UK geospatial intelligence company EvaluateLocate has launched Geospatial Predictor 2 (GSP-2), a predictive model designed to provide a continuously updated view of economic activity across the UK at one-kilometre resolution, as the company looks to expand its technology into the institutional investment market.
The model produces current estimates and forecasts for more than 130 economic, demographic and social indicators, with EvaluateLocate claiming it can predict economic trends up to three years ahead of official statistics. Gross Value Added (GVA), a measure of economic output closely related to GDP, is one of its core metrics, alongside employment, earnings, disposable income and residential property indicators.GSP-2 grew out of a problem the EvaluateLocate co-founders identified in commercial real estate: investment reports could contain extensive information about a building but relatively little quantitative analysis of the economy surrounding it. Their premise was that location itself represented a factor that could be measured.
From National GDP to Local Exposure
The model ingests official data, including statistics from the Office for National Statistics, then uses machine learning to fill gaps both spatially and through time. Rather than waiting for the next release of historical data, the aim is to estimate conditions today and their future direction.
EvaluateLocate is positioning that more localised view of economic activity for investors with geographically concentrated holdings.
“Companies have exposure to locations, or they’re thinking about changing their exposure to locations,” CEO and co-founder Adam Kirby tells Market & Alt Data Insight. “Those are the two categories. If you’re a house builder, you have a portfolio of land and projects, and you’re also looking at where to go next. If you’re an investment manager, you have a portfolio of exposure and you’re thinking about where next. Fund managers don’t have exposure simply to UK GDP. They have exposure to the catchment area for a shopping centre, for example, or whatever their business is. If you have a living model of every number, you have an advantage because you have a number for today, rather than last year’s number.”
Redrawing the Economic Map
Underpinning GSP-2 is a kilometre-level grid intended to remove another limitation of conventional economic statistics: their dependence on administrative boundaries.
Around 60,000 defined areas are currently available through EvaluateLocate’s API and can be aggregated for different requirements, with access to the underlying grid planned. For investment firms, that could allow economic geographies to be constructed around assets, company locations or catchment areas rather than relying on council boundaries or postcodes.
Says Kirby: “It should be more a function of the local economy than of the Local Government Act 1974 or the latest reorganisation. I was looking at the boundary between two London boroughs and it literally follows a Roman road. People are making economic decisions based on data published in tables that say one side of that road is one thing and the other side is something else. Because we have this continuous field of the economy across actual space and time, we don’t need to express it in the tables the data originally came in. The flexibility is there because we have the underlying grid.”
Accounting for Uncertainty
Predicting economic conditions between official observations inevitably introduces uncertainty. Kirby says GSP-2 correctly identifies the direction of travel 98% of the time at a one-year horizon, while the model also produces confidence ranges intended to show how much weight users should place on individual estimates.
“Inside the model we have uncertainty, and I think that’s really important. It’s an uncertainty-aware model, so under the hood it gives you a band. Take the Highlands Council area: it’s a relatively small economy spread over a huge area, and there isn’t much data. We think its economy was somewhere around £8 billion to £10 billion in 2023 pounds. But underneath that number, because the official data is years out of date, we can say we actually think it’s between £6.5 billion and £11.2 billion. We’re not simply projecting a straight line forward. There are bands, and that’s really important.”
EvaluateLocate is only now beginning to target the wider investment community, having gained its initial traction in commercial property, house building and other real-asset applications. It is also exploring expansion beyond the UK and the incorporation of Earth-observation data, including satellite-derived signals that could help tighten estimates where conventional economic data are sparse.
For institutional investors, the test will be whether that combination of timeliness and geographic precision can provide information that isn’t already captured in established economic and alternative datasets. Kirby’s ambition is straightforward: just as people consult a local weather forecast before deciding where to go, investors should eventually be able to consult an accurate local economic forecast before deciding where to put capital.
Subscribe to our newsletter
