You Don’t Have a Data Problem. You Have a Seeing Problem.

At most investment firms, someone maintains capital-structure records by hand. Every round, every valuation, every investor’s check and ownership stake, entered manually into a spreadsheet. The backlog compounds. One firm put it plainly: a good two or three years of backlog that needs to be inputted… since fund five we’ve stopped tracking it.

Baked into that work is an assumption: gathering the data is the hard part.

It isn’t.

What a keyhole shows

Every funding document at this firm — term sheets, SAFEs, purchase agreements, cap tables, side letters — had already been delivered, signed and timestamped. Straight into email. The record existed in full.

Search it the obvious way — two partners’ inboxes, last twelve months, emails with attachments, standard funding keywords — and it returns 73 documents. That looks like a result. It is a rounding error.

Open the aperture — the right mailboxes, a multi-year window, searching the way funding paper actually moves rather than the way it is imagined to — and the picture changes entirely.

The candidate surface — every message worth examining at all — was roughly thirteen times larger the moment the search stopped looking through a keyhole. Not because new data arrived. Because the aperture changed.

Why the gap is invisible from the inside

Each of those blind spots is individually reasonable. Signature pages arrive as images, so keyword search cannot see them. Cap tables live inside spreadsheet attachments, where the relevant content is a cell rather than a phrase. An increasing share of documents never arrives as an attachment at all — just a link to a signing platform or a shared drive.

None of that registers as a failure, because there is no error message for a document you never knew was there. The search returns results. The results look plausible. Nobody is told what the aperture excluded.

It degrades in a chain, one defensible link at a time. No domain on file, so no way to identify the CEO; no search terms, so no mail associates with the company; no associated mail, so the company looks inactive; and because it looks inactive, nobody goes back to fix the original gap.

What follows

The useful question is not what data to acquire. It is whether the data already present is being read at all, and at what aperture. Most firms have never measured the second number, which is why the first question keeps getting asked.

What the shape of the work implies

Manual data entry is not a task with a beginning and an end. It is a shape — a recurring pattern that expands with the portfolio and never fully resolves. Which is why it survives every attempt to fix it with discipline: the backlog is not evidence that someone fell behind, it is the natural state of a job whose input grows faster than any person’s throughput.

That distinction changes what a good outcome looks like. The goal is not to help someone do the entry faster. Time saved on a task that regenerates itself is not saved. The goal is for the category of work to stop existing — for the record to be current because reading is continuous, not because someone caught up.

The firms that have watched this happen describe the before-and-after in units that don’t match: months of accumulated manual work against hours of reading. That gap is not a productivity gain. It is the difference between a shape and a task.

Reading at full aperture is not something a person can do — not across years of mailboxes, thousands of attachments, and signature pages that arrive as pictures. It is exactly what machines are now good at, and it produces a specific, unglamorous result: the record the firm already owned, finally legible.

Written by Oded Noy, ContextApp.