Poor customer experience, and where it actually comes from

What a poor customer experience actually is

A poor customer experience is any point in an e-commerce purchase where what the customer was promised and what they actually got come apart: on the product page, at checkout, and most often after checkout, in the parcel and the refund behind it. The standard list of poor experiences is a list of things a company decided: a slow site, a rude reply, a thirty-day returns window. The ones that cost the most aren't decisions at all. They're orders that broke after the money was taken, and they get filed as support tickets rather than as operational failures. EHP Labs hit that point after a major sale. The team spent days digging out of a backlog their own customers had flagged before the team did.

Poor customer experience and poor customer service are also not the same thing, and treating them as synonyms is why so much advice here ends in a support fix. Service is one interaction, in a channel, with a person. Experience is the whole journey, most of it unattended, which is what customer experience management misses after checkout.

The two kinds of poor experience

Almost every page on this topic puts all of it in one numbered list. It's two lists, and the split that matters isn't by channel or by journey stage. It's by who caused it.

• Chosen: a decision somebody made and can unmake: the eight-field checkout, the thirty-day returns window, the support hours. Predictable, visible in aggregate, fixable in a planning cycle.
• Happened: an order that broke after checkout: the split shipment, the carrier exception that never updated, the item picked wrong, the refund sitting unprocessed. One order at a time, invisible in aggregate, and nobody finds out until the customer says something.

The chosen kind gets roadmapped. The happened kind gets a ticket. One in five orders hits an operational break after checkout, and that whole share lands in the second column, where nobody owns it. That asymmetry is the expensive one, and shoppers price it in long before they meet either kind: among shoppers who abandoned for a reason other than simply browsing, Baymard Institute's synthesis of cart abandonment research finds 20% naming delivery being too slow and 13% an unsatisfactory returns policy.

Where the breaks come from

Independent measurement doesn't put the problem where the advice puts it. The American Customer Satisfaction Index scores online retail's post-purchase benchmarks no higher than any of its storefront ones: mobile app quality at 88 and site performance at 83, against quality of delivery at 83 and ease of returns at 80, with its 2026 shipping study reporting differences of 15 to 20 points between companies on those measures. Brands are badly placed to grade themselves here: Bain surveyed 362 companies and 80% believed they delivered a superior experience while 8% of their customers agreed. That was 2005, and it's still the number everyone reaches for.

Before the money is taken

• Costs that only appear at the last step, after the customer has already decided.
• Forced account creation in front of a first-time buyer.
• A product page that describes something other than what arrives.
• A checkout that breaks on the phone most of the traffic is using.

All four are real, worth fixing, and covered at length everywhere else.

After the money is taken

• The parcel stops moving in a carrier facility; the customer notices first.
• The order ships in two boxes and nobody says so; the customer counts them first.
• The wrong item, or a damaged one, leaves the warehouse; the customer opens it first.
• The address fails validation after the confirmation email goes out; the customer asks first.
• The refund is approved and never processed; the customer checks the statement first, on a returns leg that now carries 19.3% of online sales.
• The stated delivery date quietly passes; the customer counts the days first.

Six different breaks, one pattern. The customer found out before the brand did.

What a poor customer experience costs

Take the independent number first. ACSI's national index sat at 76.7 in the first quarter of 2026, down on the year, and the index reads the pattern as pent-up defection: customers who are unhappy and haven't moved yet. The channel meanwhile keeps growing, at 16.9% of US retail sales in that same quarter, and every point of that share is more orders in flight, so more of them breaking after checkout and landing on the same team.

• The order: the break is cheapest to fix on the day it happens and most expensive to ignore: a reship, a refund, a discount, or a customer who quietly doesn't come back.
• The customer: they weren't promised a fast apology. They were promised a delivery, and an apology doesn't deliver it.
• The team: every break nobody caught arrives later as a ticket, at the worst moment and at the highest price, which is the shape of what waiting for complaints costs.

Why the usual fix does not work

Every fix prescribed for a poor customer experience starts after the customer has already had one: reply faster, apologize better, empower the agent, route the ticket. That's recovery, and the category underneath it's why it caps out. A helpdesk, built two decades ago to run customer service over email, is a system for replying about problems, not for resolving them, and no amount of tuning turns a reply into a delivered parcel. Fixing the order is a different job, and it belongs to a different function: proactive e-commerce operations.

