What is actually being compared, and what the comparison is not about
Two different things get called the sheets app. One is the order interface inside the agent platform: your item list, warehouse records, and parcel records, shown as a table. The other is a spreadsheet file, kept outside any platform, that a person updates by hand. The second one only ever contains what the first one already has, copied at some point in the past.
So the useful question is not which one has nicer formatting. It is which copy of the data is authoritative at the moment you act, and how many hours or days separate a change from the moment you can see it. Everything below compares the two on latency, authorship, versioning, and whether you can act from them.
- Latency: how long a warehouse change takes to appear where you are reading.
- Authorship: who created the row you are looking at, and who can edit it.
- Versioning: whether an old value leaves a trace after it is replaced.
- Actionability: whether the thing in front of you can start or stop a process.
Read the four together. Three of four can favour one option and the fourth can still decide the case, which is why the arguments in comment sections never converge.
The four dimensions side by side, scored on what each can prove
| Dimension | Order interface | Hand-maintained copy | Which one settles a dispute |
|---|---|---|---|
| Latency | Reads the live record | As fresh as the last manual refresh | Order interface |
| Authorship | Written by the platform | Written by whoever holds edit rights | Order interface |
| Versioning | Keeps state changes in order | Keeps only the value someone typed | Order interface |
| Actionability | Buttons that change the record | Cells you can highlight | Order interface |
| Readability | Fixed columns, dense rows | Free layout, colour, comments | Copy, for orientation only |
Every row ends in the same place, and that is not an accident of the table. A copy can beat the interface on readability and lose on all four of the others, because readability is about presentation while the rest are about who is allowed to write the record.
Dimension one: how many hours behind the copy runs
Ask a sheet owner how often they refresh it. Answers of daily, weekly, and monthly all exist, and the refresh interval is the latency, not a detail. A warehouse scan at 09:40 is visible in the interface at 09:40 and in a weekly copy at some point within the next seven days.
Latency changes which decisions the copy can carry. Cancellation eligibility, return windows, and storage clocks are counted in days, so a monthly copy cannot be used to judge any of them. Reading a trend over several weeks is different: there, a weekly snapshot is often better than a live table, because a live table cannot show you last week.
No refresh frequencies or service windows are quoted here as platform facts. If someone tells you a specific number, ask them for the page it came from and check it on the day you rely on it.
Dimensions three and four: history, and whether a click does anything
A record keeps order. If a status went from paid to arrived to packed out, the interface shows the sequence and the dates attached to it. A copy keeps only the last value someone typed, so a status overwritten twice leaves no trace, and the row that would have proved a timeline is gone.
Actionability is the shortest of the four. Cells do not book freight, merge parcels, or open a return. A sheet can tell you what you meant to do next; it cannot do it, and it cannot stop a clock that has already started. The [merge and pack stage](/steps/merge-parcel/) is a good test case, because the decision there depends on a storage clock rather than on a note in a cell.
| Your question | Read this | Why |
|---|---|---|
| Is my item in the warehouse yet | Order record | The intake scan is the record |
| How has my spend moved this month | Your own copy | The record does not aggregate for you |
| Can I still cancel this item | Order record | Eligibility is tied to a live state |
| Which of my items are unpacked | Either, if the copy is dated | A dated list is enough for planning |
A copy is built for its maintainer, and you are a guest in it
Every shared copy has one person who refreshes it, and that person refreshes it for their own rhythm. If they ship in batches every two weeks, the sheet will be accurate in the days around a batch and quietly stale in between. Nothing about that is dishonest; it is just somebody else workflow showing through the columns.
The practical test takes one minute. Find the newest row and the oldest row, and note the spread between their copied-on dates. A spread of three days means an active maintainer. A spread of six weeks means you are reading an archive, and archive rows are fine for orientation and useless for eligibility.
- Look for a copied-on column before you look at any status word.
- Check whether rows carry a platform item ID or only a name.
- Check who owns the file, and whether revisions are visible.
- If two rows disagree, keep the later row and mark the earlier one as superseded.
A copy you cannot date is not a worse copy; it is a different object. Read it as a snapshot of somebody else fortnight and re-check anything you plan to act on against the live record, including the storage and merging view if packing decisions are on the table.
When a copy is the right tool, and when it quietly costs you
Keep a copy for planning, not for deciding. Write one row per item with the platform item ID, the date you copied it, and the source stage you copied it from. A copy with those three columns is a usable index; a copy without them is a screenshot with nicer fonts.
- Start from the record, copy the ID, not the words.
- Add a copied-on date to every row you paste in.
- Re-check any row before you act on it; never act on an undated row.
- Delete rows whose item left the warehouse more than one cycle ago.
Keep the arithmetic in view. If a copy is refreshed every 30 days and an item typically sits in the warehouse about 5 days, then roughly 5 divided by 30 of the lines still match reality, about 17 percent, and about 83 percent of the status column can be wrong at the same moment. The fix is not a better sheet; it is a shorter interval and a date on every row.
A last practical note for anyone comparing platforms: the same four dimensions apply to every agent you look at, including [Kakobuy](/partners/kakobuy/). Ask about refresh interval and edit rights before you ask about layout.
Worked example: with a 30 day refresh cycle and an average item staying in the warehouse about 5 days, a copy that old has roughly 5 divided by 30, or about 17 percent, of its lines still describing the current state, so about 83 percent of its status column can be stale at once. (stated basis: Worked example)
This note explains the mechanics; the list is where the items are.