Act I: The Trouble With Books About Replatforms
How To Read The Agency Proposal
Two proposals on Owen Caldwell’s desk at Greycott & Co. The first one is fourteen pages, full color, printed on a paper stock that suggests confidence. There is a Gantt chart on page six that resolves to weeks, with milestones colored in a four-step palette that progresses from a pale blue (kickoff) to a bottle green (launch). The dependencies are drawn with thin arrows. The arrows look correct. Owen does not know what makes a Gantt chart’s arrows correct, but he is fairly sure these are.
The second proposal is six pages. It does not have a Gantt chart. It asks for a paid discovery phase before committing to a timeline. The line item for the discovery phase is, by Owen’s reading of similar discovery phases at other agencies, expensive. The proposal is on regular paper. Whoever wrote it has used em-dashes in a way that suggests they are familiar with em-dashes.
Owen has been told, by the operations consultant who has been advising Greycott on this migration for three months, that he should pick whichever one feels right.
Owen has already decided. He doesn’t know it yet. He thinks he is still deciding.
The polished pitch is selling you confidence. The agency that asks uncomfortable questions is selling you reality. The first is a magician. The second is a doctor. The agency that makes you feel best during the sales process is often the worst fit for the months that follow it.
We did not invent this principle. Every Head of E-Commerce who has been through a replatform—been through in the way one is been through a divorce—will tell you the same thing, but only in the third drink. The second drink is when they tell you they wish they had done more research. The third drink is when they tell you which agency they should have picked.
The first drink is when they tell you the migration went fine.
Confidence is easy when you are applying a template.
The first agency on Owen’s desk—fourteen pages, full color, Gantt chart with the correct arrows—is applying a template. They have done this migration eight times. They have a tool that handles most of the data migration automatically. They have a preferred partner for search, a preferred partner for reviews, a preferred partner for the customer-account portal that Owen has not yet realized he is going to need. They have a project manager who will shield the team from stakeholder chaos. They have a methodology, and the methodology has a name, and the name is in the proposal.
The methodology is good. The methodology has worked. The methodology has, in fact, worked eight times.
The methodology has never worked on this migration, because this migration has never happened before. Yours hasn’t either. (Ask us how we know.)
This is the thing about templates: they tell you what to do until they don’t. They tell you what to do all the way up to the point at which they don’t, and they do not, in the moment, announce the transition. The transition arrives quietly, in the form of the operations consultant who turns out to have a strong opinion about wholesale invoicing, or the warehouse manager who mentions a label format in passing, or the CFO who reads the contract on a Sunday and notices that the data-migration line item does not specify what data. By the time the template stops working, you are eleven weeks in, and the agency that was applying the template is the agency that has to figure out what comes after the template, and they are figuring it out on your time.
The difference between the agency that has a template and the agency that has depth is hard to spot in a proposal and obvious six months in. The proposal will not tell you which one you are looking at. The proposal is, on this question, constitutionally unable to tell you. The proposal is a document produced during the sales process, which is the process specifically engineered to make this question hard to answer.
An agency with real depth surfaces complexity you hadn’t anticipated.
This is its own form of unpleasantness. The second proposal on Owen’s desk does it in the second paragraph. It says, in language that is almost but not quite apologetic, that it would like to look at the data structure before committing to a migration approach, because it has seen Magento installs with custom EAV models that do not map cleanly to Shopify’s product model, and it would like to know whether Greycott has any of those before estimating anything.
Owen does not know what an EAV model is. Owen knows that Daniel, Greycott’s lead engineer, has at various points in the last two months said the words attribute, catalog, and please. Owen suspects, correctly, that an EAV model is the kind of thing Daniel knows about, and the kind of thing the first agency’s automatic tool would silently mishandle, and the kind of thing Owen would discover, in month four, when somebody on the wholesale team asked why three hundred gift sets had lost their bundled-pricing logic.
The discomfort of the second proposal is that it makes Owen aware of a problem he didn’t have. Or rather: that he had it the whole time, and was previously enjoying the version of his life in which he didn’t know about it.
This is what the right agency is selling. It is not selling certainty. It is selling the willingness to find the things that will hurt you, first, before they hurt you.
That is a terrible sales pitch. (Ask us how we know.)
The right agency will make you slightly uncomfortable with how much they want to learn before committing to a plan. That discomfort is a sign they understand what they’re getting into. It is also a sign you are about to pay for a discovery phase, which is the kind of line item that gets a procurement director out of bed.
Now, governance.
If the proposal anticipates that the agency will disappear for months and only surface for status updates, the proposal is wrong. Not bad. Wrong. It is describing a kind of project that does not exist at this scale. Projects at this scale are not linear. You do not gather requirements upfront, lock down scope, and emerge six months later with a finished product. You discover requirements as you dig in. Stakeholders realize implications of decisions only when they see working prototypes. Business priorities shift mid-project, because they always do, because the world does not stop while you migrate.
The agency that wants to disappear for months and only surface for status updates is the agency that, six months in, will reappear with status updates and a change-order form. The change-order form will be the first time you find out that scope has shifted. The change-order form will not, on its own, tell you why.
The agencies that succeed at this level treat governance as iterative and transparent. Direct access between your team and the agency’s developers and strategists. Continuous communication, not just at milestones. Flexible planning that absorbs changes without derailing the timeline. Trade-off conversations conducted in actual sentences rather than in change-order forms.
A proposal that promises a single project manager who will handle communication is a proposal that, in practice, will end up handling communication by not communicating. The PM will surface the good news. The PM will manage the bad news privately. The PM is, in fairness, doing this because the PM has been told to do this, by the proposal you signed. The communication structure you sign for is not, in any practical sense, the communication structure you will get.
Head of E-Commerce Reading This At The End Of A Long Tuesday: the polished proposal is the one your CFO will prefer. We know. The polished proposal has a fixed price on it. The CFO likes a fixed price the way some people like a clean kitchen. We are not telling you to ignore your CFO. We are telling you that when the CFO asks why the second proposal is more expensive, the right answer is not because the second agency is worse at pricing. The right answer is because the second agency is being honest about how much we don’t know yet. This is a harder sentence to say in a steering committee than the first one. We know.
E-Commerce Director Considering The Other Agency: if you have ever inherited a Shopify Flow setup that nobody can untangle six months after launch, you have met the deliverable a templated agency leaves behind. The automations work. The automations probably even work well. They just do not make sense to anyone who did not build them, including, eventually, the agency that built them. There is a name for this. The name is technical debt that arrived in branded packaging. We have, by way of disclosure, delivered some.
Reader Whose Procurement Department Is Going To Hate The Second Proposal: yes. We know. Procurement is going to hate it. Procurement hates anything that does not have a fixed total. This is procurement’s job; do not be angry with procurement for doing it. The job of getting procurement comfortable with a paid discovery phase is, regrettably, your job. It is one of the many jobs that became your job the moment the board approved this migration.
Owen makes his decision on a Thursday afternoon, two weeks after the proposals arrive. He picks the second one. He cannot tell you, on the day, exactly why. He will tell you, eighteen months later, that it was the em-dashes.
Here lies the migration that came in at fixed price.
The price was correct. The migration was somewhere else.