever got around to configuring DKIM, SPF, and DMARC correctly, or because the email has too many hyperlinks and/or spammy language so a good chunk of the company's transactional email quietly reads as suspicious to Gmail. She never sees it. As far as she knows, her refund is still pending three weeks later. She calls again, this time visibly irritated, and the agent on the phone has no idea the email even went out, because from the company's dashboard, that ticket looks resolved.

None of this, taken on its own, is a disaster. It's not a security breach, a product recall, or a PR crisis. It's five small, individually forgivable things stacked on top of each other in under two weeks. And she's not coming back.

The Project That Got the Budget

That same year, this company spent eight months and a meaningful chunk of its engineering roadmap building an AI-powered product recommendation engine. Leadership loved it. It got its own slide in the quarterly business review, complete with a demo showing how it would surface the right armchair to the right shopper at the right moment. It was, by most technical measures, a genuinely impressive piece of work.

It also launched to a churn rate that hadn't moved a single point, because the address lookup, the coupon fine print, the delivery tracking, and the spam-filtered emails were all still exactly as broken as they'd been the year before. Nobody had ever formally owned fixing them. Each one individually lived in a different team's backlog, filed somewhere below "nice to have," because none of them were dramatic enough on their own to compete for a roadmap slot against a project with "AI" in the name and a demo that made the room go quiet.

This pattern shows up everywhere, not just in e-commerce. It shows up when a hospital spends a year and a substantial capital budget rolling out a new patient portal while the phone line to reschedule an appointment still puts people on hold for twenty minutes, every time, with no callback option. It shows up when a bank redesigns its entire mobile app, complete with a new spending-insights dashboard, while a wire transfer can still silently fail without ever telling the customer why. The big, visible project photographs well in a board deck. The small, ugly annoyance sits in a backlog that nobody ever revisits, because fixing a broken coupon rule doesn't make anyone's promotion case, and nobody wants to be the person who spent a quarter on DNS records instead of "innovation."

Why the Small Stuff Isn't Actually Small

According to PwC's Future of Customer Experience research, 32% of customers say they'll stop doing business with a brand they genuinely loved after just one bad experience. Not a scandal. Not a data breach. One bad experience. That number is worth sitting with, because it should reframe how a company thinks about its backlog: the coupon fine print isn't a minor UX nitpick sitting quietly at the bottom of a Jira board. For roughly a third of the people who hit it, it's the entire relationship, decided in one moment they'll probably never even mention to anyone at the company.

And losing her isn't a cheap mistake to shrug off, either. Harvard Business Review has cited research putting the cost of acquiring a new customer at anywhere from five to twenty-five times more expensive than keeping an existing one, depending on the industry. Every customer this company loses to a fixable annoyance is a customer it now has to pay five to twenty-five times as much to replace, just to get back to where it already was. That's not a marketing problem. It's a quiet tax on the entire business, and it's invisible on any dashboard that only tracks new signups.

Checkout friction specifically has been studied to death, and the pattern holds up everywhere researchers look. Baymard Institute, which has spent years auditing live checkout flows across thousands of real e-commerce sites, has found that unexpected costs revealed at the last step, things like shipping fees, taxes, or coupon exclusions nobody warned the customer about, are the single biggest driver of abandoned carts, cited by roughly half of shoppers who leave without buying. Being forced to create an account before checking out drives away roughly another quarter of them. A checkout that simply feels too long or too complicated accounts for a further chunk on top of that. None of these are exotic, one-in-a-million failures. They're the digital equivalent of a store clerk who's rude exactly once, to exactly one customer, who then quietly never comes back and never says why.

The Math Nobody Runs

Here's the part that rarely gets calculated, and it's worth doing once with real numbers, because it explains how a company can be "mostly fine" at every individual step and still lose the customer anyway.

Say the address lookup works cleanly for 95% of shoppers. The coupon terms are clear and honest 95% of the time. Delivery arrives as promised, on the first attempt, for 95% of orders. And the confirmation and shipping emails actually land in the inbox 95% of the time which, given that fully authenticated domains see roughly 2.7 times better inbox placement than unauthenticated ones, is already a generous assumption for a company that's never configured DKIM, SPF, and DMARC correctly.

