A signed agreement is not an operating business. Between the two sits a period in which nothing is yet running and everything is being made ready, and that period is where this kind of work most often dies quietly. It rarely dies of a bad decision. It dies of the distance between what the client expected the early weeks to look like and what those weeks actually contain.
The period starts with access, and access starts with the system asking for it. The onboarding is run by the operation rather than handed to the client as a form to fill in. It requests the accounts it needs, connects them, and produces its first output, which is not a campaign but an inventory: what exists, what is reachable, and who actually owns each thing rather than who happens to hold a password to it.
Then the existing digital side is mapped against what the business sells. Not against a checklist of good practice, and not against whatever a competitor appears to be doing. Against the actual products, the actual margins, and the actual route the money takes to arrive. A store that converts well on a product line the company would rather stop making is not a strength. A tracking layer reporting in fine detail on a channel that carries a small share of the revenue is decoration.
What comes back from that map is a list of gaps with an order to it. The order is not aesthetic. It follows what has to exist before anything else can run. There is no point warming a sending domain for a store that is about to be replaced, and no point directing media spend at a page that cannot be measured.
The gaps become the build list, and the build list becomes the contract's deliverables, written item by item with a fee attached to each. That structure exists for one reason. If a deliverable in the contract does not happen, the fee for that deliverable is refunded. Not carried into next month, not offset against something else, refunded. A refund clause only means anything if the thing being refunded is small enough and named enough to be identified, which is why the build appears in the agreement as items rather than as a phase.
Execution starts on whatever is ready instead of waiting for the whole list. If the email infrastructure is sound before the store rebuild finishes, sending starts. This is not partial scope. Partial scope is a permanent arrangement in which some functions stay outside the operation, and it does not work, because the functions feed each other and taking over the media without the creative trades control for speed. Sequencing inside a scope that is already whole is a different thing entirely.
That is why the early weeks look like construction rather than operation. Building comes first because nothing can be run until it exists. A rebuilt tracking layer produces no headline in its first week. It produces the possibility of a headline that is true. The same holds for a brand foundation, a store, a sending domain.
This is also the point at which an honest system is at its least reassuring. The reported number is revenue divided by every pound of spend the operation directs, and in the early weeks the attribution chain is usually incomplete and the window is too short to say anything. The rule is published in advance: where data is missing, the chain is broken, or the period is too short, the number is not shown and the reason is written in its place. A client who wanted a figure in week two gets an empty slot with an explanation attached.
A less honest supplier gives that client a number. It will be a flattering one, because a number built on two weeks of partial data can be made to say almost anything, and the client will feel better for having seen it. The feeling is the product in that arrangement. Six months later there is no way to tell whether anything worked, because the baseline was invented rather than measured.
What the client has to do in this period is small, but it is not zero, and pretending otherwise is how the period slips. The first thing is access, granted from the owner's own authority rather than delegated down to whoever happens to hold the password, because delegated access arrives late and arrives incomplete.
The second is a set of answers that exist nowhere except in the owner's head. Which products carry margin and which are kept for other reasons. Which customers the business does not want more of. What has been promised to a retailer, a distributor or a landlord that constrains what can be said publicly. What cannot be claimed for legal reasons. None of this is in the order data, and a takeover that guesses at it will build the wrong thing competently.
The third is one pass on the decisions that belong to the company rather than to the operation: the name, the register of the language, what the business will and will not say about itself. One pass, early, in writing. Left open, it turns into an argument in month three about work that has already shipped.
The fourth is telling the previous supplier. Two parties publishing into the same accounts is worse than either of them alone, and the handover date belongs in a message from the owner rather than in an inference drawn by an agency that has noticed its access change.
Beyond those, the client operates nothing and configures nothing. Access is the only continuing input. The rest is construction, and the honest description of construction is that it looks like very little from the outside and is the reason everything after it is possible.
Book the call, and bring the answers that live nowhere but in your own head, because that is the only part of this period the business itself has to carry.