The one-star review went up on a Thursday.

“Showed up for the 9am Saturday class. Locked door. Happened again three weeks later. I emailed. Nothing changed.”

The owner read it at the front desk and felt blindsided.

But the business wasn’t blindsided. It had been told — twice, in writing, months earlier. Both messages were sitting in the inbox: read, replied to, and archived.

This is the strange thing about feedback. Most businesses ask for it, receive it, and answer it, and still nothing inside the business ever changes because of it. The channel works. The process behind the channel doesn’t exist.

In the last post I modeled work flowing out of a business — one job, from “customer says yes” to money in the bank. This post is its mirror: the signal flowing back in. The technique is the same — a process model, which I covered in an earlier post on the general method — so we can go straight to tracing.

A government runs dozens of feedback channels at once — I inventoried that whole landscape separately. A small business usually runs two: an email address and a form on the website.

The process: from “a customer tells you something” to “something changes because they did.”

The business

Picture a family-owned fitness studio. An owner, a manager, and three part-time front desk staff who share an info@ inbox. The website is maintained by an outside contractor who bills by the hour, so website changes get saved up until there are enough to justify an invoice.

The Saturday morning class moved from 9am to 8am back in the spring. The schedule taped to the front door was updated the same week. The schedule on the website was not.

As before, the details are a composite, but every step comes from real organizations I’ve sat with. Swap “studio” for “clinic” or “permit counter” or “after-school program” and the shape holds.

Path one: the shared inbox

Here is what happened to the first customer’s message, traced day by day.

Day 0, Saturday. The customer arrives at 9am to a locked door, checks the website in the parking lot, and sees 9am confirmed. That afternoon she emails info@ and explains exactly what happened.

Day 2, Monday. A part-time staffer checks the inbox during a quiet stretch — the inbox gets read when the desk is slow, and weekends are never slow. He reads the message. He doesn’t know why the schedule changed and doesn’t have access to the website, so he stars the message for the manager.

Day 4, Wednesday. The manager works Wednesday through Sunday. She reads the starred message, winces, and replies with a sincere apology and a free class pass. She means to tell the website contractor. The message, now replied to, gets archived.

Day 5 onward. Nothing. The note to the contractor never gets sent — there is no step that would send it, only the manager’s memory, and the manager’s memory is running a front desk.

Day 41. A second customer emails about the same locked door. A different staffer reads it, apologizes, offers a pass, archives it. She has no idea a nearly identical message arrived six weeks earlier, because archived messages are invisible and nobody counts them.

Day 67. The first customer, burned twice, posts the review. The review cannot be archived.

Count it: one sentence of feedback took seven steps, crossed three sets of hands, and produced two apologies, two free passes, and zero changes. Notice what “handled” meant at every stage: replied to. The reply is real work, done with genuine care — and it is aimed entirely at the person, never at the cause. After the reply, the message goes to the archive, and the information in it dies there.

flowchart TD
    locked["Customer finds a locked door (day 0)"] --> email["Customer emails the shared inbox"]
    email -->|2 days waiting| read["Staffer reads it, stars it for the manager"]
    read -->|2 days waiting| reply["Manager replies: apology and a free pass"]
    reply --> archive["Message archived — counted as handled"]
    archive -.->|never| fix["Website schedule corrected"]
    archive -->|37 days waiting| second["Same complaint from a second customer"]
    second -->|26 days waiting| review["One-star review, public and permanent"]

Path two: the online form

The studio’s website also has a contact form.

Trace the second customer’s path, because she actually tried the form first.

Day 0. She taps “Contact” on her phone. The form wants her full name, email, phone number, a topic from a dropdown — General, Membership, Billing — and a message of no more than 500 characters. None of the topics fit, so she picks General, types her complaint, and hits send. No copy arrives in her own inbox. No reference number. Just a page that says “Thanks! We’ll be in touch.”

Where the submission goes. The form sends an email — subject line “New submission from Website Contact Form” — to the same info@ inbox.

Day 3. A staffer opens it, replies with the apology, archives it.

That’s the whole model:

flowchart TD
    locked2["Customer finds a locked door (day 0)"] --> form["Customer fills out the website contact form"]
    form --> submit["Submission emails the same shared inbox"]
    submit -->|3 days waiting| read2["Staffer reads it, replies with an apology"]
    read2 --> archive2["Message archived — counted as handled"]
    archive2 -.->|never| fix2["Website schedule corrected"]

Look at the two diagrams side by side. They are the same diagram. This form is the email path wearing a costume — with extra steps for the customer and nothing extra for the business. She filled out five fields, fought a character limit, and got less than an email would have given her — she doesn’t even have a copy of what she sent.

This is worth saying plainly, because forms get sold as the upgrade: a form that dumps into an unwatched inbox is not an upgrade. It is the same dead end with a worse doorway.

Where the paths actually differ

So is the form pointless? No — but its advantages are all potential, and none of them arrive automatically.

