From subscription docs to welcome letters: Reconciling at scale with DwellFi
.png?table=block&id=3979cd9a-2bee-8063-b51c-c4a1f78c1092&cache=v2)
By Deepak Sheoran, Founder and CTO, DwellFi
A fund administrator we work with opened Monday with 400 LP subscription documents to reconcile. The close was Friday.
Each document carried a commitment amount that had to match a contribution form. Four hundred pairs of numbers, pulled from PDFs that never quite agree on formatting, and keyed into one master spreadsheet by a team that already had a full week ahead of them. That is the job at scale. It is not hard in the way a math problem is hard. It is hard in the way that carrying water is hard, one bucket at a time, four hundred times, without spilling.
They found the first error around noon on Wednesday. Row 212. A commitment logged as $2,500,000 that should have read $25,000,000, a missing zero from a fast morning three days earlier.
Here is the part nobody outside operations understands. The problem was not fixing row 212. The problem was that once you find one error in a spreadsheet you have built by hand, you can no longer trust any of the other 399 rows. The error does not stay in its cell. It spreads backward through every number you already checked and forward through every number you have not. Was row 88 right? You checked it Tuesday. Are you sure? A single wrong cell turns three days of finished work back into unfinished work, because the only way to trust the sheet again is to re-verify the whole thing.
So they did. They started over Wednesday afternoon and reconciled through the night. The welcome letters went out Friday, barely, and the team spent the weekend recovering instead of doing anything that mattered.
Why the spreadsheet fails at scale
A spreadsheet is a wonderful tool for ten rows. It is a liability for four hundred.
The reason is that a hand-built spreadsheet has no memory of how it was built. It shows you the number in the cell. It does not show you where the number came from, who typed it, or whether it was ever checked against the source. When every value is entered by a person, the sheet is only as reliable as the least caffeinated hour of the week that produced it. And you cannot see which hour that was.
At ten rows you can hold the whole thing in your head. At four hundred you are trusting a process you cannot audit, run by people who are tired, against a deadline that does not move. The math does not get harder as you scale. The trust does.
What reconciliation looks like when the machine holds the line
DwellFi reconciles the same 400 documents differently, and the difference is not speed for its own sake.
The workflow reads each subscription document and each contribution form directly. It pulls the commitment amount from the source and matches it against the funded amount, and it does this the same way every time, because the check is deterministic. Row 212 does not depend on which hour of the week it was processed. A $2,500,000 figure that should read $25,000,000 surfaces as a flagged exception with both source values attached, so a person looks at the mismatch instead of hunting for it.
That last point is the one that matters at scale. Your team stops reconciling 400 rows and starts reviewing the 6 that disagree. The other 394 are matched, logged, and traceable back to the exact page they came from. When someone asks on Thursday whether row 88 is right, the answer is not "I checked it Tuesday, I think." The answer is on the screen, with the source.
And when an error does turn up, it stays contained. A flagged exception is one row. It does not poison your confidence in the rest, because the rest were never reconciled by the same tired hand that missed the zero. They were reconciled by a process that runs identically on row 1 and row 400.
The scale argument, plainly
Onboarding one investor by hand is annoying. Onboarding four hundred by hand is a structural risk, because the failure mode is not a slow week. It is a wrong number that you either catch late or ship.
DwellFi runs the reconciliation inside your own cloud, on deterministic checks you can trace to the source document, built for fund administration rather than borrowed from a general tool. The commitment amount that reconciles today reconciles the same way at your next close, and the audit trail is already written by the time anyone asks for it.
See it on your own data
Send us your next batch. Not a demo file, your real subscription documents, the messy ones with side letters and a commitment amount that does not match the contribution form.
We will reconcile them live and show you the exceptions before the call ends. You will see which rows disagree, why, and where each number came from. Then you decide whether your team should ever spend another Wednesday night re-checking row 88. Bring the batch that scared you last quarter.