Skip to main contentSkip to main content

Vening™ Independent Depth-Model Assurance Engine

An assessment, not a reprocessing project.

You are not buying a six-figure reprocessing contract. You are buying a pre-registered decision on whether your depth model clears measured gates — with a signed receipt either way.

Gates
4pre-registered
Dispositions
REFUSE · PROMOTE
Discrimination
12.1× · 2.06 m floor
Last field run
REFUSED

The four gates

Cheapest refusal first.

Each gate answers a question that makes the next one meaningful. Any REFUSE stops the run before inversion spend.

GateQuestionPass means
01Instrument clearanceIs the acquisition outside the absorbing sponge?All sponge transits at unity; headers self-consistent.
02Blind-well illuminationDo withheld wells lie inside the recorded footprint?Every blind well inside midpoint coverage.
03Target depth reachableDoes the grid reach the deepest marker?Grid bottom below deepest pick TVDSS.
04Spatially blocked holdoutCan holdout shots be separated from training?Train–holdout separation and alternation above floor.

The floor

Every threshold sits beside a measured floor.

A requirement without a floor is a contract defect. Blind-well mistie uses a truth floor measured independently at checkshot-constrained picks — not a velocity-domain default that would read as vacuous in depth metres.

Pre-registration

Why the pass mark is frozen before we look

On our own reference survey the marker leg carried a bias of −25.69 m against an RMSE of 33.21 m, decided against a 25 m ceiling. Removing that bias — a defensible thing to do — drops the statistic to 21.04 m and turns REFUSE into PROMOTE. One analyst choice, made after seeing the number, was worth more than the inversion.

That is why criteria are pre-registered and digest-sealed before data is seen. Your team can run the measurement. Your team cannot freeze the pass mark before seeing the result, because your team is the one under drilling deadline.

What the data can support

The precision on a depth-converted volume is mostly assumed

Depth is a linear functional of a velocity column, and a surface reflection survey is far better at reflectivity than at the vertical average velocity that positions a marker. Measured on our own operator, with a perfect noise model and a perfect prior, a surface survey removes 3–6% of the depth-positioning variance. The resolution threshold is 50%.

So the interval on a depth model built from surface seismic is largely the interpreter’s prior, not a measurement. We say so, and we refuse to report a positional interval where the data cannot support one.

The receipt

Replay without our stack.

Artifacts carry bundle digests, pinned commits, and reproduce commands. Your auditor verifies signatures and re-runs the gates — not a PDF attestation.

Qualification

Is your survey eligible?

We need licensed field data with well picks and checkshots, in a region you specify. Reference runs on public benchmarks do not substitute for your acquisition geometry.

  • 01Pre-stack gathers with trace headers that agree with geometry
  • 02Blind-well picks with independent time–depth control where available
  • 03Explicit hold on turnaround promises — eligibility is decided by gates, not sales

Scope fence

What this is not.

  • Not a commitment to deliver a better image
  • Not an interpretation-suite workflow
  • Not a GPU reprocessing farm — contractors with thousands of GPUs already exist
  • Not a confidence percentage on a slide — REFUSE and PROMOTE are the only dispositions
Independent Depth-Model Assurance Engine

Request Assessment Session

Institutional buyers receive a desk-style walkthrough with signed lineage artifacts and offline verification steps — not a consumer demo.

Request Assessment Session

sales@seismicswiftai.com