Twelve months of profit and loss
Revenue, cost of sales and operating costs, monthly if you have it. This is the single most valuable file, because it sets revenue, margin and cost base, and every one of those otherwise arrives as a default.
Start
Build a twin of your company
Make an account, upload one profit and loss and one customer list, and run the question you have been putting off. The first twin takes about twenty minutes, most of which is waiting for you to find the files.
Your files are stored outside the media library on the server and are never shared between accounts.
What happens next
Three steps, in this order, and you can stop after any of them and come back.
- 1
You put in what you already have
Drag in the files, or paste text straight into the box if it is easier. Each one is read, cut into chunks and turned into facts with the line that supports them quoted. Files that will not read are told to you plainly rather than contributing an empty result.
- 2
It builds the twin and tells you what is missing
Facts become a graph, the graph becomes agents in nine classes, and every number ends up in one ledger with its origin. A readiness score shows how much of the model is standing on your documents rather than on industry defaults, and the missing files are listed in the order it is worth fixing them.
- 3
You ask the question and read the band
Pick a scenario or build one from the levers, run it, and read the result: a median with a tenth and a ninetieth around it, the events the agents caused and how often, the assumptions the answer rests on, and the earliest month the real world can tell you it was wrong.
Before you sit down
What to have ready
None of this is required. The first two make the difference between a twin built on your numbers and a twin built on industry defaults.
A customer list with revenue
One row per account with a revenue figure. Contract end dates and a segment column help a lot. Names can be replaced with labels: the model works on sizes and relationships, so Account A through Account X gives the same concentration analysis.
Your price list
What you charge and when it last changed. Pricing questions are the most common reason people arrive here, and a price history narrows the assumption that everything else hangs off.
Who you compete with
Even a list of names in a text file. Without it the twin invents competitors from industry priors and marks them as inferred, which is honest but weaker than the truth.
Headcount by department
People are capacity and people are cost. An org chart or a payroll summary sets both, and makes any question about hiring, cutting or restructuring worth asking.
One decision, written in a sentence
The thing you would actually do differently depending on the answer. Write it before you start. It is the difference between a good run and a beautiful simulation of the wrong question.
Before you upload anything
What happens to the files you are about to hand over
They are stored on the server in a folder outside the media library, under randomised filenames, with a deny rule so they are not served over the web. They are parsed into chunks so that every number can be traced back to the line it came from.
Nothing is shared between accounts. Nothing you upload is used to train, fine tune or evaluate any model, and there is no training pipeline in the product to do it with. The simulation itself makes no network calls at all.
Delete a source and the file, its chunks and the facts extracted from it go together, immediately. Delete the account and the twins go with it. If you would rather none of this touched anybody else's hardware, the same two plugins run on your own WordPress.
- Files outside the media library, with a deny rule on the folder
- One owner per twin, checked on every read and write
- No training on customer data, ever
- Deleting a source deletes the file, the chunks and the facts together
Rather look before you upload?
The worked example is an invented contract manufacturer called Harborline Components: 24 customers, one of them 18 percent of revenue, 31 percent gross margin, and a price unchanged since 2023. Every question on this site is already set up on it.