Who owns client data: the agency, the client or the vendor.

There is one question that settles this, and it is not in most contracts. If you cancelled every tool tomorrow, what would you still have?

If the answer is a set of tables in an account you control, you own your data. If the answer is an export you would have to request, or a dashboard that stops rendering, you were renting access to it. Both arrangements are common and only one of them is ownership.

An agency has to ask it twice: once about the vendors it pays, and once from the client's side. If the client left tomorrow, what would they still have?

Most tools fall into one of these, and the difference is worth knowing before you sign rather than after.

• The vendor holds it

  Data is collected onto the vendor's platform and served to you from there. Convenient, and your history lives inside their system. Ending the relationship means an export, in whatever shape they provide, if they still hold the full history.

• The vendor sends you a copy

  Data is collected by the vendor and forwarded to storage you control. Better, and the usual middle option. Check whether the destination is on your plan or an upgrade, and whether the copy includes the full history or only what arrived after you set it up.

• It lands in your account first

  The vendor writes into storage you already own and works inside it. Ending the relationship leaves every row where it is, because it was never anywhere else.

Under the vendor question sits a second one, and it is the one clients ask. The ad accounts and analytics properties should normally belong to the client, with the agency granted access. An agency holding a client's ad account in its own name has a dispute waiting for the day the relationship ends.

The reporting history is less clear-cut. The agency built it, from the client's accounts, in storage the agency pays for. Decide at onboarding what the client receives if they leave, and write it into the contract. Then make sure the data is shaped so you can deliver it: when every row carries the client it belongs to, one client's history is a query rather than a project.

Five, and they are all answerable in a sentence. A vendor who needs a call to answer them is telling you something.

Where does the data physically land, and in whose account. Can I choose the region. Is the destination included on my plan or is it an upgrade. If I cancel, what exists the next day and for how long. Can I see how a given number was calculated, or is that logic inside your product.

Two reasons, and neither is about switching vendors.

The first is history. Marketing data gets more useful with age, because comparing this year to last year needs last year. Data you do not hold is data you cannot compare against once the arrangement ends.

The second is AI. Anything you point a model at has to be somewhere the model can reach, under access rules you set. Data sitting inside a vendor's platform, reachable only through their interface, is not available to your own tooling on your own terms.

SightShark writes into a warehouse you already own, in a region you choose, from the first sync. It is not something on a higher plan.

The modelling runs in SQL inside your own account, so the logic behind every number is readable by you and does not leave when we do. Stop paying and every row is still there, in the account it was always in. Every row also carries the client it belongs to, so a departing client's history can be separated from the rest of the roster.