The honest comparison:

	Email / shared inbox	Online form
What arrives	Prose, readable one message at a time	Structured answers you can count and sort
Where it goes	Whoever happens to check	Can route by topic — if a route is built
Spotting repeats	Someone’s memory across archived threads	A column you can sort — the same complaint, eleven rows tall
What you can measure	Nothing until a message arrives	How many people started, finished, or gave up partway
Effort for the customer	Low — one sentence is fine	Higher — fields, dropdowns, limits
What gets lost	Little — people say it their own way	Anything the form didn’t think to ask about

A form produces structure: you can count submissions, sort them by topic, and see that the same complaint has appeared eleven times — something no one can see across eleven archived email threads. A form can route: a “class schedule” topic can go straight to the person who can fix schedules. And a form can be measured before anything arrives: you can know that forty people opened it this month and thirty-one quit at the phone number field. An email channel is invisible until the moment a message lands.

But the form pays for that structure in ways worth naming. It only collects what it asks about, so the problem you didn’t predict has no field. It asks for effort — and every required field turns away some person who just wanted to tell you one sentence. It can feel bureaucratic exactly when someone is already annoyed. And a badly designed form — wrong topics, tight limits, a demand for a phone number — collects almost nothing useful while suggesting you’d rather not hear it.

The channels genuinely differ. And yet in this studio, both ended in the same archive, because the difference that matters isn’t between email and form. It’s between having a process behind the channel and not having one.

What the models reveal

1. Feedback is answered one message at a time and never added up. Each complaint got a reply and a free pass — call it ten minutes and fifteen dollars, per message, forever. The cause was a one-line edit to a website. When nothing aggregates the messages, the business pays the per-message cost eleven times and the fix-it-once cost never, because no single message ever looks important enough to chase upstream. The archive isn’t storage; it’s where patterns go to be unfindable.

2. Nothing closes the loop back to the customer. The first customer did everything right — she reported the problem clearly, in writing, to the address the business gave her. From where she stands, the message changed nothing, so why send another? People stop telling you things when telling you visibly doesn’t work, and then the quiet inbox reads as satisfaction. The feedback doesn’t stop existing; it moves to channels you don’t control. The review was not her first attempt to help this business. It was her third.

3. Nobody owns the channel, so “handled” means “replied to.” Three staff and a manager all touch the inbox, and each one reasonably assumes that someone else follows up on causes. No one is wrong and no one is responsible. Without an owner, the definition of done quietly collapses to the reply — because the reply is the only step any one person can complete alone. “We answer every message within three days” can be completely true while the business changes nothing for a year.

Should be

As with the last post: redesign nothing, buy nothing. Just add rules where there is currently only habit.

One named person owns the feedback channel each week — not to answer everything, but to be the one who follows up on causes. Every message gets a topic scribbled on a running tally — paper is fine — so repeats become visible the second time, not the eleventh. “Handled” gets redefined: replied and either something changed, or the reason nothing will change is written down. Fifteen minutes a week, the tally gets read aloud: what came in, what repeated, what we changed. And when something does change because a customer spoke up, that customer hears about it — “you told us, we fixed it, thank you” — which is the cheapest customer loyalty available anywhere.

Same people, same inbox, same form. The locked-door problem dies at the second message instead of surviving to the review.

Could be

Then there’s what today’s tools make possible, honestly scoped.

A well-built form can route by topic, so a schedule complaint reaches the person who can fix schedules while the customer is still in the parking lot. Submissions can land in one shared list instead of an inbox — visible to everyone, sortable, countable, with a status beside each one that distinguishes “replied” from “resolved.” The customer can get a copy and a reference, so telling you something feels like it went somewhere.

AI has a place here, and it is narrow and unglamorous. It can group similar messages so themes surface instead of hiding in prose. It can suggest the topic and the route. It can draft the first reply for a person to correct rather than compose. It can raise a hand when the same issue crosses some threshold — this is the eleventh locked-door message this quarter.

Notice what it is not doing: it does not decide what matters, it does not apologize on your behalf, and it does not know which complaint is the quiet early warning about your business. Grouping and counting, yes. Judgment, no.

This problem is one I keep working at directly. I spent years building public feedback forms for government, and I’m now building Touchpoints at Civic Studio, where every feedback form is attached to a model of the process it belongs to — so a submission arrives already knowing which service, which step, and which owner it concerns. That design choice is this entire post in software form.

The real point

Feedback is only worth collecting if it can change something. And you cannot tell whether it changes anything until you draw the path — because “we always respond” and “nothing ever changes” turn out to be perfectly compatible.

The channel is not the decision. Email versus form is the small question. The big question is what happens on day 4, after the reply — and for most organizations the honest answer is a dotted line to nowhere.

Draw your path. Take the last three pieces of feedback you received and trace each one, day by day, to the thing that changed because of it. If the trace ends at an archive folder, you now know something your customers figured out months ago.

If you’d like help modeling how feedback moves through your business or organization — with your team in the room — get in touch or look at process modeling. A single coaching session ($195) is enough to draw one path and find the first thing worth fixing.

A Process Model of Customer Feedback