We rebuild your assets as one auditable economic model, fit it to what they actually produced, and hand you a ranked list of what to do next with a price and a probability on every line. Not a dashboard. A decision.
Every operator already owns a model. It lives in a spreadsheet, it was built by someone who has since left, and nobody has checked it against what the asset actually did since the year it was made. The tools that would replace it are licensed by the seat, siloed by commodity, and priced for a major.
Gossans works the other way round. We bring the modelling engine, you bring the data, and what you receive is a study: a written report, the underlying model files, and a workshop with the people who run the asset. There is no seat to license and nothing to administer after we leave.
The engine handles the physics each commodity actually has. Wells decline on a hyperbolic curve nobody chooses; mines run at a rate somebody does choose, against a strip ratio that climbs; uranium leaches on its own kinetics years after the wellfield was paid for; helium is separated out of a gas stream that is declining underneath the plant. One cash-flow engine values all of them, which is what makes the comparison honest.
Fees are fixed on scope agreed before we start. Travel and third-party data are billed at cost. Multi-asset work is discounted against the portfolio rate rather than repeated per asset.
The most common reason a modelling project stalls is a data request that reads like a discovery motion. Ours does not. For a producing asset in the United States, roughly two thirds of what we need is already public, and we pull that ourselves before asking you for anything.
From you, once
We collect ourselves
Nothing leaves our custody. Data arrives under an NDA signed before the request goes out, is held in a single-client workspace, and is returned or destroyed at your instruction when the engagement closes.
The engagement below is a representative example, modelled on a six-well tight oil pad. The figures are model output used to show the method, not a named client's results.
Two CSV exports. Monthly oil, gas and water by lease with producing days and well count; monthly operating cost from the ledger. Plus one page of lease terms. Nothing else was requested.
The in-house model fitted a single decline curve to total field rate. Three more wells had come online in month twelve, so the rate rose in the middle of the history and no decline curve could fit it. The regression was reported without its quality measure.
Normalised rate per producing well, so adding wells stopped looking like a reservoir doing something impossible. Segmented the history at the last step up in rate, so the fit saw one curve instead of two vintages. Dropped the flowback month. Fitted on the logarithm of rate, because production spans two orders of magnitude and a fit on raw rate ignores the tail the reserve lives in.
The annual model charged drilling capital and first-year flush production to the same period, which flatters the cash flow a discount rate weights most heavily. Separating spud from first sales moved two numbers that matter to different people.
Four improvements were priced with the operations team and ranked on risked value. The one everybody wanted did not survive it. The one that ranked first needed no capital at all.
The gas lift conversion returned nothing because the recovery cap already bound the well: flattening the tail moved no barrels. That is not a judgement about gas lift. It is what this asset's own physics says, and it is invisible without a model that carries it.
These are the errors we find most often, in models built by competent people. Each one is invisible in the output and each one moves the answer in the same direction, which is toward approval.
Adding wells makes total rate rise. No decline curve can fit that, so the regression collapses onto a straight exponential and the recovery estimate follows it.
A "70% decline" is three different numbers depending on whether it is secant, tangent or nominal. Type curves are quoted one way and modelled another more often than not.
A well spudded during a year does not produce for that whole year. Charging both to period one moves cash into exactly the period a discount rate rewards most.
A pit deepens and its best ore goes first. A mine modelled at one strip ratio and one grade across its life is not conservative; it is describing a different mine.
By the time closure is paid there is no production left to depreciate it against. Pooled into capital under units of production, the deduction is stranded entirely.
A plant can be exactly the right size while something upstream is binding. Expansion capital is routinely proposed against the constraint that is easiest to measure, not the one that is holding the asset back.
Every asset team's best idea wins its own study, because none of them carry a cost or a probability. Ranked on risked value across a portfolio, the ordering usually inverts.
"We were eleven million light" points nobody at anything. Split into volume, price, cost usage and cost rate, the same eleven million names the department that can act on it.
What the asset is, what decision is pending, and whether we are the right people for it. We will say so if we are not.
Sent after the call, scoped to the specific asset. We start collecting the public half while you assemble yours.
The model is built and then fitted to what the asset actually did. Any gap between the two is reported before anything is optimised, because a model that has not been checked against outturn is an opinion.
Your operations and engineering people put a cost, a probability and a lead time on each candidate improvement. We supply the value; they supply the judgement. This is the step that makes the ranking defensible.
A self-contained report, the model files, the year-by-year cash flow as a spreadsheet, and a two-hour session with whoever needs to act on it.
The model stays yours. We re-run it against new actuals when it is useful to, and not on a calendar because a subscription says so.
A gossan is the iron-stained crust that forms where an orebody reaches the surface and weathers. It is not the ore. It is the rust the ore leaves behind, and for most of the history of mining it was the only evidence anyone had. A prospector who could read a gossan knew roughly what lay underneath, how deep, and whether it was worth the shaft, before a single metre was drilled.
That is the discipline this firm is named for. Your production history, your cost ledger and your public filings are the surface expression of an asset whose real behaviour is underground and cannot be observed directly. The data is already there, and it is already telling you something. The work is reading it correctly.
The name is plural because the work is. One reading tells you about one asset. The comparison across a portfolio, on a single consistent basis, is where the decisions actually get made.
The Health Check exists to be the cheapest possible way to find out whether we are worth the larger engagement. Two weeks, fixed fee, and a written answer either way.