A seventy percent decline is three different numbers depending on how it was defined. Type curves are quoted one way and modelled another more often than not.
Decline can be stated three ways, and they only agree when the curve is exponential, which a shale well never is.
Nominal decline is the instantaneous rate of change, the parameter the hyperbolic equation actually consumes. Tangent effective decline is the annual drop implied by that instantaneous rate if it never changed. Secant effective decline is the plain observed drop from the rate today to the rate twelve months from now.
Public type curves and investor decks almost always quote secant effective, because it is the one a non-specialist can check against a production plot. Reservoir software almost always consumes nominal. The conversion between them depends on the hyperbolic exponent, so it is not a constant you can memorise and it is not close to unity for a well with an exponent near one.
Type a secant number into a field expecting nominal and the well declines too slowly for its whole life. Every year of the forecast inherits the error, and it compounds into the tail where most of the remaining reserve sits.
Ask what basis the number is on. If nobody can answer in one sentence, the conversion was not done. This single question resolves the issue more often than any amount of re-derivation.
Recompute the first year. Take the fitted curve, integrate the first twelve months, and compare it against the first-year volume the type curve claims. If those two disagree, the parameter was entered on the wrong basis. This is a five-minute check and it is decisive.
Look for a conversion anywhere in the workbook. In most spreadsheets there is no conversion step at all. The number was copied from a slide into a cell.
Label the basis on every decline parameter, everywhere. Not in a
comment. In the field name. A parameter called Di tells you
nothing; one called Di_secant_effective cannot be misused
silently.
Convert once, at the boundary. Take the published basis in at the edge of the model, convert immediately, and work in nominal everywhere inside. Conversions scattered through a workbook get applied twice as often as they get skipped.
Test the first-year volume automatically. Any model worth keeping should assert that its own first-year output matches the type curve it was built from. That test catches this class of error permanently, and it catches it the day somebody changes the parameter rather than a year later.
An Asset Health Check finds this one on a producing asset in two weeks, fixed fee.
The Asset Health Check is the cheapest possible way to find out whether we are worth the larger engagement. Two weeks, fixed fee, and a written answer either way.