Moving inventory off a spreadsheet without losing a week

Every row you type costs about a minute, and the spreadsheet is the last place you will ever be able to sort, filter and delete in bulk. Use it while you still have it. The catalog that goes into the app should be smaller than the one on the sheet, and the order you enter it in decides whether week one leaves you with something useful or just leaves you tired.

When the spreadsheet has actually run out

Most inventory spreadsheets are fine, and retyping one that works is an afternoon wasted. Work out first whether yours has actually failed, because "it feels messy" is not a reason and "nobody trusts it" is. Sheets fail in six recognizable ways, and each one has a test you can run in five minutes. Three or more symptoms and the file has stopped being a record of what you hold and become a list of candidates.

Be honest about the third one, because it is the symptom an app is least likely to fix. Stockout does not show who changed a quantity: stock movements carry no user, and there is no movement-history screen on Windows, Android or the web. What you gain is one record instead of four, not a name against every change. If attribution is the main argument for moving, this is not the tool that supplies it.

The first symptom is the expensive one. Two live copies diverge at the rate stock moves, and once they have diverged there is no third record to arbitrate, so the only repair is a full recount. That is the argument for fixing a cutover date before you type anything, which is the last thing this guide asks you to do.

There is no CSV import, and that decides the plan

No file Stockout produces is a spreadsheet: the outputs are an analytics PDF, a backup JSON and a label PNG. The only bulk paths into an account are restoring a backup JSON that the same account produced, and Reset to Sample Data, which loads demo records. A backup binds to the account that created it, so it is neither a handoff to a colleague nor a route in from a spreadsheet. Everything else is typed into the item editor or created from a scan. The entry work is real, and nobody can do it for you.

Because every row costs keystrokes, the sheet has to earn its rows before anything is typed. On the 620-row sheet this guide follows, typing it exactly as it stands is about nine hours of keying; typing what survives an hour of cleaning and a ranking by value is about two. The rest of the guide is that difference, in order: decide, clean, rank, type the top, scan the tail, count what you typed, freeze the sheet.

Six decisions to make before you type anything

These six are cheap now and expensive later. Stockout has no bulk editing, no multi-select and no mass update, so changing your mind after four hundred records means opening them one at a time. Write the answers at the top of the sheet before the first item goes in.

1. What counts as one item. One item is one thing you count and reorder as a unit. If you would never order half of it separately, it is one item. A bundle you assemble from other items is a kit rather than a second stock line: Make deducts component stock, Unmake returns it, and Remove/Sold reduces the built count without returning components. Get this wrong and the same physical unit is counted twice, once as a component and once as a finished thing.

2. Unit of measure. One unit per item, written the same way on every row. An item is tracked by Quantity, Weight, Length or Volume — the three non-Quantity modes add a Unit dropdown in the item editor — and nothing converts between units or turns boxes into eaches, so a sheet that mixes them carries the mistake forever. Where the shelf is still ambiguous, say it in the name as well: "Cable, 3-core — per meter" leaves nobody guessing.

3. Locations. The most expensive decision to reverse. Every item, kit, piece of equipment, checklist, folder and work order belongs to exactly one location, so the same product at two sites is two records with independent quantities and thresholds, and there is no transfer — moving stock is two adjustments. Deleting a location permanently removes every item, kit, piece of equipment and checklist at that site.

Only Owners and Admins can add, edit or delete a location, and a new account has to create one before it can use the app. If a second site is anywhere in your plans, settle the structure with how stock is run across sites before you type; the mechanics are in locations.

4. Naming format. Pick one pattern and apply it in the sheet, not in the app. Category — Description — Size suits most catalogs. Search in every list is live, case-insensitive and partial across name, SKU and category, so the pattern you settle on decides how findable the catalog is for years. Renaming four hundred items by hand later is a day you will not get back.

5. SKU scheme. One primary SKU per item, and keep the manufacturer's code as a second SKU rather than choosing between them — an item can carry multiple SKUs or barcodes, and search and scans match any of them. Resolve collisions on the sheet rather than in the app: two items sharing a SKU string scan to the wrong record every time, and nobody notices until a count disagrees.

6. Who may type. Creating and editing items needs the User role or above. Owner or Admin is required for locations, teammates and devices. Viewers cannot adjust stock. Decide this before you hand the sheet to a temp, and set the roles as you send the invites — see roles and permissions.

