How a dossier's band is set, and what happens when it changes

Every dossier sits in one of four bands — standard, extended, complex, federal — and the band sets the price. This page explains how the band is decided, why it sometimes moves, and what a change means if you have already bought.

We publish this because a price tied to a classification is only fair if the classification is legible. You should be able to check ours.

What a band actually buys

A band is a statement about the market, not a grade for our work.

The dividing line that matters is not the count. It is the regional layer — state, province, or municipal closures on top of the national calendar. That layer is most of the work and most of the value, and it is what a complex or federal price is charging for.

How the band is computed

The band is generated from the data, not typed by anyone. It is recomputed every time the corpus is rebuilt, from the observance count and the presence of a regional layer, and written to contracts/country_classes.json in the public repository. The same figure prints on the dossier cover and appears on the invoice, so the two cannot describe different products.

There are two exceptions, and both lower a band rather than raise it:

1. A market with no regional layer is capped at extended, however rich its national calendar. Nepal has 36 national observances and no subdivisions; there is no regional layer to sell, so there is no complex price to charge. 2. A market whose regional layer cannot be evidenced is capped at extended. This one is a finding rather than arithmetic, so it is declared by hand in contracts/scope_points.json — each entry carrying a written reason and a reference into our data-quality register, where the work behind it can be read.

The second is the harder call and we would rather explain it than bury it. Indonesia is the current example. Its national calendar is one of the strongest we hold — 17 observances, every one resting on the retrieved three-minister decree. Its regional layer is 32 observances that are real, that people genuinely keep, and for which no government instrument sets a date: they are fixed by custom, by temples, by adat councils. We searched five regency legal repositories and read eight located documents before concluding it. A complex band would charge for a regional layer we cannot evidence, so Indonesia is priced as extended and the regional observances are included anyway, marked for what they are.

The subnational label

Every dossier cover states, in one line, what you get **below the national calendar**. There are three possibilities and the line is generated from the data, not chosen:

The third case deserves the explanation, because it is unusual and we would rather be asked about it than assumed about.

Indonesia's 32 regional observances are real days that people keep. What they lack is a government instrument that sets their dates — they are fixed by custom, by temples, by adat councils, and we established that by searching five regency legal repositories and reading eight located documents before concluding it. Iran's seven are unreachable for a different reason: every provincial host we can name refuses us, and part of the layer duplicates the national calendar.

That left three options and only one of them is honest:

• Drop the layer — removes true information from a dossier, to flatter a

statistic.

• Charge for it — implies an evidence standard those rows do not meet.
• Include it, mark it, and take it out of the price — which is what we do.

You are getting more than you paid for, clearly labelled as what it is. If those rows later gain instruments, the layer becomes priced and the market re-bands upward, under the rules in the next section.

When a band changes

A dossier is sold at the band it is in on the day you buy it. That is the product you bought and it does not get re-opened later.

We do not promise a band will never move. Markets change — a state adds a holiday, a regional layer turns out not to be evidenceable, a calendar we relied on stops being published. What we do promise is that a move is **predictable before it happens**, not discovered in an invoice: band changes follow from counts and evidence we track continuously and publish.

No refunds when a band falls. You bought a dossier at the price the market was worth at the time, and you received it. A later reclassification does not retroactively make that a different transaction.

No surcharge when a band rises. If a market grows and moves up a band, you are not billed the difference, and you receive the added coverage immediately rather than having it withheld pending payment.

Your renewal price stays where it was. This is the part most likely to be assumed wrong, so plainly: if you bought at extended and the market later becomes complex, you keep renewing at extended. The higher band applies to new purchases. We do not use reclassification as a mechanism for raising prices on existing customers.

In both directions the change is published. The band lives in a generated file in the repository with its full history, and the reasoning lives in the data-quality register. Nothing about a reclassification happens quietly.

What a band does NOT tell you

A band describes scope. It says nothing about how well-documented the dates are. That is a separate measure and it is printed on page one of every dossier — counted from the rows, never composed by hand — telling you how many of the dates come from the instrument itself, how many from independent secondary sources, and how many from published compilations.

A market can be federal and thinly documented, or **standard and fully documented from primary instruments**. The two axes are independent and we report both because collapsing them into one number would hide exactly the thing a careful buyer wants to know.

If you think a band is wrong

Tell us. The classification is generated from published data and the reasoning is written down, so a disagreement is a factual one and can be settled. If we are wrong, we correct it, republish, and apply the refund rule above.