Healthcare services
Highest switching cost and near the top on regulatory exposure at 0.85, with contracts averaging eighteen months. Decisions take a long time to show up.
By industry
Where the model starts before it has read anything of yours
A twin has to run on day one, including the parts your files never mention. So every twin begins from a table of industry starting points: a gross margin, a churn, an elasticity, a competitor reaction and ten more. They are prior beliefs, they are marked as prior beliefs everywhere they appear, and each one is replaced the moment a document of yours says otherwise.
Build a twinHow the ledger works
Four industries are written up in full below. The product carries twelve.
What an industry prior actually is
The least interesting and most important number in the system.
When the engine needs a value it does not have, it takes one from this table rather than stopping or inventing something. A manufacturer with no churn history starts at 1.2 percent a month. A software company with no elasticity in its files starts at 1.1. The run then proceeds normally, and the ledger records that the number was a default.
Three things are true about every one of them
- It is a starting point, not a finding. It is not a benchmark, it is not what companies in your sector achieve, and it is not evidence of anything about you.
- It is replaced, not blended, the moment a document supports a real number. There is no weighted average of your figure and the default.
- It is drawn with a band on every replication, and the bands on the priors are wide on purpose. Elasticity is drawn within forty percent of its value, churn within thirty five percent, monthly market growth within fifty percent. Gross margin, which your accounts usually pin down, is drawn within fifteen.
That last point is the whole design. A prior with a wide band is a statement that the model does not know, and it produces a wide answer, which is the correct answer to give somebody who has uploaded very little. As documents arrive, values get replaced and bands narrow, and the band on the final number tightens because of evidence rather than because of confidence.
Where they came from
Ranges commonly reported in public benchmark surveys and in textbook industrial organisation, rounded to one decision. They are not derived from any customer of this product, there is no proprietary data set behind them, and no claim is made that any particular company sits near any of them. The honest description is: plausible, defensible, and the first thing you should replace.
If a page on this site ever reads like "companies in this sector achieve X", it is a mistake. These numbers are the model's ignorance, written down in a readable form so it can be corrected.
Four industries, side by side, on the five priors that decide the most.
The full table for each industry, all thirteen values, is on its own page. This is the short version, and it is worth reading because the differences between the columns are the reason a price rise is a different decision in a factory than it is in a shop.
| Starting point | Manufacturing | Software and subscription | Professional services | Retail and ecommerce |
|---|---|---|---|---|
| Gross margin | 32 percent | 78 percent | 45 percent | 42 percent |
| Monthly customer churn | 1.2 percent | 1.5 percent | 2.5 percent | 6 percent |
| Price elasticity of demand | 1.8 | 1.1 | 1.4 | 2.2 |
| Revenue under contract | 55 percent | 62 percent | 35 percent | 5 percent |
| How hard competitors match a move | 0.55 | 0.45 | 0.35 | 0.70 |
| Months before competitors react | 4 | 3 | 2 | 1 |
| Regulatory exposure | 0.45 | 0.15 | 0.20 | 0.18 |
Elasticity is quoted as a magnitude; the engine carries it negative. Competitor reaction, switching cost and regulatory exposure are scores from zero to one. Every value here is a default that your own documents replace.
ManufacturingThin margin, most of the cost bought in, customers on long contracts and competitors who move slowly but hard. The industry where supply is usually the question.Software and subscriptionMargin is not the constraint, retention is. Low churn compounds, contracts delay everything, and the cost of winning a customer takes the better part of a year to come back.Professional servicesCapacity is people and people are the cost. Short contracts mean a price decision reaches the whole book quickly, and competitors move fastest of the four.Retail and ecommerceAlmost nothing under contract, the highest elasticity of the four, and a competitor who matches you within a month. Everything happens quickly, including the mistakes.
What replaces them
One document is worth more than the whole table
A profit and loss replaces gross margin, cost per head and monthly operating cost immediately, with the line quoted. A customer list with dates replaces churn, revenue per customer and both concentration figures. A set of contracts replaces revenue under contract and average months remaining, which are the two numbers that decide how quickly any decision reaches your customers.
The readiness score on the overview is exactly the share of the model still standing on this table. It is not a grade, it is a queue: the missing documents are listed in the order that replacing them would narrow the answer most.
- Financial statements replace the money assumptions and narrow their bands
- A customer list with start dates and values replaces churn, contract position and concentration
- Contracts replace locked share and months remaining, which drive the timing of everything
- Supplier agreements replace cost share, concentration and lead time
The other eight
Not written up, but carried in the product
These have the same thirteen starting points and behave the same way. They do not have their own page yet because there is nothing useful to say about them that the table does not already say.
Financial services
Regulatory exposure 0.90, churn 0.8 percent a month, twenty month contracts. The slowest moving twin in the set.
Construction and trades
Twenty four percent gross margin with fifty eight percent of cost bought in, which makes supplier moves land almost immediately.
Logistics and transport
Twenty two percent margin, sixty six percent supplier cost share, elasticity 2.0. Very little room between a cost rise and a price rise.
Hospitality and food
Eight percent monthly churn, almost nothing contracted, and competitors who react inside a month.
Education and training
Ten month contracts, 3.5 percent monthly churn and slow competitor reaction at four months.
Nonprofit and public
Elasticity 0.6, the lowest competitor reaction in the set at 0.15, and fifteen month agreements.
Something else
A deliberately middling set for companies that do not fit. Wide bands, and a strong reason to upload more.
Questions about the priors
My industry is not in the list.
Pick the closest, or pick "something else", which carries a middling set with wide bands. Then replace the values that matter with your own, which you would be doing anyway. The industry choice matters most on day one and matters less with every document you add.
What if the default is obviously wrong for me?
Change it. Every prior is an editable assumption with an editable band, and the edit is recorded against your name. If you know your elasticity is nothing like the default, setting it correctly is a two minute job that changes every run afterwards.
Do these numbers say anything about how my sector performs?
No, and it matters that they do not. They are rounded ranges taken from public benchmark surveys and textbook industrial organisation. They are a way of not stopping when a number is missing. Treating them as a benchmark to measure yourself against would be a misuse of them.
Do two twins in the same industry share anything?
Only this table, which is public and sits in the product's source. Nothing is shared between accounts, nothing is learned across customers, and no twin is influenced by any other twin.
Start on the defaults, then replace them
Pick your industry, upload one profit and loss and one customer list, and watch how much of the table disappears in the first ten minutes.