The setup order for a brand-new account is short: create the account, create the first location, invite the people who will be typing and set their roles, then start. Getting started covers the account steps.

Make the sheet earn its rows

Clean in the spreadsheet, where you still have sort, filter, find-and-replace and the delete key. None of those exist once the data is in item records — there is no bulk edit and no multi-select in the app, so the cleaning either happens on the sheet or it does not happen. Five passes, in this order, in one sitting, with one person deciding. A dedupe run by committee produces a third version of the sheet.

1. Dead rows. Sort by last issue date. Set aside the critical spares first — the parts that stop the job, anything held for a regulatory or safety reason, and anything on a replacement cycle longer than a year — and keep those whatever their last issue date. Then delete every remaining row with no movement in twelve months, or park it on a tab called ARCHIVE and never type it. These are the rows that make a catalog look impressive and make a count take all day.
2. Duplicates. Sort by name, then by SKU, and merge rows that are the same part under two spellings. Keep the better name and keep the discarded code for the second SKU field. Do not add the quantities together on trust — the commonest duplicate is one physical stash written down twice, and adding doubles it on the day it becomes your record. Go and count the part, enter what you counted, and add only where the two rows really did cover separate stashes.
3. SKU collisions. Filter for repeated SKU strings and resolve them now. A scan matches by SKU, so two items sharing one string is a permanent mis-scan rather than a tidy-up job.
4. Units. Fix any row whose quantity is in a different unit from its neighbours, per decision two. This is the pass people skip and the one they regret.
5. What not to bring at all. The long tail is usually not worth typing. It arrives by itself the first time somebody handles the part, and anything never handled again is never typed — which is the correct outcome rather than a gap.

Budget it at about five seconds a row to scan and decide.

Enter by value, not alphabetically

The failure this prevents is starting at row one and typing until morale runs out — you finish with every item beginning with A and none of the items that cause problems.

Rank the survivors by annual usage value (annual units used × unit cost) and cut at roughly 80% and 95% of cumulative value. Ranking items before you do anything else works that through with a table; if you already ranked for a count program or a reorder-point review, reuse the list rather than building a second one. Stockout has no A/B/C field, so the ranking stays in your working file. Override it wherever you know better — a $4 fuse that stops a line is a day-one item whatever its usage value.

Three groups, three treatments. A items get a full record on day one, with a Low Stock At value you have actually worked out. B items get a short record — name, SKU, quantity, Low Stock At — in the same week. C items are not typed at all; they arrive by scan-to-create as they are handled. Star the A items on the Inventory screen — bookmarks are per device, so each person stars their own.

If the A list runs past about forty items, the cut is wrong. Forty full records is a bit over an hour of typing and about half an hour of counting a month later. A hundred and fifty is neither, and it will not get done.

The mechanics of fast manual entry

Nine fields make what this guide calls a full record: name, SKU, quantity, Price At Cost, Price At Retail, Storage Spot, Max Capacity (units), Low Stock At and Near Capacity At. Four make a short one: name, SKU, quantity, Low Stock At. Decide which band gets which before you start rather than item by item — the deciding is what costs the time, not the typing. The item editor documentation covers every field.

Fill Storage Spot standing in front of the shelf, not from the sheet. It is free text with no zone or bin hierarchy, and the funnel sheet filters on the whole string — so "Aisle 1, Shelf A" and "Aisle 1 Shelf A" become two separate entries in that filter and neither one lists the whole aisle. Write the pattern down next to the naming format in decision four and use it on every record.

Set the threshold as you go and never as a second pass. "Thresholds later" is a second trip through a hundred and twelve records. What to put in Low Stock At is arithmetic you do outside the app: Stockout does not compute a reorder point, does not store one, and has no safety-stock field. Working the number out takes about ten minutes an item once usage and lead time are to hand, and a rough number entered now beats a precise one entered never.

Let the scanner type the SKU. Barcode scanning works two ways: the device camera on Android, and a USB or Bluetooth reader in keyboard-wedge (HID) mode on Windows and the web. When a scanned code matches nothing, Stockout offers Create New Item with the SKU pre-filled, or Add to Existing Item. That is scan-to-create, and it is the entire strategy for the C tail. The scan buttons live in the footer of the Inventory screen, they are hidden for Viewers, and scanning applies to inventory items only.

