At 4:40 on a Tuesday afternoon, a quote went out to a Beirut contractor with a 30-day price that was really a 7-day price. The supplier's message said seven days. The chat assistant's summary kept it. The spreadsheet did not. The quote template, which prints "Prices valid 30 days" whenever nobody tells it otherwise, filled the silence.
This is a composite. The tile importer, the operations manager and every number below are illustrative, assembled from the shape of situations I have seen in more than one office. It is not a study and it is not one client. But I would be surprised if you could not find the same shape in your own week: the tool never made the error. The carry did.
The Quote That Lost Its Expiry Date
Picture Rima, who runs operations for a tile importer with 24 staff. On Monday morning her Italian supplier sends a price update on WhatsApp for three porcelain lines, in euros, with one line at the end: "Prices valid 7 days, container closes Friday."
Rima pastes the message into her chat assistant and asks for a clean summary. The summary is good. It lists the three lines, the prices, and the seven-day validity. She copies the prices into the pricing spreadsheet. The validity line stays behind, because the spreadsheet has no column for it and she is thinking about the numbers. On Tuesday a contractor asks for a quote on two of the lines. She builds it from the spreadsheet into the quote template, a colleague sends it as a PDF at 4:40, and the template prints its default: valid for 30 days.
On the twelfth day the contractor orders at the quoted price. The supplier's price has moved back up. The importer honours the quote, because the PDF says it must, and the difference comes out of its margin. It takes two weeks for anyone to trace it, and the first suspect is the assistant. The assistant had it right.
Here is the same job as a list of hand-moves. Five tools touched it: WhatsApp, the chat assistant, the spreadsheet, the quote template and email. Between them there were nine carries, and three of them lost something.
| # | Carry | What went missing |
|---|---|---|
| 1 | Supplier's WhatsApp message to chat assistant | Nothing |
| 2 | Assistant's summary to the spreadsheet | C: the "valid 7 days" line had no place to land |
| 3 | Contractor's WhatsApp request to the spreadsheet | Nothing |
| 4 | Spreadsheet to quote template | F: two tile sizes that sat on separate lines ran together as one |
| 5 | Assistant's intro wording to quote template | Nothing |
| 6 | Template to PDF to email | S: the sender assumed Rima had checked it against the supplier's message; she had checked it against the spreadsheet |
| 7 | Email to the contractor's site manager on WhatsApp | Nothing |
| 8 | Contractor's reply to the spreadsheet | Nothing |
| 9 | Spreadsheet to the order form | Nothing |
Nine carries; three marks; one of them touching a date that carried money.
Draw the Path, Then Circle Every Carry
This is the first part of what I call the Carry Map, and it has three parts: list the carries, rate what each one drops, and fix one seam in a set order. The first part sounds trivial, and it is, which is why it works.
Part 1 is to list the carries. Take one real job and walk it from first input to final output. Every time a person moves content between two tools, write that move on its own line, such as "WhatsApp to chat assistant" or "assistant to spreadsheet". If you did it by selecting, copying, retyping, forwarding or exporting and re-uploading, it is a carry.
People undercount, and there is a reason. Copy and paste feels free. It takes four seconds, nobody budgets for it, and it leaves no record. Ask Rima how many tools her quote touches and she says "the assistant and the spreadsheet", because those are the tools she thinks of as work. The phone, the template and the email are where the content sits in between, and the carries are in those places.
A carry also belongs to nobody. Each tool has an owner and a login. The space between two tools has no one watching it.
What a Carry Drops: Format, Context, Sign-off
Part 2 is to rate what each carry drops. I use three letters, and only three, because a longer scheme does not survive a busy week.
F is for Format. The columns, the currency, the line breaks, the digits. In a bilingual office this shows up often: a price written with Arabic-Indic digits arrives in the spreadsheet as text and will not sum, or "1.250" means one thousand two hundred and fifty in one convention and a little over one in another. A pasted price list loses its line breaks, and two tile sizes become one. Format losses are loud. Somebody usually sees them, because the thing looks wrong.
C is for Context. The why behind the content: the validity date, the exception, who the customer is, what was promised on the phone. Context is the quiet one. After the carry the number still looks perfectly healthy, and nothing on the page says that something is missing. In the example, the seven days vanished at carry 2 without a single visible error, and the template then supplied a confident, wrong sentence in its place.
S is for Sign-off. Who last looked, and what they looked at. A check is only as good as the thing it was checked against. Rima verified the quote against the spreadsheet, which was accurate to what had been pasted into it. The sender took that as a check against the supplier. Both did their job and the gap sat between them.
To rate a carry, mark any letter that has gone missing in the last month. Then count the marks per carry. The carry with the most marks is the leaky one, with one override: any C mark on money or dates wins, whatever the count. A single lost expiry date will cost more than several garbled line breaks.
Fix One Seam, Cheapest First
Part 3 is to fix one seam, in this order.
First, a paste template. This is a fixed block of text that travels with the content: Source, Valid until, Customer, Checked by. It costs nothing, it takes ten minutes, and it works because it makes the missing thing impossible to leave behind. For Rima, the block goes at the top of every supplier summary, and the quote template loses its default expiry line so that an empty "Valid until" shows up as empty.
Second, if the paste block still leaves marks, give both tools one source file. If the supplier's validity sits in the same record as the price, nothing needs to be copied, so nothing can be dropped.
Third, and only third, a connector or an agent. Not because they are bad, but because they automate whatever the carry currently does, including the loss. If the seam is badly defined, a faster seam loses the date faster. Try the first two for two weeks; if marks remain, then you know what a connector must be told to preserve.
The tension I meet most is the instinct to buy. The tools have names, so they feel like the problem, and an "agent that does it all" promises one tidy answer. Sometimes that is right. But if I can find the damage in one carry on one sheet of paper, I would close that carry first and keep the budget for the day the map says a connector is warranted. The cheapest fix that closes the worst carry wins. Then re-walk the same job once and stop.
How to Draw Your Own Map in One Hour
You need one job you do at least weekly, the last finished example of it, and a sheet of paper.
- Pick one job done at least weekly, such as quotes, invoices, meeting follow-ups or product listings.
- Take the last finished example and find every tool it touched, in order.
- Write each hand-move between two tools as one line.
- Compare input and output on each line. What is missing: Format, Context or Sign-off?
- Put an F, C or S next to anything missing, and count the marks per carry.
- Choose the single carry with the most marks, or any C mark on a date or a price.
- Add a paste block (Source / Valid until / Customer / Checked by) at that carry only. Run the next three real jobs through it and recount.
This week, check the carry where a price or a date leaves one tool and enters the next.
Composite note: the importer, the operations manager, the supplier message, the contractor and every figure above are an illustrative composite built from situations I see in my work. They are not one client and not a measured result.
