The bravest scoping decision I saw this quarter was about what NOT to migrate.
A client acquired a company and needed to fold the acquired marketing database into their own instance. The instinct in every migration like this is completeness: bring everything, because someone, someday, might ask for it. Instead, they decided to migrate only this year's contacts from two key events. No historical attribution. No decade of dormant leads. Every asset gets reviewed for relevance before it moves, because the goal is a clean, integrated brand by next year, not a bigger database.
I think this is the right call far more often than it's made, and I want to defend it properly.
The completeness instinct feels like risk management. It's risk transfer. Every record you migrate without a reason becomes someone else's problem later: the rep who calls a contact that went cold in 2019, the marketer whose deliverability tanks because a third of the list hasn't opened anything in years, the ops person who can't tell which lifecycle stages mean anything. You didn't eliminate the cleanup by migrating everything. You just paid to move it first, and you'll pay again to clean it.
There's precedent inside this same client, and it cuts the other way, which makes it useful. In an earlier acquisition, they migrated only accounts and opportunities, purely for finance. That was too narrow; marketing lost context it later wanted. So the answer isn't "migrate as little as possible" either. It's that scope should be a decision per asset class, made against a stated goal, not a default in either direction.
The filter I now push clients toward: if this didn't exist, would you rebuild it today? Apply it to everything. Nurture flows, forms, lists, historical campaign data, dormant contacts. If you wouldn't rebuild it, don't pay to move it. Archive the old system read-only for the one time in three years someone needs to check a 2018 campaign stat, and move on.
The honest counterargument is compliance and attribution continuity. Some industries need history retained, and some boards want multi-year attribution unbroken. Fine, but notice that both needs are usually satisfied by keeping the old system accessible in read-only mode for a while, at a fraction of the cost of migrating and maintaining that data in your live instance forever.
A migration is the one moment where leaving things behind costs nothing extra and cleaning up later costs everything. It's the cheapest data hygiene you will ever do.
What's the oldest asset you migrated that nobody has touched since? I'll go first if you make me.