For parts that carry no code, you can print one — but only for the parts you could not scan, and expect the first pass to be slower than typing. A label template puts the item's SKU or its Name in the code, and you export one filled-in label at a time as a PNG. Choosing between a printed barcode, a QR code and an NFC tag is a decision worth making once; the steps are in label printing.

Work in 45-minute sessions. As a rule of thumb accuracy falls off after that, and every error costs more to find than it did to make.

If two people type at once, split the sheet by row range rather than by whoever gets there first — when two people change the same field at once the last change saved wins, and only changed fields sync.

Before typing a row, search its SKU. Search is live, case-insensitive and partial across name, SKU and category, and during a migration it is the only duplicate check you get.

A realistic time budget

These are planning figures, not measurements: 90 seconds for a full record, 45 for a short one, 30 for a scan-created one, each with 20% added for interruptions. Time your own first ten records and replace them — a sheet that needs a decision on every row runs at half these speeds. Plan against the right-hand columns.

The bottom row is why the tail is cheap: the SKU arrives from the barcode, so only three fields get typed. It is also why the tail should never be scheduled as a block — those items appear at the bench, a few a day, for months.

Pairing helps less than it looks. Assume a reader-and-typist pair manages about 1.6× one typist rather than 2×: the 131 minutes of keying becomes roughly 82 minutes elapsed but costs 164 person-minutes — 49 minutes of calendar time and a second pair of eyes, bought for 33 extra person-minutes. Worth it on the A list, where an error is expensive and hard to find later. Not worth it below that.

One caveat about the tail. Those 304 C items would be 3 hours 2 minutes if you sat down and did them in a block, and you never will. They spread across months at a handful a day, and the ones nobody handles again are never typed at all — if half of them go untouched for a year, the tail costs about an hour and a half in total, thinly enough that nobody notices.

Verify what you entered with a count

Count what you typed, not the whole stockroom. The point is to test the transcription rather than the shelf, and transcription is where a migration goes wrong — the picking was already fine yesterday.

Score it as exact matches, by item rather than by unit, because "close" is not a match.

If the B sample misses 90%, the sample was the warning — count the rest of the B list rather than only fixing what it turned up.

Correct anything that misses the same day. Single units go through the + / − buttons on the item card; a bulk add or remove goes through the three-dot menu → Adjust, which shows a live preview before you commit. Adjusting stock needs the User role or above, so a Viewer cannot do it. And an adjustment can never take quantity below zero — it stops there, and the movement quietly under-records by the difference.

Verify once, then stop. How often to count afterwards, how many items, and who owns the round is a separate discipline with its own arithmetic.

Back up at both ends

Take a backup the moment the A list is in, and another when the tail stops trickling. Two files, dated, stored somewhere that is not the laptop doing the typing. Each takes about a minute.

The file is one timestamped JSON covering items, kits, equipment, checklists, suppliers, customers, folders, work orders and label templates, together with their child rows. Activity logs and stock movements are not in it.

One file covers the whole account, not one site. That is exactly what you want on the day somebody deletes a location and takes every item, kit, piece of equipment and checklist at that site with it.

A backup binds to the account that created it: not a migration path between two Stockout accounts, not something you hand to a colleague to import, and not a route in from a spreadsheet. Restore, Reset to Sample Data and Clear Data all write to the shared cloud, so they affect the whole team rather than the device that ran them. Backup and restore has the steps.

Kill the sheet on a fixed date

Pick the date before you start typing and write it at the top of the sheet. Ten working days from the first entry session is enough; fourteen is the outside limit. Miss it and you are running two records indefinitely.

Two live records diverge at the rate stock moves and there is no third record to arbitrate: five issues a day recorded in only one of the two leaves 50 movements out of step after ten working days, with no way to tell which record is wrong.

During the overlap, one direction only. Stock changes go into the app and the sheet is read for reference. Never the reverse, and never both.

Save a copy named ARCHIVE with the date on it, set it read-only for everyone, and delete every formula in the live copy that computes a quantity. A sheet that still adds up is a sheet somebody will still trust.

The tail is not a reason to extend the date. C items appear through scan-to-create as they are handled, and a part nobody touches during the overlap window was never going to be typed anyway.

