Switch when the problem is structural rather than personal. That distinction sounds vague and is sharp in practice: it comes down to whether another agency would run into the same thing.
Four reasons that are structural
The team has changed three times. Then you buy your own learning curve again each time. With median staff turnover of 34.6 percent a year in this market that is not exceptional, but three changes on one account in a short period points at something other than chance.
You have to rebrief every proposal. If the first version is consistently off while your briefing is sound, they do not understand your market, and that will not change by itself.
The agency has outgrown your category. You have become their thirtieth client instead of their fifth. That is not a reproach, it is a fact with consequences for who sits on your account.
The assignment changed and they did not. You moved into e-commerce, or international, or a different kind of work, and that is not where this agency excels.
Three reasons that are not
One disappointing campaign. That happens at every agency, and the second attempt is usually better because they have learned something in the meantime that a new agency still has to learn.
A price discussion. That gets solved with a conversation about scope, not with a move costing you six months and the learning costs of a new team.
A new marketing lead who brings their own agency. Understandable and human, and not a reason. Measure first, switch after.
The order
Before you launch a pitch: name the problem, ask for a different team if that is where it sits, revise the agreements if the scope is wrong, and put a three-month deadline on it. If nothing changes, switching is the right decision and you can explain it. See my agency is not performing.
Measuring before you decide is the cheapest thing you can do here: an evaluation on both sides costs a fraction of a pitch and points at whether the problem is the team, the briefing or the agreements. IKAg does that with COLLAB, under Agency evaluation & remuneration.
And work out the cost first, because that number is usually missing from the decision. See what switching agencies costs.