Closing The Operating-Model Gap You Left Open · The Reluctant Guide to Shopify Migrations

Three months after Greycott went live, Owen Caldwell walked past the marketing team’s bank of desks on a Wednesday morning, the way he had walked past it every Wednesday morning for nine years, and something was wrong. It took him most of the corridor to work out what. They were launching a campaign—he could see the proofs on someone’s second monitor, a spring gifting promotion, olive oil and the good vinegar bundled into a box that photographed well—and they were launching it today, and nobody had told him. Nobody had filed a ticket. Nobody had asked engineering to schedule a deploy. Nobody had booked the forty-minute meeting where the launch date got renegotiated against the development queue, the meeting that used to be load-bearing, the meeting that used to be how he knew a campaign was happening at all. He stopped. He nearly leaned in to ask whether everything was okay, whether they needed anything, whether someone should be looping him in. Then he understood that the absence of all that machinery was not a problem to be solved. It was the thing they had spent a year and a great deal of money trying to build. He walked on.

That is what done looks like, and it is so quiet that the person it happens to almost mistakes it for something going wrong.

We told you at the very start of this book—back when you had every reason to think you were buying a guide to a software project—that the real subject was never the software. The platform was always the easy half. We can say that now without it sounding like a provocation, because you have been through the hard half with us: the discovery that wouldn’t sit still, the scope that grew teeth, the staging store that lied, the four weeks after launch when the bug list climbed before it fell. All of that was the technology migration, and the technology migration is the part that has a launch date.

The other migration has no launch date, which is exactly why it is the one almost everyone underestimates. The technology migration gets you onto Shopify. The operating-model migration is what determines how much of Shopify you actually use. They are not the same project, they do not finish on the same day, and the second one finishes so softly that you can miss the moment entirely—which is the whole reason this chapter exists, and the reason the book ends here rather than at cutover. The platform went live in an earlier chapter. The migration finishes here, on a Wednesday, in a corridor, when a campaign ships and nobody schedules a meeting about it.

The shape of it, if you want a shape, is the thing every post-launch deck on earth promises and almost none deliver honestly: thirty, sixty, ninety days. The deck version is about feature adoption and training completion percentages. The honest version is about what the organization can do without asking permission.

Day thirty is training has happened. Somebody ran sessions, somebody recorded them, somebody built a wiki page that will be out of date by day forty-five. The content team has technically been shown the theme editor. The operations team has technically been walked through Flow. On day thirty, everyone can do the new thing if a person who already knows how is sitting next to them. This feels like progress, and it is, but it is the cheap kind, the kind you can buy with a calendar invite.

Day sixty is people are using it without help. This is the part that cannot be bought with a calendar invite, because it isn’t knowledge, it’s nerve. The content lead changes a hero image at four in the afternoon without sending a single Slack message first, because she has done it eleven times now and nothing broke any of the eleven times, and the twelfth no longer feels like skydiving. The reflex that took years to install—file a ticket, wait for prioritization, wait for QA, plan the campaign around however long all of that takes—has finally started to dissolve, not because anyone gave a speech about empowerment, but because the tools were right there and the building didn’t burn down.

A lone chair under a lamp in a room where the dust outlines of a table and many surrounding chairs are still visible on the floor, the furniture itself long gone. — The Thursday meeting. Still on the calendar. No longer on the floor.

Day ninety is the one nobody predicts. Day ninety is people are using it better than the agency taught them. The content team has found a section configuration you never showed them, because they had a problem on a Tuesday and the platform let them solve it on the same Tuesday, and now they have a way of building landing pages that is genuinely theirs. This is the moment the operating model has actually migrated, and you can tell because the knowledge has stopped flowing outward from the people who built the thing and started generating inside the people who run it. You taught them to read. They have started to write.

(Not every team gets to day ninety. Some teams get to day sixty and decide that sixty is comfortable, that self-service is fine for hero images but real changes still want a ticket, and they settle there, running a platform built for distributed ownership as though it were the centralized one they left—leaving, as we warned you somewhere back near the front of the book, between a third and half of the thing’s practical value sitting unused on the table. Day ninety is not automatic. It is sponsored, or it doesn’t happen.)

Then there is the queue. You have been so good about the queue.

