Fiscus — money auditing by Op2ra

The situation

Finding the savings is the easy half. Proving one happened is the other.

Somebody reports saving $40,000 last year. Against what? A saving is a difference from what would otherwise have happened, and if the counterfactual was never written down the number cannot be checked by anyone — including the person who produced it. Meanwhile three subscriptions have crept up 30% without a single approval, and two vendors are billing for the same job.

$ nullius spend statements.csv --claimed-saving 40000 money audit ============================================================== period 2026-01 to 2026-06 total observed spend 412,880.00 recurring vendors 31 committed monthly spend 48,210.00 resolution 18.00 — the smallest single charge; nothing below this is visible at all -------------------------------------------------------------- FINDINGS -------------------------------------------------------------- [CRITICAL] SAVING_WITHOUT_BASELINE: 40,000 claimed against no stated counterfactual [WARNING ] PRICE_CREEP: 6 charges up ≥10%; ≈31,000/yr [WARNING ] OVERLAPPING_SERVICES: 'analytics' has 3 vendors

What it checks

Four questions, asked in the same order every time.

Price creep on recurring charges — vendors whose amount rose 10% or more across the period, annualised. None of these increases required anyone's approval
Overlapping services — categories carrying more than one recurring vendor. Overlap is not proof of waste, but it is the shortest list worth reviewing
Committed spend concentration — what fraction of total spend is recurring — the part where negotiation has leverage and cancellation has immediate effect
The savings claim itself — whether a baseline was stated, whether the periods are comparable, and whether the claim is plausible against observed spend

What it does on real statements

findings on one 6-month set

of them about the savings claim

declared limits, always

calibrated thresholds — stated

The recurrence rule is deliberately crude: same merchant, near-monthly, amount stable within 35%. A tighter rule needs a cadence model, and cadence detection on six statements is guesswork. The honest move is a loose threshold and a declared limit, not a confidence score nobody calibrated.

What this audit cannot see

The section a competent buyer reads first.

Every report this product emits ends with its own version of this list, generated from the run rather than written by hand. A report cannot be constructed without one — the validator refuses.

whether a recurring charge is still earning its keep. Statements show payment, never usage — a tool nobody has opened in a year looks identical to one the whole team depends on
whether a saving actually happened. Statements can show that spend fell; they cannot show it would not have fallen anyway, and no amount of statement analysis closes that gap
whether cancelling is cheaper than keeping. Termination clauses, migration cost and the salary time to switch are not on the statement and routinely exceed the subscription
anything billed annually. Recurrence is detected from monthly cadence, and one or two charges inside a twelve-month window cannot be told apart from a one-off purchase without a contract to read. Measured: 0 of 641 annual contracts detected. It is usually the largest line
what the calibration covers. The measured part is the recurrence classifier — 100% of monthly subscriptions detected at 99.4% precision, against simulated statements rather than bank exports. The rest of this surface is deterministic: a baseline was either written down or it was not, and there is nothing there to calibrate
the deflated-Sharpe engine is not part of this surface. It returns no report, declares no limits, and none of the numbers above describe it

Engagements

Fixed scope, fixed price, and a report you can argue with.

Free

Browser audit

The documents surface, running in your browser — the real engine, installed into the page. Nothing is uploaded, because there is no server to upload it to. Or send one artifact and we will look at it: you get the finding either way, including if the finding is that nothing is wrong.

Fixed

One surface, two to three weeks

One product run against your artifacts, with a written report: measures, findings by severity, the resolution floor, and the limits. The price is fixed before the work starts and quoted from the size of your company, not from how the conversation goes.

Quote

Multiple surfaces

Where it gets interesting — the findings on one surface routinely explain the numbers on another.

The rest of the suite

Five surfaces, one decision procedure. Each deploys separately, so one product's failure cannot take another down.

Product	Surface	In one line
Nullius	documents	Your retrieval score is measuring your wording.
Auctus	growth	Most campaign wins are smaller than the experiment could see.
Rima	code	Two checks, chosen because they are high-precision and commonly missed.
Arbol	assistant	The layer that answers from your documents, and shows you which ones.