A fixed asset conversion looks like a technical exercise and behaves like an accounting one. Records move from one structure to another, and along the way every ambiguity in the source data has to be resolved into a specific value. Whoever writes the mapping is making accounting decisions, whether or not anyone in the project plan called it that.
It is also very close to a one-way door. Once a converted register is reconciled, signed off and used to produce a depreciation run, unwinding a bad mapping means unwinding everything built on top of it. The cheapest moment to get this right is before the first record moves, and it is cheaper than the second-cheapest moment by an order of magnitude.
One note on where I sit in this, because it shapes nothing on the list below. I run conversions hands-on inside Sage Fixed Assets — Depreciation, CIP and Tracking. On any other destination platform I do the readiness, mapping decisions, cleanup, reconciliation and validation alongside whoever owns the load. The decisions in this article are identical either way; only the hands on the keyboard change.
Decide what the fields mean before you decide where they go
Mapping is usually approached as a matching problem: source field to target field, tidy arrows, done by Thursday. The harder question is what the source field actually contains, which is very often not what its name suggests. Description fields that quietly accumulated location codes. Acquisition dates that are sometimes invoice dates. A class field that three different eras of staff used three different ways, each of them convinced they were being helpful.
Before mapping anything, work through the fields that carry accounting consequence and write down, for each: what it is supposed to mean, what it actually contains, and what the rule will be going forward. The fields that earn this treatment are usually asset class, acquisition date, in-service date, cost, useful life, depreciation method, location, entity, and whatever links a record back to the project that created it.
Settle the accounting questions
These are the decisions that determine whether the converted register is defensible a year from now.
- Depreciation history: carry or recalculate? Carrying history preserves the audit trail and the errors. Recalculating produces internal consistency and a variance you will be explaining. Both are legitimate; the choice has to be deliberate and written down.
- Which books come across? Financial, tax, state, alternative minimum — and whether they currently agree. Conversions regularly reveal that tax and GAAP parted ways years ago for reasons nobody recorded, which is always an interesting Tuesday.
- Fully depreciated and disposed assets. Whether they convert at all, and if so with what flag. Bringing everything increases fidelity and clutter; leaving it behind loses history you may want later.
- Componentization. Assets recorded as lumps cannot be componentized later without effectively re-creating them. If the future state needs component-level detail, conversion is the natural — and cheapest — moment.
- Lives and methods. If the target environment enforces a class-based standard and the source data never followed one, conversion surfaces every exception at once. Decide in advance which exceptions are intentional.
- In-service dates. Dates set for convenience rather than readiness are extremely common, and conversion is the last easy chance to put them right.
Fix the reconciliation first, not after
If the sub-ledger does not tie to the general ledger today, it will not tie after conversion, and the conversion will get the blame. Establish the reconciliation before the move so there is a known starting position. A documented, explained variance is a workable baseline. An unexplained one becomes a permanent, unwinnable argument about whether the conversion caused it.
The same goes for population completeness. Confirm what the register is supposed to contain — against the project system, against capital spend, against the physical estate where that is practical — before declaring the source to be the source of truth.
Decide who owns the decisions
Every conversion generates a steady stream of small judgement calls: this record has no in-service date, this class does not exist in the target, this cost includes something that should never have been capitalized. Under deadline pressure those get resolved by whoever is closest to the file, which is frequently a technical resource with no mandate whatsoever to make accounting decisions. That is not their failing. It is a design failing, and it is ours to prevent.
Name the person who owns those calls before the work starts, and log every one that gets made. The log is not bureaucracy — it is the document that explains the converted register to an auditor, and to whoever holds your role in three years and is trying very hard to understand you.
Plan validation before you plan the cutover
Validation is where conversions are genuinely won, and it needs to be specified in advance rather than improvised at two in the morning during go-live weekend.
- Control totals. Cost, accumulated depreciation and net book value by class and by entity, source versus target, before and after.
- Record counts with reasons. Any difference between source and target counts should be explained by a documented decision, not discovered later by an auditor with good instincts.
- A parallel depreciation run. Run one period in both environments and reconcile the difference. Unexplained variances found here are the cheapest ones you will ever buy.
- Targeted sample testing. Pick the hard cases on purpose — the oldest assets, the componentized ones, the ones with mid-life adjustments, transfers, partial disposals and impairments.
- Report parity. Reproduce the reports leadership and audit actually use, and confirm they answer the same way. This is the test that decides whether anyone trusts the new system in month two.
What "clean data" really means
Clean data is not data without blanks. It is data whose every field has an agreed meaning, a documented rule, an owner and a lineage you can explain out loud. A register with gaps you understand is in far better shape than a complete one whose values nobody can account for.
So treat conversion as the accounting event it has always been. The technical half — extract, transform, load — is the straightforward part, and it keeps getting easier as the tooling improves. The half that decides whether the new environment is trusted is everything settled before the first record moves.