All through this book we have been telling you no on its behalf—no to the wishlist that wanted to be a loyalty program, no to fixing the taxonomy on launch day, no to the while we’re at it that quietly turns one migration into two. The deferred-changes backlog has been sitting there the entire time, patient, timestamped, accumulating the small reasonable wants of every team who was told not yet, park it, after launch. After launch is now.

And here is the genuinely lovely thing, the reward for all that discipline: you can finally touch it. Slowly. One change at a time. The reason you can touch it now and couldn’t before is not that you have more time—you have less, if anything—but that the ground has stopped moving. During the migration, every change you made was a change to a system that was itself in flight, which meant you could never cleanly tell whether a number moved because of your change or because of the eleven other things shifting underneath it. That is over. The system is stable enough now to read what each change actually does, which means each change is, for the first time, reversible—you can ship it, watch the one number it was supposed to move, and roll it back on Thursday if the number went the wrong way, knowing it was your change that did it and not the tide. That reversibility is the entire prize you bought with a year of saying not yet. Spend it the way you earned it: deliberately, one deliberate thing after another, not by emptying the whole backlog into production the first calm week because you finally can. Emptying it all at once is just the old while we’re at it wearing a party hat.

A glass jar crowded with folded paper slips sits on a sunny windowsill; one slip has been unfolded and set beside it while the rest remain inside. — After launch is now. One at a time.

An Inventory Of Things You No Longer Need.

It is traditional, at the end of a long project, to take stock of what you built. We would like to do the opposite.

Here is a partial inventory of the things that, somewhere between day thirty and day ninety, quietly stopped existing—not decommissioned in a meeting, not migrated, not sunset with a memo, just one day no longer there, the way a habit you’ve broken stops being a habit you’re breaking:

The standing Thursday meeting where marketing and engineering negotiated the campaign calendar against the deploy schedule. Nobody cancelled it. It just stopped having an agenda, and then it stopped having attendees, and the recurring invite is still on the calendar like a tooth that fell out painlessly in the night.
The shared document titled Content Change Requests (Q+)—a spreadsheet, of course it was a spreadsheet—into which the content team used to type the changes they wanted and could not make themselves. Its last row is dated nine weeks ago.
The Slack channel where you asked an engineer to please change a button color, and the small private shame of having to ask.
The role, never written down anywhere but understood by everyone, of the person who knows how to deploy on a Friday. That person is still employed and still valuable. He is simply no longer a single point of failure for whether the homepage can change before the weekend.
The phrase “we’ll need a developer for that,” which used to be the true answer to roughly half the questions the business asked, and is now the true answer to a much smaller and much more interesting half.

None of these had a funeral. That is the point. The operating-model migration does its most important work by subtraction, and subtraction never announces itself. You will know it happened the way Owen knew it: by walking past where something used to be and finding the space empty and calm.

The agency, somewhere in here, leaves. Or rather—the agency changes shape, because the version where it simply evaporates the week after launch and the version where it stays forever are both failures, and we spent the previous chapter telling you why. By the fourth month the handoff should be substantially complete: the in-house team has taken the wheel on progressively larger pieces while the people who built the thing were still in the room to catch the swerve, and what remains is a maintenance retainer, a relationship rather than a dependency. The brand owns operations. The agency is on call for the genuinely differentiated work—the bespoke build that is actually worth building—and is no longer the answer to who changes the button. That is the correct end state, and it is correct precisely because it is undramatic. The good handoff feels like nothing happening. The people who built the thing get quieter and quieter, and then one week you notice they have not been needed, and that week is the deliverable.

You can watch this same transition from the outside, in the brands whose stories we’ve told you by name. Tannico’s marketing team came off a Magento setup where every content change meant touching code and waiting in a queue, and on the other side they were launching campaigns on their own—the visible metric was campaign velocity, the number that goes in the case study, and it is a real number and a good one. But the invisible metric is the better one, and it is this: nobody at Tannico had to schedule a meeting about a button color anymore. The velocity is what the operating-model migration produced. The absence of the meeting is what it actually was. MeUndies tells the same story from a different organ of the business. The subscription processing runs at a success rate north of ninety-nine percent, which is the technical outcome and the one an engineer is proud of; the operational outcome is that the customer service team stopped alt-tabbing between two systems all day, because the subscription logic was surfaced where they already lived—inside the support console, where, as an earlier chapter argued, the black box finally got dragged into the room. The resilient subscription is the technology migration. The support agent who no longer keeps two tabs open is the operating model.

