Build vs buy: the crypto subledger decision | Cryptoworth
Build vs buy

Your engineers can build a crypto subledger. It won't survive the audit.

Pulling transactions off-chain and pushing them into NetSuite is a few sprints of work. Any competent team ships v1. What v1 cannot do is prove itself to an auditor, handle the long tail of edge cases, or be right about tax in a way someone will sign their name to.

We sell a subledger, so read this with that in mind. The section on when you should build is real, not a setup.

The real ceiling

It doesn't fail on engineering. It fails on compliance.

Teams assume the constraint on an in-house subledger is technical — more chains, more throughput, more integrations. It isn't. The constraint is that a financial system has to be defensible, and defensibility is not something you can ship in a sprint.

Auditors want proof

"It's correct" is not an audit position

An auditor does not accept an output. They test a method. If your cost-basis rules exist only as implementation, there is no documented method to test — only code someone reverse-engineers under time pressure, at the worst point in your year.

Tax engines

A wrong tax engine is worse than none

An engine that is subtly wrong produces numbers that look right, get booked, get filed, and get discovered later. The remediation cost isn't fixing the code. It's the restatement, the amended filings, and a finance team explaining a number it did not compute.

AI is not a shortcut

A model that hallucinates a classification is a liability

Pointing a general-purpose model at raw chain data feels like leverage — until it confidently mislabels a transaction type it has never seen. Accounting has no tolerance for a plausible answer. A ledger has to be right, and has to know when it isn't.

The long tail

The edge cases are the work

The first 95% of transactions are easy and every build gets them. The last 5% — the wrapped-then-migrated-then-slashed position, the failed bridge, the airdrop with no price feed — is where the close breaks. That tail never ends, and it is on no one's roadmap.

The maintenance curve

Two years in the life of an in-house subledger

None of these events are scheduled. Each arrives as an interrupt against a roadmap that was supposed to be about your product.

Month 0 → 24 Unplanned work, in order of arrival
M0
v1 ships
Three chains, two exchanges. It works. Everyone is pleased.
M3
Treasury adds an L2
New RPC, new gas semantics, new bridge events to classify.
M7
First reorg past finality
Transactions you already booked no longer exist. The close reopens.
M11
Protocol changes rewards
Staking accrual silently starts producing the wrong number.
M14
Auditor asks for the method
Cost-basis rules live in code, not documentation. Someone rewrites both.
M17
Token migration
Historical basis must carry across a contract that no longer exists.
M19
The engineer leaves
The system now has no owner who can explain why it does what it does.
M24
The rebuild conversation
Now framed as risk, not cost — and decided by the CFO, not the CTO.
The pattern is not unique to crypto. Firms that built proprietary accounting platforms are revisiting those decisions industry-wide, as the engineers who wrote them move on and true cost of ownership becomes impossible to ignore. Digital assets just run the clock faster.
What you cannot build

The edge cases are already in our data.

Since 2017, every correction is a lesson we already learned

Cryptoworth has been reconciling digital assets since 2017. Every break, every mislabeled transaction, every reorg and migration and strange protocol event we corrected became part of how the system classifies the next one. That corpus is the product. It is the one thing a new build cannot start with at any budget — because it isn't code. It's years of being wrong in public and fixing it.

Our AI agents run on top of that history, not instead of it. They're constrained by classification logic that CPAs, public-company controllers, and audit teams have already put through the wringer — which is the difference between a model that assists a close and a model that contaminates one.

Correction history
Since 2017 — the long tail, already encountered and resolved.
Tested by
CPAs, controllers at public companies, and audit teams in live engagements.
Provable
SOC 1 and SOC 2. A documented method, not an inferred one.
The cost you actually pay

Cheaper on the invoice is not cheaper on the business.

Set a license line against a salary line and building can look competitive. Set fully-loaded cost against it — engineering time, delayed closes, audit remediation, and the roadmap you didn't ship — and it usually isn't close.

Build
Cryptoworth
Engineering — a permanent share of your team, on plumbing that is not your product.
Engineering — none. Your team ships what customers pay for.
Edge cases — discovered at close, one painful quarter at a time.
Edge cases — already encountered, already corrected, already in the classifier.
Audit — your engineers explain your method to your auditor, in your busiest week.
Audit — documented method, SOC-certified, audit-ready by default.
Tax risk — carried by you, found late, remediated expensively.
Tax risk — logic public companies and their auditors have already tested.
Focus — the ledger becomes a recurring interrupt on the roadmap.
Focus — it becomes something you stop thinking about.
The honest part

Sometimes you should build. Here's when.

If any of these describe you, a vendor subledger is probably the wrong tool — and we would rather say so now than nine months into an implementation.

Build if

  • Your accounting treatment is the product — you're an exchange, a custodian, or a protocol whose ledger is customer-facing
  • You have a standing platform team already funded to own financial infrastructure indefinitely
  • Your regulatory position genuinely forbids a third party touching the data

What these share: the ledger is a source of advantage. If it isn't — if it simply needs to be correct and never thought about again — you're paying engineers to maintain plumbing, and paying finance to live with its failures.

Before you decide

Five questions worth answering out loud

01
Can your auditor test your cost-basis method without reading your code?
If the method exists only as an implementation, it isn't documented. It's inferred — and auditors don't accept inference.
02
What happens when your tax logic turns out to have been wrong for four quarters?
The cost isn't the code fix. It's the restatement, and whose name is on it.
03
Who is on call when a reorg breaks a closed period?
If the answer is a name rather than a team, that name is a single point of failure in your financial reporting.
04
Would you describe this system in a job listing as "how we win"?
If the honest answer is "it's table stakes," it's a buy.
05
What did your engineers not build this quarter?
That number belongs in the build estimate. It never is.
Next

If it's plumbing, let us maintain it.

Cryptoworth is the subledger between your on-chain activity and your general ledger. 1,000+ sources, SOC 1 and SOC 2, native NetSuite, QuickBooks, and Xero sync — with agents built on nearly a decade of digital-asset corrections. When a chain reorgs at 2am, it's our pager, not yours.