International Location Data: Build a Coverage Matrix You Can Test

A country on a coverage map does not tell you whether a location feed can support the decision you need to make there. Before comparing international location data, ask for a dated evidence row for each market. The useful question is which observations survive your requirements, in the places and periods that matter.

Start with the decision, then define the row

Write the intended output above the spreadsheet: store visits, district-level flows, or regional trends. Add the geographic boundary, observation period, and reporting interval. A country-wide total cannot settle whether a rural catchment or a particular shopping district has enough evidence for your analysis. For smart-city planning, use the zones in the proposed study, including the areas where you expect coverage to be difficult.

Create a separate row when the source, delivery format, or observation rules change within a country. A national row can be a summary, but keep the underlying regional rows available. Do not combine a ready-to-load aggregate with an event feed that still needs visit attribution and call both the same deliverable.

The evidence matrix to send with a sample request

Use the fields below as columns in your evaluation sheet. They are a request template, not a published inventory of country coverage. Require a sample reference or a written definition beside the answers so another analyst can reconstruct the assessment.

Keep the sample filename, schema version, filter definition, and extraction date with the matrix. If the vendor substitutes another sample during evaluation, preserve both assessments. Otherwise a revised country total can quietly replace the evidence behind an earlier decision.

Count identifiers without turning them into people

Ask what the distinct identifier represents and over which interval it is stable. Daily distinct identifiers and distinct identifiers across the full evaluation period answer different questions. Adding daily counts counts returning identifiers again; it does not produce a deduplicated period total. Keep the interval and deduplication rule beside the result.

The UN mobile-phone-data handbook's data-quality chapter explains why subscriber, SIM, and handset identifiers in call detail records need not correspond directly to people. It also describes how operator coverage and phone usage affect representation. That discussion concerns network records; it is not a specification for an app-derived location feed. Use it to ask which unit and selection process apply to the sample in front of you.

Then request a reconciliation from the delivered sample to the records your analysis would use. Keep rejected records grouped by reason, such as missing fields or failure to meet your documented quality rule. Record where filters overlap so the same rejected record is not subtracted repeatedly. A large input file may still leave a thin usable sample in a priority region.

Inspect quality and gaps before ranking markets

For feeds with coordinates, ask what the accuracy field means, whether it is populated, and whether its definition changes by source. Android's Location documentation defines horizontal accuracy as an estimated radius with a stated confidence level. A coordinate alone therefore does not establish the quality of a location fix. Do not assume a vendor field uses Android's definition unless the vendor documents that mapping.

In the matrix, distinguish an interval that was delivered and contained no qualifying observations from an interval that was never delivered. Give suppressed results their own status too. An empty cell cannot tell an analyst which happened. Ask for the delivery manifest and suppression explanation before interpreting missing output as an absence of activity.

Separate event time from delivery time. Record the timezone and handling of late arrivals, then check whether rerunning an earlier period changes its output. Compare markets on the same reporting rules. A daily file schedule alone does not establish how recently its underlying observations occurred.

Test the output you intend to buy

Choose evaluation locations before seeing the sample, including dense areas, rural areas, and difficult boundaries relevant to your project. For visit analysis, compare against reference evidence you can obtain, such as site-operated counters. Document differences in what each source measures before judging agreement. The foot-traffic accuracy guide explains why panel and place-assignment choices matter.

Set acceptance rules for the intended use before the vendor returns results. A market might support a regional trend while remaining unsuitable for a store-level estimate. Record that narrower approval explicitly. Keep technical suitability separate from permitted-use review; the site's sensitive-location buyer guide addresses that additional diligence.

Suppose a retailer wants to compare store catchments across markets. The sample for one market arrives as visits assigned to store polygons, while another arrives as raw coordinates. Pause the comparison. Ask whether you can apply a common visit definition, which place boundaries were used, and how ambiguous assignments are handled. If the inputs cannot support that common definition, label the outputs as different analyses. Do not rank countries by the apparent visit totals. Keep the unresolved transformation in the matrix, with an owner and the evidence needed to complete it. Procurement can then separate a coverage gap from unfinished processing work in that market.

Turn the matrix into a buying decision

Mark a market as supported for the tested output, needing more evidence, or unsuitable for the current project. Retain the reason and the sample reference. Revisit the assessment when source composition, filters, or delivery change. This gives procurement a defensible scope instead of a country list that silently promises the same capability everywhere.

For a GSDSI evaluation, the Global Mobility & Location Data page is the relevant product starting point. Bring your target geographies, observation period, and intended output when you request a mobility sample, so the evaluation matrix addresses your actual buying decision.