Which brings us, gently, to Priya Shah, and to her spreadsheets.

When this book opened, Greycott’s Director of Operations was running her corner of an eight-figure business on seventeen spreadsheets. Wholesale invoicing in one, retail forecasting in another, a QuickBooks reconciliation that existed because no system in the building agreed with any other system about what had actually been sold. Seventeen workbooks, each one a small monument to a thing the platform couldn’t do and a person therefore had to. She had been doing it for nine years. She was very, very good at it. Nobody ever told her that the migration would obsolete most of it, because she didn’t need to be told; she could read an architecture diagram as well as anyone and she had counted the spreadsheets that were about to lose their reason to exist.

Three months post-launch, she is down to two.

It would make a tidier story if the number were zero, and there is a kind of migration salesman who would promise you zero. But two is the honest and better ending, because the two that survived are not the two nobody got around to automating. They are the two that shouldn’t be automated—the judgment calls, the wholesale exceptions for the three accounts that have earned a phone call instead of a workflow, the forecasting hunch that lives in nine years of knowing which Charleston restaurants over-order at Christmas. Fifteen spreadsheets were doing the platform’s job badly because the platform wasn’t there yet. Two spreadsheets are doing Priya’s job well, and Priya’s job was never the thing the software was supposed to replace. They are, now, the two she wants to keep. The migration didn’t take her work away. It took away the fifteen things standing between her and her work.

Priya at a desk with fifteen bulky binders shoved to the far edge and only two slim open notebooks in front of her, her hands resting on them in quiet ownership. — Fifteen spreadsheets were doing the platform's job badly. Two are doing hers well.

And Andrew Whitfield, finally, who has been asking questions this entire book.

He is the CFO. It is his job to ask, and he has asked continuously and well—about the timeline, about the burn, about the second invoice that was larger than the first, about whether the thing the agency promised in the room was the thing the agency would deliver once push came to shove. For a year he got point-in-time answers, each one true on the day it was given and slightly stale by the next board meeting, which is the native condition of being a CFO during a migration. He was never the antagonist. He was the organization’s anxiety, given an office and a reasonable salary, and the anxiety was correct to exist.

The way you know the migration is over is not a dashboard. It is that Andrew stops asking about it.

Not because he gave up, and not because someone finally produced the report that satisfied him—there is no such report—but because the question quietly stopped being interesting. Three months in, in a hallway, he asks Owen something, and Owen braces for how is the migration going, the question he has been answering in some form for fourteen months. That is not the question. The question is whether the new self-serve campaign capability means they could finally test the wholesale storefront idea the founder has been muttering about since before any of this started. Owen does not have an answer. Owen is delighted not to have an answer, because it is a question about the future of the business instead of a question about the cost of the past, and somewhere between those two questions, without anyone declaring it, the migration ended.

So here, at the close, is the whole thing said once more, as plainly as it deserves.

Launch was the beginning. You know that now, three months and one chapter past it; you have lived the part where the platform going live turned out to be the easy half, and the part where the organization slowly, quietly learned to operate differently turned out to be the half that decided whether any of it was worth doing. The brands that get this right are not the ones that swapped platforms most cleanly. They are the ones that migrated their assumptions—about who is allowed to change the homepage, about whether a button color is an engineering problem, about how many of a person’s spreadsheets are load-bearing and how many are just scar tissue—and the operating model finally clicks for them because the assumptions underneath it moved first. You migrated the code in an earlier act. You migrated the assumptions in this one. That second migration is the one that doesn’t show up in the cutover runbook, doesn’t have a go-live date, and doesn’t end with a celebration, because the celebration is the absence of something—a meeting that no longer convenes, a ticket no one files, a question the CFO no longer needs to ask.

The organizations that treat all of this as an afterthought tend to finish their migrations exactly on time, and then spend the following year slowly discovering how much value they left behind. You did not do that. You treated it as the real project, because it was, and the reward is the quietest possible thing: an ordinary Wednesday, a campaign that ships without you, and the strange, earned, slightly disorienting feeling of walking past where a problem used to be and finding nothing there at all.

That is the end of the migration. It is not, as we keep promising you, the end of the book.