de
All references

[ geodata × sales containers ]

Where the next sales container pays off.

A sales container opens up new business: open around the clock, without staffing, placed where customers pass by. The only question is: where exactly? What emerged is not a static evaluation but a dynamic system: the search area is rastered without gaps, every candidate scored on six weighted criteria, and in the workshop the selection is condensed live with the customer on an interactive map.

Every spot, not just the obvious ones

The entire search area is split into cells of roughly 174 meters. No place is missed just because nobody thought of it.

Six criteria, defined

Passing footfall, distance to the business, parking, competitor density and substitutes, plus infrastructure and safety, each with a visible weight.

Live in the workshop

Search area, criteria, and street view in one map: the shortlist emerges in conversation, not in the follow-up.

Site decides the range

Every candidate gets a location type and, from it, a product-range recommendation. A commuter spot sells differently than a residential area.

[ the question before the investment ]

The most expensive mistake is the wrong site.

Staff is scarce, and classic retail space pays off ever more rarely in rural areas. Sales containers are one answer, also called smart stores, vending shops, or 24/7 stores depending on the format. Setting one up is manageable; choosing where decides everything: it determines whether anyone passes by at all.

Traditionally that decision comes from the gut: you know the area, you have a feel for the spot. The problem is not that the feel is wrong, but that it only tests the places someone thought of. A raster across the entire search area does not share that bias.

So this analysis scores every point in the radius by the same rules and makes the weighting visible. And because it is built as a system rather than a report, it does not end at handover: search area and criteria can be changed interactively and live on a map, and the ranking shifts before your eyes.

[ approach ]

From search area to a reasoned recommendation

Data space

Open geodata instead of purchased market research

The basis is open map and statistics data. Within the defined search area around the business, thirteen categories are collected: from supermarkets, bakeries, and gas stations through bus stops and stations to restaurants, schools, leisure grounds, and residential areas.

Two further sources join in: the official population grid for residential density, and Google Places, whose rating counts serve as a proxy for actual footfall. Alongside the pure location, that gives a signal for where people really go.

The search area is sampled twice: as a hex raster for the attractiveness landscape, and as a list of real siting spots, because a container needs space, power, and visibility.

Scoring

Six weighted criteria instead of one opinion

Every candidate is scored on six criteria: footfall, distance to the business, parking, distance to competitors, infrastructure, and safety. The weights sit in one place and can be discussed with the business.

With competitors the nuance sits in the detail: distance is good, because the offering should stand apart, but too much distance means no customers. Finding exactly that sweet spot was the real goal of the analysis: close enough to the footfall, far enough from the full-range store.

So that outliers do not distort the ranking, all criteria are normalized at the 5th and 95th percentile. The result is a score per candidate, broken down into all six parts: you can always see why a place ranks first.

Workshop tool

An interactive map, driven live rather than handed in

From its surroundings, every candidate gets one of five location types: residential, commuter, industrial, leisure, or shopping. The product-range recommendation follows from that, because a commuter spot calls for coffee and snacks to go, a residential area for the evening's shopping.

What is handed over is not a report but an interactive map: an attractiveness heatmap, switchable layers per data category, the best candidates as markers with all part scores, location type, and range suggestion. In the workshop everyone works it together until the shortlist is settled.

The analysis deliberately ends at a shortlist, not at a decision. What follows is manual work: visiting, approaching owners, clarifying power and permits. That is exactly what the list is short enough for.

[ lessons ]

Without AI tooling, this analysis would not exist.

That is the real insight from this project. Built the classical way, an evaluation of this depth would have been too expensive and too slow: the business would simply not have made the investment and would have picked the site by feel. With AI-assisted development, a level of analysis emerged in a short time that was previously unthinkable. New methods do not just get better that way, they become affordable in the first place.

And it did not stop at a static evaluation. What emerged is a dynamic system driven together with the customer in the workshop: change the search area, reweight or add criteria, compare candidates, and drive virtually through the streets via street view to condense the shortlist. Everything in one place, and whatever was missing could be added on the spot. That is a different workshop experience: decisions happen in the room instead of in the follow-up.

And it is not a one-off. Further sites can be identified with the same system, and for the next business with a similar plan, all it takes is swapping address, radius, and weights.

[ numbers ]

The analysis in numbers

154 km²

of search area are split without gaps into cells of roughly 174 meters and scored individually

6

weighted criteria make the score, calculated traceably and adjustable together with the business

13

categories of open geo and statistics data feed in completely, from bus stops through leisure grounds to footfall signals

5

location types, each with its own product-range recommendation: the place decides the offering

[ technology ]

Technology in use

Python
Jupyter Notebook
OpenStreetMap
Google Developer Platform
Destatis
Claude Code

Put the potential of this technology to work.

Are you looking for sites for sales containers, smart stores, or vending shops, and should the decision rest on data instead of gut feeling? The system carries your plan too. Then let's talk: you bring the challenge, I bring the experience and the tools.