Luno makes the analytic service, rather than the image, the product

Buying a satellite image leaves the customer to find the relevant change. Buying an analytic service asks the supplier to do some of that work and deliver a finding. The shift can reduce the volume an analyst must inspect, but it also moves part of the judgement into a commercial process the customer may not fully control.

NGA’s July 2, 2025 Luno update records delivery orders for facility monitoring, human geography and automated change analysis. Breaking Defense’s earlier task-order coverage describes the move toward commercial analytic services. These awards establish what the agency is purchasing. They do not publish comparative accuracy across the services or demonstrate that every automated finding is ready for an operational decision.

A finding has an information contract

Geospatial intelligence, usually called GEOINT, relates information to locations and activity on Earth. A commercial analytic service might count objects, flag a changed facility or describe movement. Each has a different burden of evidence. A useful purchase specifies what the output means and which original observations remain available to check it.

An automated count can be useful even when it is imperfect, provided the error is understood and the decision can tolerate it. A missed object, a duplicated object and an incorrect identity have different consequences. Combining all three into one score makes it difficult for the customer to judge the service’s suitability. Evaluation should connect the error to the actual question being asked.

Collection conditions matter too. A service can appear stable while the source imagery changes in angle, quality or time interval. An apparent change at a facility might result from those differences rather than activity. The supplier should preserve enough context to let the user distinguish a changed world from a changed observation. An attractive dashboard cannot remove that distinction.

The service also needs to explain absence. No alert could mean nothing changed, the location was not observed, or the available evidence was insufficient. Those states should not share the same quiet output. Otherwise a downstream analyst or software system may turn a gap in collection into unwarranted confidence that the situation remains normal.

Commercial delivery does not remove verification

The acquisition approach can expand available expertise and make it easier to purchase specialised analysis. The countercase is a service that returns findings faster than the customer can validate them. More output is not a net saving if every uncertain result requires a new manual reconstruction of the source record.

One proposed assessment would give each service comparable, previously unseen cases and follow the results through analyst review. Count missed changes, unsupported alerts and the time to resolve disagreement. The comparison should include the original observation, not merely an agreed answer sheet, because source availability determines how readily a finding can be verified.

Contract ceilings and delivery orders answer procurement questions rather than quality questions. A large overall buying vehicle can include several unrelated services and options. It should not be converted into a market-share or effectiveness claim without the relevant ordering and performance evidence. The useful unit here is the particular analytic task and its accepted output.

A model update can alter that output without changing the service’s name. Customers should retain the method or version, collection conditions and evaluation supporting each change. That makes it possible to determine whether a new alert reflects new activity or a different classifier. It also prevents a past acceptance result from becoming a permanent assurance about every later build.

When an alert becomes a service obligation

Consider a hypothetical customer monitoring a logistics facility. A service flags a sudden increase in parked vehicles. That is a useful starting point, but the customer still needs to know whether the previous image covered the same parking areas and whether the objects were present but obscured. The acceptance question is not simply whether the alert arrived. It is whether another analyst can reproduce the comparison and explain the remaining uncertainty without contacting the supplier for every case.

That changes how an analytic service should be bought. A delivery specification can require an observation interval, an explanation of the counted category and a route to the supporting scenes. It can also define what happens when evidence is unavailable. Those requirements are proposed purchasing choices, not additional terms disclosed in the Luno announcement. They make the distinction between an informative alert and an attractive but uncheckable notification concrete.

The larger consequence is that intelligence customers increasingly purchase part of their interpretation workflow. Competition then concerns more than sensor capacity or model sophistication. A supplier that makes disputed results easy to inspect can reduce the customer’s total burden, even if its first alert is less dramatic. Luno matters because the service contract is becoming one of the places where analytic accountability must be specified.