The industry consoles itself with the service recovery paradox, the idea that a failure handled well leaves a customer more loyal than no failure at all. The meta-analysis is narrower than the story: recovery lifts satisfaction, and doesn't measurably move repurchase intention, word of mouth or corporate image. Handled well, a customer will say they're satisfied. It doesn't make them buy again. Harvard Business Review's study of more than 75,000 customers points the same way: loyalty tracks how well a company keeps its basic promises. The scores most teams watch grade the recovery rather than the failure, which is the trouble with customer satisfaction metrics.

The promise has a deadline

In the US the delivery promise is also a legal one. Under the FTC's Mail, Internet, or Telephone Order Merchandise Rule, 16 CFR Part 435, a seller must ship within the time it stated, or within 30 days of receiving a properly completed order where it stated none. On learning it cannot, it has to seek the buyer's consent to the delay, and without that consent it has to promptly refund. It has covered all internet and mobile orders since December 2014.

Read that as an operator. A late order isn't just a disappointment, it's a commitment with a clock on it, and a brand that hears about the delay from the customer has already run past the point where it could have asked. Every order is a promise, and the promise has a date on it.

What to fix first

Start with the happened column, because nobody owns it yet. The shape is three verbs, in order. Detect the break from the order data rather than from the inbox: the shipment with no carrier scan, the refund past its window, the delivery date that has gone by. Decide the right action for that one order, against what it was promised. Act before the customer writes in. That's the difference between proactive and reactive service, and why a healthy on-time delivery rate can sit above a queue full of late orders.

The edges, plainly. Keeyu doesn't fix your checkout, your product page, your mobile site, your returns policy or the tone of a reply. We're not a helpdesk, a chatbot, a carrier, an order management system, a returns portal or a reviews platform. We act on the post-purchase break, one column of the two and not the whole of it.

The poor customer experiences that cost the most are the ones nobody chose and nobody saw: the parcel that stopped moving, the refund that never processed, the delivery date that passed in silence. They're not in a survey, and by the time they reach the inbox it's too late to fix them well. They're in your order data. Every order is a promise, and Keeyu keeps the promise: we detect the break, decide what should happen and act on it, usually before the customer knows anything went wrong. See what the Keeyu platform does.

• American Customer Satisfaction Index. Online Retailers industry benchmarks, 2026. Quality of delivery 83 and ease of returns 80 against site performance 83, mobile app quality 88 and app reliability 87, across 22 major online retailers.
• American Customer Satisfaction Index. Retail and Consumer Shipping Study 2026. The two post-purchase benchmarks and the note that company-level scores differ by 15 to 20 points on them.
• American Customer Satisfaction Index. National ACSI, Q1 2026. The national index at 76.7, down on an annual basis, and the release's framing of pent-up defection.
• Baymard Institute. Cart abandonment rate research. Reasons for abandonment excluding browsing: delivery too slow 20%, unsatisfactory returns policy 13%. An independent research institute, synthesizing 50 studies.
• National Retail Federation and Happy Returns, a UPS company. 2025 Retail Returns Landscape. An estimated 19.3% of online sales returned in 2025. Produced with a commercial returns provider, disclosed here so the provenance is on the record.
• US Census Bureau. Quarterly Retail E-Commerce Sales, Q1 2026. E-commerce at 16.9% of total US retail sales, growing faster than total retail.
• Federal Trade Commission, via Cornell Legal Information Institute. 16 CFR Part 435, the Mail, Internet, or Telephone Order Merchandise Rule. The shipping window, the 30-day default, the obligation to seek consent to a delay and to refund promptly without it.
• Bain and Company. Closing the Delivery Gap, 2005. A survey of 362 companies: 80% believed they delivered a superior experience, 8% of their customers agreed. Cited with its age as part of the point.
• Dixon, Freeman and Toman. Stop Trying to Delight Your Customers, Harvard Business Review, 2010. A study of more than 75,000 customers finding loyalty tracks how well a company delivers on its basic promises.
• de Matos, Henrique and Vargas Rossi. Service Recovery Paradox: A Meta-Analysis, Journal of Service Research 10(1), 2007. The recovery effect is significant for satisfaction and non-significant for repurchase intention, word of mouth and corporate image.