Institutional finance memory, and why your CFO's leaving takes it with them

Every fractional CFO I know has lived this scene: you roll off an engagement after a year of careful, compounding work. You wrote 12 board packages. You taught the CEO to read unit economics the way her board reads them. You logged 300 Slack messages documenting the “why” behind every tricky accounting call.

And then a new fractional CFO takes over. They inherit a Dropbox folder, four PDFs from three different tools, and a half-populated context doc that nobody updated after month three.

The client feels it immediately. The first month back is worse. The second is the same. The third, maybe, finds a groove. But the compounding curve that made the first year valuable is gone — because the memory is gone.

Memory is the product

The work of a senior CFO is not the package. It’s the accumulated judgment — how this particular CEO handles bad news, which metrics the board actually reacts to, which revisit triggers are still active, which decisions are on their second draft because the first one went poorly.

A dashboard doesn’t carry that. A Notion doc doesn’t either. Both are read-only. Both are frozen in time. Both start over with the new reader.

What you actually want is an analyst. Not “AI” in the abstract — a specific analyst who remembers every correction you’ve ever given her, every preference the board has ever signalled, every decision you’ve ever confirmed. One who gets sharper the longer she works with you.

The Rule of 40 test

A quick concrete example. When you show a cash-burning Series B SaaS company a Rule of 40 score, which margin do you use?

The default most tools reach for is FCF margin. That’s the variant Brad Feld originally wrote about. But for a Series B with a deliberate multi-cloud-migration investment, FCF margin badly understates the story — it turns a “temporary investment” into a “headwind” and hands the board the wrong mental model.

The operating-margin variant is the one your board actually wants to see. And the first time you explain that to your CFO and she edits the narrative accordingly, that correction should live forever. Not until next month. Not until the next Slack thread. Forever — attributable, searchable, ready for the next person who opens the package.

What we built

Marlow is the longitudinal memory of your practice. She writes every board package grounded in confirmed prior decisions. She cites her sources. She dedups her signals so preferences don’t pile up as 16 duplicates. She refuses to export a board deck that doesn’t reconcile with the PDF. And when your CFO rolls off, she stays.

That’s what “the compounding curve” actually looks like when you build it on purpose.