Reading geolocation-accuracy claims in commercial imagery

A commercial imagery product can be sharp, attractive and consistently misplaced against another map. Image detail and positional accuracy describe different qualities. The first concerns what the data can distinguish; the second concerns where the represented information sits relative to a stated reference. A supplier comparison needs both where both matter to the customer's product.

The phrase geolocation accuracy is therefore incomplete without an explanation. A buyer needs to know the metric, the reference used to assess it and the population of observations behind the result. Those details determine whether two proposals make comparable claims and whether either supports the customer's intended map or analytical service.

A number needs its full label

USGS describes Landsat Collection 2 Tier 1 imagery as meeting prescribed image-to-image georegistration tolerances of no more than 12 metres radial root mean square error. The wording identifies a product class, a relationship between images and a statistical measure. Removing those qualifiers changes what the figure says.

In particular, an RMSE value is not a promise that every point lies within that distance. It is a summary of an assessed set of differences. A purchasing comparison should retain the name of the statistic instead of rewriting the result as an unqualified maximum error.

Other proposals may use a different metric or express error in pixels rather than metres. The customer should ask the provider to explain the definition and basis. Converting between unlike measures without the necessary assumptions can produce an apparently neat comparison that the evidence does not support.

The useful commercial record is short: named metric, units, assessment basis, product version and stated limitations. Technical specialists can examine the underlying method while the purchasing team preserves the meaning in the order and acceptance documents.

Identify what the imagery was compared with

Positional agreement is always assessed against something. An image may be compared with a geographic reference, another image or a set of documented control observations. Each comparison answers a particular question. Agreement with one source does not automatically establish agreement with every other dataset the customer uses.

Consider a hypothetical company providing environmental summaries for industrial properties. Its customers supply property boundaries from several systems. A consistent imagery archive may support comparisons through time, yet a boundary overlay can still show a mismatch. The company needs to investigate the relevant references rather than assuming the latest image is wrong or the customer map is exact.

A data proposal can clarify whether the quoted result concerns absolute position, relative alignment or another defined comparison. It should identify the reference dataset or assessment method at a level sufficient for the buyer's quality review. Where access to the full reference is restricted, a scoped assurance statement may be useful, but the restriction should remain visible.

This distinction also affects the value of a paid preparation service. A supplier may be improving alignment with a particular customer reference rather than changing the original sensor's performance. That is potentially valuable work. It should be described and priced as the actual service delivered.

Spatial coverage of the assessment matters

USGS's Geometric Verify Image and Report documentation provides a concrete example. For specified Landsat MSS and TM Tier 1 products, the verification material compares the scene with a reference dataset and reports both quadrant and whole-image summaries. USGS warns that places without measurements have not received an accuracy assessment in that record.

The purchasing inference is that a scene-wide statistic should be read alongside the distribution of its evidence. A dense set of measurements in one part of a scene does not establish an equally detailed assessment everywhere else. The report's gaps are information about the scope of the conclusion.

A vendor can communicate that scope through a summary of assessed coverage and exclusions. The customer need not receive an overwhelming amount of technical material at the first meeting. It should nevertheless be possible to discover whether the relevant area was represented in the evidence used for the claim.

This is especially useful when comparing two evaluations with different selection rules. One may report only cases that passed its internal quality screen. Another may describe the complete set of attempted assessments. Their numerical summaries are not directly comparable until the customer understands those populations.

Keep product processing in the comparison

An imagery supplier may offer several processing levels or a prepared analytical product. The positional characteristics can depend on which version is being delivered and the reference information used in preparation. A family-level performance statement should therefore be connected to the actual product on the order.

USGS's tier structure illustrates that product selection can preserve meaningful quality distinctions. The practical lesson is not that a commercial supplier should copy Landsat's categories. It is that different delivered products deserve different evidence where their characteristics differ.

A customer can ask whether the demonstration sample, the bulk archive and the ongoing subscription use the same supported preparation. If they do not, the supplier should explain the difference. A carefully prepared example can demonstrate an available capability without proving that the same work is included in every standard delivery.

The guide to SAR resolution and pixel spacing covers another specification distinction. A fine output grid can make an overlay look precise without establishing the positional accuracy of the underlying data. Keeping grid spacing and geolocation evidence separate prevents that visual impression from becoming a contractual assumption.

Ask what uncertainty the downstream product can tolerate

The right purchasing threshold follows from the customer's output. A broad regional summary, a visual reference and a detailed comparison against a customer dataset may need different levels of positional assurance. Requiring the strongest available number for every use can add cost without improving the delivered decision.

The product team should state what would become unreliable if the imagery were misaligned by more than the acceptable amount. That turns a technical preference into a testable business requirement. It also helps determine whether the existing data are sufficient, whether additional preparation is needed or whether a different product is required.

An early evaluation should preserve unresolved cases instead of hiding them within a successful average. If a supplier cannot establish suitability for a particular subset of the customer's inputs, the team can restrict that part of the offer or commission further evidence. The scope then expands when the supporting work exists.

This approach is useful for sales teams as well. A bounded claim is easier to defend through delivery than a universal accuracy statement copied from a sensor brochure. It also gives the company a clear explanation for the cost of customer-specific preparation.

Preserve the assessment through data updates

Historical imagery can be reprocessed with revised reference information. A customer may welcome the improvement while still needing to explain why a later delivery no longer aligns exactly with an earlier one. The product record should identify the versions used in the original assessment and the subsequent change.

The commercial service should say whether reprocessed data replace existing files, arrive as separately identified versions or require a new purchase. It should also explain what notification the customer receives. These details affect downstream reproducibility and support, especially when an earlier report must be reconstructed.

Our BlackSky company profile and reporting on commercial radar augmentation awards provide context on imagery businesses and buying activity. The technical purchase needs a narrower conclusion: which delivered product has which documented positional characteristics, against which reference, and for which assessed customer use. That is the foundation for an accuracy claim a commercial team can stand behind.