Almost every refurbishment operation starts in a spreadsheet, and that is the right decision. A spreadsheet is free, immediate and shaped exactly like the problem on the day it is created. The difficulty is that it degrades gradually rather than failing, so the moment to move is easy to miss - and the cost of missing it is paid in data that was never captured.
Threshold 1: more than one person edits it
The first failure is concurrency, and it arrives earlier than expected. Two people editing the same file produce two truths, and reconciling them is manual work that nobody has been assigned.
Shared cloud files solve the mechanics and not the problem. Simultaneous editing works; simultaneous interpretation does not. One person records a grade as B, another as Grade 2, a third leaves it blank because the device is still on the bench.
The tell is the appearance of a person who is the only one who really understands the file. That person is now a single point of failure for the operation, and the situation will not improve on its own.
Threshold 2: the device has a history, not a state
A spreadsheet row holds a current state. A device has a sequence: received, triaged, erased, repaired, failed test, repaired again, graded, sold, returned, re-repaired.
The usual workaround is one sheet per stage, and it works until a device has to be traced across them. Now the question of what happened to this serial number requires opening six files and trusting that the identifier was typed identically in all of them.
The information genuinely lost here is timing. When a row is overwritten, how long the device spent at each stage disappears, and with it any ability to find where the operation is slow. Cycle time is the first metric a spreadsheet operation cannot produce.
Threshold 3: parts have to be reserved
Stock counting in a spreadsheet works while stock only leaves. It breaks when stock is committed in advance: a part allocated to a device that is waiting, but still physically on the shelf and therefore still counted as available.
The visible symptom is a technician discovering an empty shelf the sheet says is stocked. The invisible one is worse - parts ordered that were already reserved, capital sitting on a shelf, and a purchasing decision made against a number that was never accurate.
Harvested parts make it sharper. A part recovered from a dismantled device has no purchase price and a provenance worth keeping. A spreadsheet can record that it exists; it struggles to record what it cost and where it came from, which is exactly what deciding to dismantle another unit requires.
Threshold 4: a customer asks for evidence
This is the threshold most operations actually migrate on, and it arrives as an email. A client wants the erasure certificates for a batch delivered eight months ago, or an auditor picks five serial numbers and asks for their complete history.
The evidence usually exists. It is in a folder of PDFs named by date, in a second folder named by client, in an inbox. Producing the answer takes a person two days, and the two days are not the point - the point is that the operation cannot demonstrate control of its own records.
Once a certificate is filed against the device it belongs to rather than in a parallel archive, the question becomes a lookup. That is the whole difference, and it is not a small one commercially: corporate and public buyers increasingly ask before signing, not after.
Threshold 5: you cannot tell what is profitable
The last threshold is the quietest. The operation is running, revenue is real, and nobody can say which device categories, which supply sources or which repair types actually make money.
Cost per device requires parts, labour and overhead attributed to a serial number, including the devices that consumed resources and never sold. A spreadsheet can hold any one of those. Holding all of them, per device, across months, is where it stops.
Operations that cross this threshold without noticing make decisions on gross margin at the batch level, which averages profitable work with unprofitable work and hides both. The first bad quarter then has no diagnosis.
Migrating without losing the year you already ran
Three things make the difference between a migration and a restart.
Decide what the identifier is before anything else. Serial number, IMEI, or an internal reference - but one of them, applied consistently, because every record you hope to carry across has to attach to it. Historical data whose identifier was typed four different ways does not migrate.
Do not import everything. Open work, current stock and the evidence you are contractually required to keep will earn their place; three years of closed rows in a format nobody trusts will not. And run the two in parallel for a defined, short period with a stated end date - long enough to catch what the new system does not cover, short enough that nobody settles into maintaining both.