April 2036: EO Became an Output Delivery Business

Reviewed and revised:

Maintained as a living reference by practitioners working across Earth observation data infrastructure, automation, and downstream operations.

It’s April 2036.

The industry finally stopped pretending the product is imagery.

Imagery is evidence. Sometimes it’s required. Often it’s useful. But it is not what most end customers actually need day to day. They need outputs they can action, at a cost base they can sustain, with a trust model that survives scrutiny.

So EO changed its default behaviour.

For the end customer, the deliverable is the result package: counts, detections, change vectors, tracks, confidence bands, timestamps, provenance. It lands as structured data, it plugs into workflows, it triggers decisions. That’s the product.

Imagery sits behind that, and it moved into a different role. It became a compliance artefact for audit, not the thing you hand to every user because it’s what we’ve always done.

The separation that made deletion possible

Back in the 2020s we used pixels as the audit layer. If someone questioned an output, we pointed back to the scene. That forced a heavy operating model: downlink everything, store everything, keep it “just in case.” Petabytes in, kilobytes out, with a human analyst as the conversion layer. Expensive, slow, and structurally incapable of scaling.

The fix was not better dashboards. It was separation of concerns.

Onboard processing became the default

Once that separation became real, onboard processing became default. A capture happens, processing runs in orbit, and the downlink is the output package. Raw imagery is retained only when policy requires it for audit, and only for the right subset of captures.

Most customers never touch the pixels. They don’t want them. They want the answer.

The audit layer matured beyond image-only trust

Imagery alone is not a trust framework. It’s just a file you can argue about. So every output began arriving with a verification record that is machine-checkable: model identity and version, onboard runtime hash, sensor state, calibration snapshot, inference parameters, confidence distribution, lineage to the capture event, and signatures.

The output became an auditable artefact on its own, without requiring an analyst to interpret the image as the source of truth.

Normal compliance in April 2036

Imagery is sampled and retained for audits against a published standard. Not indefinitely. Not everywhere. Not for every customer. It’s retained for verification cycles. It’s used to score operators against benchmark suites. It’s used by independent third parties to keep the standard honest and prevent the scoring system becoming a weapon.

Operators stopped being bought on claims and started being bought on audited capability. If an operator says they can count ships in coastal conditions, the statement is meaningless unless it comes with a task-specific accuracy score, tied to an evaluation suite, with audit dates and independent verification. Procurement treats these scores the way it treats revisit and latency: measured, comparable, and hard to game.

Why EODI was the prerequisite

EODI looked boring in the 2020s: common metadata, lineage, entitlements, dissemination, delivery states, audit trails. But without that structure you cannot distribute outputs safely, you cannot run audits across multiple operators, and you cannot compare results without rewriting everything per vendor.

EODI made outputs portable and made compliance enforceable. Once distribution and dissemination were standardised, the verification and audit layers could be standardised too.

So the end state in April 2036 is not “imagery is gone.” It’s sharper than that.

Imagery is no longer the default product. It’s a controlled compliance artefact used for standard audits. That’s what lets most workflows run on onboard outputs, and that’s what lets the industry delete raw captures in orbit for the majority of routine tasks without losing trust.

In 2026, everyone thought deleting imagery was reckless. In 2036, downlinking everything looks reckless. Different era, different constraint. The only way through was a standard that separated customer delivery from compliance evidence.

EO scaled when it stopped selling pictures and started delivering audited results.