Multiply those four "pretty good" numbers together: 0.95 × 0.95 × 0.95 × 0.95. The odds of a single customer sailing through all four steps without hitting a single snag come out to about 81.5%. Add a fifth step at that same 95% rate, say, a support interaction that goes smoothly if something does go wrong, and it drops to roughly 77%. Add a sixth, a seventh, an eighth, as most real customer journeys have, and the number keeps falling well below what any single team would ever describe as a problem. That means for every five customers, more than one is going to hit friction somewhere along the way, even though every individual piece of the system is passing with what most teams would proudly call a solid grade.

This is the part that catches people off guard the first time they actually run the numbers: nothing in this journey is "broken" by any single team's own metrics. The checkout team hits their 95% and calls it a good quarter. The email team hits their 95% and moves on to planning the next campaign. Nobody owns the compounding, because nobody's dashboard shows the whole chain end to end; everyone's dashboard shows their own link in it, looking perfectly healthy. The customer, meanwhile, only ever experiences the whole chain. She doesn't grade each step separately and average them out. She either had a fine two weeks or she didn't, and she's already decided which one it was.

What Losing Her Actually Costs

"She probably won't come back" sounds abstract until it's translated into money. Say her couch was $900, and a reasonably loyal customer like her would have bought two more pieces of furniture over the next three years, worth another $1,400 combined. That's roughly $2,300 in customer lifetime value walking out the door over a broken address field, a misleading coupon, a delivery mix-up, and an email that never reached her.

Now the company has to go replace her. Using the low end of that HBR range, five times the cost of retention, if keeping her happy would have cost next to nothing (a DNS record, a clearer coupon disclaimer, a corrected delivery address), replacing her with a new customer of similar value costs real advertising and sales spend to land. Multiply that single story by the percentage of customers hitting similar friction across an entire order volume, and the eight-month recommendation engine starts looking like it was solving the wrong problem at a very real cost, not just an opportunity one.

What Actually Gets This Fixed

The fix isn't a bigger project. It's closer to the opposite: someone has to actually walk the full journey, start to finish, as a real customer would, on a real phone, with a real credit card, and write down every point of friction they hit, no matter how small or embarrassing it feels to log. Not a survey. Not a dashboard pulled from last quarter. An actual person, placing an actual order, noticing that the coupon math looked off and the confirmation email never showed up in their own inbox.

From there, the work is almost boring on purpose, and that's the point. Each friction point gets a name, an owner, and an honest size. The coupon disclosure issue is a copy and layout fix a designer can turn around in a day. The DKIM, SPF, and DMARC gap is a DNS configuration change an IT Admin can usually close in an afternoon, not a quarter, once someone actually assigns it to a person instead of leaving it in a shared "tech debt" list that never gets prioritized. The address lookup issue might genuinely need a new data provider, which is a real project with a real budget line, but it's one clearly scoped project instead of a vague, permanent backlog item nobody's ever accountable for. You might also consider a simple override for the customer to manually add their address to avoid such edge cases. Ranking these by effort against impact, and actually closing the cheap ones first instead of saving them for "someday," routinely fixes more of the real churn problem than the flashy initiative that took eight months and earned its own slide in the QBR.

This also isn't a one-time cleanup. New friction creeps back in constantly: a shipping carrier changes its API, a new coupon type gets launched without anyone checking how it interacts with the sale flag, a marketing platform gets swapped out and nobody re-verifies the DNS records on the new one. Treating this as a standing rhythm, someone genuinely walking the journey every quarter, not just once after a bad churn report, is what keeps the compounding math in the company's favor instead of quietly working against it.

None of this requires a reorg or a new platform. It requires treating "small and annoying" as a legitimate category of problem worth a name and an owner, instead of something that only gets attention once it's big enough to show up in a churn report with a dollar sign attached. By the time it shows up there, the company has usually already lost its share of the 32% PwC is describing, and it lost them quietly, one unremarkable Tuesday checkout at a time, while everyone's attention was on the demo that made the room go quiet.

As for the couch, she did eventually get her refund. She also bought her next two pieces of furniture somewhere else, from a company whose checkout, for whatever reason, just worked. Nobody at either company will ever see that decision on a dashboard. It happened quietly, the way almost all of them do.

And this same concept applies to employee experiences, where compounding small frictions lead to disengagement and, eventually, higher turnover. Read the full article here.

Frequently Asked Questions