On the freeze date, take the second backup and switch on the Low Stock and Out of Stock alert types in Settings. The four types — Low Stock, Out of Stock, Near Capacity, Full Capacity — toggle independently and apply to every item at the location that device is set to; there is no per-item alert switch, so it is the A-item thresholds you typed that decide which ones actually fire. Expect a burst over the first few days as thresholds meet reality. Alerts are edge-triggered — one fires at the moment a change newly crosses a threshold, an item that is already low does not re-alert, and there is no scheduled digest — so the burst settles on its own.

Set alert recipients on every device that should send. They are a comma-separated list saved per device or browser and they do not sync, and they do not fire at all while a device is in the All Locations view. Email alerts work on every platform; mobile push is the Android app.

When a spreadsheet is still the right tool

Under about fifty items, one person handling the stock, nothing time-critical when it runs out and no second site — a sheet is the cheaper answer and migrating is a hobby. The decision rule is who needs to read the number and where they are standing when they need it, not how many rows there are.

The arithmetic in this series stays in a spreadsheet permanently. Reorder point, safety stock, service levels and Z factors, annual usage value and the A/B/C cut, inventory turnover, days of supply, carrying cost, fill rate, dead stock and aging: Stockout calculates none of them, charts none of them and stores none of them. The single reorder-ish input in the whole product is the per-item Low Stock At threshold.

Supplier data stays with you too. Stockout does not track, store or know supplier lead time, minimum order quantity, supplier part numbers or supplier pricing. What it holds is an org-wide directory of vendor name, ordering-portal URLs and contacts, with items linked from the item editor's "Sourced from" picker. Reordering means tapping a saved portal link and placing the order on the vendor's own site.

If you want the movement data back in your own sheet, the read-only public API is the route: GET and HEAD only across 24 collections, including the stock-movement ledger, with keys minted at Settings → Developers → API Keys by Owners and Admins. Read-only means read-only. It cannot push counts back, cannot power a two-way integration and is not a way to get a spreadsheet in, and a key reads the whole organization rather than one site. See the API documentation.

One expectation to set before somebody opens Analytics on a week-old catalog. Trend lines are reconstructed by unwinding stock movements from today's on-hand total rather than stored as point-in-time snapshots, so a freshly typed catalog has no history at all — the charts start the day you start recording movements. Analytics reads at most the last 13 months of stock movements, and deleting an item removes it from analytics entirely, including its past movements, so every historical point moves when you do. Analytics ships exactly eight charts — Total Stock Per Day, Weeks Cover, Sell Through, Intake & Outtake, Stock to Capacity, Stock at Cost vs Retail, Work Order Revenue vs Cost and Work Order Profit — and the features page covers the rest of what ships.

The sheet is where you work out what the number should be; the app is where the number lives and gets watched.

The pre-flight checklist

In execution order. Nothing here takes long on its own, and doing them out of order is what turns four and a half hours into a fortnight.

• Write the cutover date at the top of the sheet before anything else.
• Set the critical spares aside — parts that stop the job, safety or regulatory stock, anything on a replacement cycle longer than a year — then delete every remaining row with no movement in twelve months, or park it on an ARCHIVE tab.
• Merge duplicate rows and resolve every shared SKU string.
• Fix any row whose unit of measure differs from its neighbours.
• Write down what counts as one item, and what is a kit instead.
• Settle the location structure — every item belongs to exactly one site and there is no transfer.
• Pick one naming pattern and apply it in the sheet.
• Pick one primary SKU per item; keep the manufacturer's code as a second SKU.
• Create the account and the first location; invite the people who will type and set their roles.
• Turn Auto-Remove at Zero off for the duration of the entry push.
• Rank by annual units × unit cost; cut the A list at roughly 80% of value and keep it under forty items.
• Type the A items as full records, each with a Low Stock At value you calculated.
• Type the B items as short records in the same week.
• Leave the C tail to scan-to-create at the bench.
• Count the A items and a 20-item B sample; correct every miss the same day with Adjust.
• Take the backup, set alert recipients on each device, and make the sheet read-only on the date you wrote down.

Week two is undramatic, which is the point of doing week one in this order. The A items get counted again, the two or three thresholds that fired wrongly get corrected, a handful of C items appear at the bench each day, and the sheet stays read-only. What makes the new record worth the afternoon is what you do with it next: the practices that stop stockouts are the reason for the whole exercise. The rest of the series works through the arithmetic and the routines that sit around them.