Information map · worked example
A nine-person wholesale bakery.
Supplies eleven cafés on standing weekly orders, plus a Saturday farmers market stall. Two bakers, one production lead, one driver, an owner who does everything else, and four part-time market staff.
01
What you told me
"We keep short-delivering people and finding out on Friday. And I can't tell you what we actually made on any given account without sitting down for an hour with the sheet."
The specific thing that broke: a café was under-delivered three weeks running. The order had changed by text, the change reached the driver but never reached the sheet, so it was baked short each week and invoiced at the old quantity — wrong in both directions.
02
Where information lives today
| System | Holds | Who touches it | Source of truth for |
|---|---|---|---|
| Paper order pad | Incoming order changes | Owner, production lead | Nothing, officially |
| Weekly Google Sheet | Orders, production quantities | Owner | Production |
| Kitchen whiteboard | Today's bake list | Bakers | What actually gets made |
| Text messages | Order changes, delivery problems | Everyone | Reality |
| Square | Market sales | Market staff | Market revenue |
| QuickBooks | Invoices, payments | Owner | What you billed |
- Systems that disagree: the Sheet and the texts. The texts are right and the Sheet is what you bake and bill from.
- Exists only in someone's head: which cafés have standing changes, which ones pay late, what the market usually sells out of.
- Exists only on paper: order changes taken by phone during a bake.
03
The trace
One real order, followed end to end. This is the core of the document.
Trace A — a standing café order, from request to payment
| # | Step | Where | Type | Notes |
|---|---|---|---|---|
| 1 | Café texts a change to the standing order | Phone | MANUAL RISK | Often mid-bake. No record beyond the thread. |
| 2 | Change written on the order pad, sometimes | Paper | MANUAL RISK | This is the step that failed three weeks running. |
| 3 | Pad transcribed into the weekly Sheet | Google Sheets | MANUAL | Second time the same fact is typed. |
| 4 | Sheet totalled into bake quantities | Google Sheets | MANUAL | Formulas exist but get overwritten. |
| 5 | Quantities copied to the kitchen whiteboard | Whiteboard | MANUAL | Third time. |
| 6 | Bake | — | AUTO | The one part nobody worries about. |
| 7 | Delivery list handwritten from the Sheet | Paper | MANUAL RISK | Fourth time. |
| 8 | Driver delivers; shortfalls and swaps texted back | Phone | MANUAL STALL RISK | Sits in a thread until someone reads it. |
| 9 | Variances reconciled against the Sheet — sometimes | Google Sheets | MANUAL RISK | When it's skipped, the invoice is wrong. |
| 10 | Friday: invoices typed into QuickBooks from the Sheet | QuickBooks | MANUAL RISK | Fifth time the same fact is typed. |
| 11 | Payment arrives, matched to invoice by eye | QuickBooks / bank | MANUAL STALL | |
| 12 | Unpaid invoices chased from memory | — | RISK | No list exists. |
Hops12 Manual hops10 Systems crossed6 Times one fact is typed5
Trace B — market sales back into production planning
| # | Step | Where | Type | Notes |
|---|---|---|---|---|
| 1 | Saturday sales rung up | Square | AUTO | Clean data, correctly captured. |
| 2 | Sales stay in Square | Square | STALL | Nobody looks at it after the day's total. |
| 3 | Next week's market bake decided from memory | — | MANUAL RISK | Sells out of some things by 10am, throws others away. |
This is the shortest trace in the document and the most interesting. The numbers you need already exist and are accurate. They never travel the last three feet.
04
What's actually broken
Ranked by what it costs, not by how hard it is to fix.
| # | Problem | What it costs | Basis |
|---|---|---|---|
| 1 | Order changes have no single home — text, pad, Sheet all partially true | Short deliveries and wrong invoices, in both directions | Three consecutive weeks on one account; you flagged two others |
| 2 | Delivery variances don't reliably reach invoicing | Under-billing. Amount not measured — a month of comparing delivery notes to invoices would size it | Step 9 is skipped under time pressure |
| 3 | The same order is typed five times | Owner time, plus five chances to fumble a number | Trace A, steps 2–10 |
| 4 | No unpaid-invoice list | Late payments found late. not measured | Step 12 |
| 5 | Market sales never inform production | Sell-outs and waste on the same morning | Trace B |
| 6 | One person holds the Sheet | The business pauses when the owner is unavailable | Steps 3, 4, 9, 10, 12 are all the same person |
05
What you can fix yourself
No developer, no project, no invoice from me. Do these first — they change what's worth building.
Do this: one shared number or address that every café uses for changes, and nothing counts until it's there. A free Google Form with four fields works. The rule matters more than the tool.
Time: an afternoon, plus one email to eleven cafés. Cost: nothing. Fixes: #1, most of #2.
Do this: standing weekly orders are already recurring; QuickBooks will raise them without you. You then only edit the exceptions instead of typing eleven invoices from scratch.
Time: two hours once. Cost: included in what you already pay. Fixes: most of #3, part of #4.
Do this: QuickBooks has an open-invoice report. Ten minutes every Friday, before you leave. That's the whole fix.
Time: ten minutes a week. Cost: nothing. Fixes: #4.
Do this: Square exports item-level sales. Fifteen minutes on Monday with last Saturday's numbers next to what you baked will beat memory immediately, long before anything is automated.
Time: fifteen minutes a week. Cost: nothing. Fixes: most of #5.
Five of the six problems above get materially better for the price of one afternoon and about half an hour a week. Do these before deciding whether you want anything built at all.
06
What's worth building
Only what the fixes above don't reach. Ranges, not quotes.
Orders arrive in one place and stay one record all the way through — production list, delivery sheet and draft invoice all generated from it rather than retyped. Delivery variances captured on the driver's phone flow straight back to the invoice.
Square sales pulled in automatically and shown against what was baked, so next week's quantities start from what actually sold.
These are estimates for budgeting, not a quote. A quote comes with a written scope that says what's excluded.
07
Recommended sequence
| Order | Do | Why now | Why not later |
|---|---|---|---|
| 1 | The four self-fixes in section 05 | Free, immediate, and they stop the bleeding on #1 and #4 | Doing them after a build means paying to automate a process you hadn't tidied |
| 2 | Run that way for a month; measure the invoice gap | Turns #2 from not measured into a number | Without it you'd be guessing at the value of the main build |
| 3 | One order record | Removes five retypes and the single point of failure together | Every week it waits is another week of manual reconciliation |
| 4 | Market sales into planning | Small, cheap, and pleasant once the hard one is done | No urgency — this one can wait indefinitely |
Deliberately not doing: replacing QuickBooks or Square. Both work, both hold correct data, and both are already paid for. The problem was never those tools — it's the gaps between them.
08
If you do nothing
Nothing collapses. This business works.
What happens instead is that it stays fragile in one specific way: the owner is a required step in five places, so growth and time off are the same problem. The short-delivery failure recurs at whatever rate it's been recurring, and the under-billing continues at a level nobody has measured. Add a twelfth café and the Friday invoicing session becomes the binding constraint on the week.
If you only ever do section 05, you'll have addressed most of the pain for free. That's a legitimate place to stop.
This is what the map produces.
A document like this one for your business. One conversation if you can walk me through a job yourself; interviews with your team if it takes several people. It starts with a free twenty-minute call, and if what you mostly need is a section 05, the map will say that.