Nixeny Dijital - Mersin / Trade Geography 2026

Trade geography

The subject of this report: where businesses stand within the province and how uneven that standing is. What is measured is location; not turnover, employment or economic size.

Base

This report uses four separate bases: 19,615 businesses, 1,851 occupied grid cells, 1,997 trade clusters and the 226 cells a rate is calculated over. Every rate names which one it is over.

Grid cell

A square 500 metres on a side. The province is divided into a grid derived from where the businesses are; a square with no business in it enters no calculation in this study.

Footprint

The total area of the cells a group put a business in. Every density denominator in this report is a footprint — never a district's administrative area. Administrative area data enters this study nowhere.

Trade core

The centroid of the block made of the 500 m cell with the highest business count and its eight neighbours. It is not an administrative centre, a town hall or a “city centre”; it is the densest block in the sample.

Hotspot

The cells in the top 1% by business count. The threshold in this study is 138 businesses; 19 cells pass it.

Gini coefficient

Summarises how unequally a resource is distributed, between 0 and 1. The resource measured here is businesses and the side it is distributed across is the occupied grid cells: 0 means an equal number in every cell, 1 means all of them in one.

Nearest-neighbour index (Clark-Evans R)

How tightly a sector's points sit compared with the distance expected if they were scattered at random inside its own convex hull. R < 1 is clustered, R ≈ 1 random, R > 1 dispersed. The edge effect pulls R slightly upward, which works against the clustering claim; no correction was applied.

Moran's I

Measures whether neighbouring cells resemble one another. A value near zero means “being side by side says nothing”; a positive value means “similar cells stand together”.

Principal axis (PCA)

The first principal component of the business points: the direction in which the distribution is widest. In Mersin that direction coincides with the coastline, but it is not coastline data — it falls out of the distribution itself.

Single-link clustering

The rule that puts businesses within 150 metres of one another in the same cluster. It chains: if A and B are neighbours and B and C are neighbours, all three are in one cluster even if A and C are kilometres apart. That is why every cluster's span is published too.

Cluster span

The largest distance between any two businesses in a cluster. A cluster a few hundred metres across is a bazaar; one ten kilometres across is a town. It is the number to read before naming a cluster.

Isolated business

A business with no other business within a 1,000 metre radius. It is a definition of distance, not a measure of success.

Standard distance

How far a group's businesses spread, on average, from their own centroid. The standard deviation of geography.

Spearman rank correlation

Summarises whether two rankings move together, between −1 and +1. A value near zero means the two measures do not predict one another.

Own website

That the address on the business's Google profile is an independent domain rather than a platform page. The definition comes from the Mersin / State of Digital 2026 study.

There is not one boundary line on the map below. The shape of Mersin is a by-product of where the businesses are: each square is 500 metres of ground, and as dark as the number of businesses in it. Coastline, district borders, the road network and population data entered this study nowhere.

The first thing the map says is an absence. Of the 18,989 km² the businesses span, only 462.75 km² — 2.4% — holds at least one business. The rest is, as far as this sample goes, entirely empty. That does not mean “there is no trade there”; it means “none of these 72 business types is there”.

Half the businesses stand in 92 cells — 5.0% of the occupied cells, 23 km² in all. The median cell holds 2 businesses; the densest holds 398.

97.0% of the variance in the business distribution is explained by a single direction. The standard deviation of the spread perpendicular to it is only 10.4 km; along it, 59.1 km. Mersin’s trade is laid out not over an area but along a line.

In the ring within 1 km of the trade core there are 354 businesses per square kilometre. At 50 km and beyond, 31. But that outer ring holds 27.8% of the businesses — the density falls, the mass does not.

Putting businesses within 150 metres of one another in the same cluster gives 1,997 separate trade clusters. 1,325 of them (66.4%) hold a single business. The largest holds 7,864.

The median distance from a business in Mersin to its nearest neighbouring business is 18 metres. Scattered at random it would have averaged 492 metres; in fact it is 72.8 metres.

That figure is a floor and should be read as one. For some businesses Google Places gives not the front door but the centre point of a complex or a shopping centre; businesses in the same building can therefore appear at nearly zero distance. The right reading is “same building or next door”, not a metre-by-metre measurement.

Not every business type clusters the same way. Psychologists lean on one another (R = 0.076), customs brokers look almost randomly scattered (R = 0.844) — but the second is not dispersion, it is being squeezed into a single district: 88.4% of customs brokers are in Akdeniz.

The share of businesses with their own website is not scattered randomly across the map: neighbouring cells hold similar shares. That means knowing one cell’s share carries information about its neighbour’s — and says nothing at all about why.

The largest difference between districts is not in density but in digital presence. Between the densest and the sparsest district there is a factor of 7.1 in businesses per square kilometre of footprint; in the own-website share the gap is 29.7 percentage points.

Not one density number in this table was divided by a district’s administrative area. Had we done that, we would have shown the difference between Çamlıyayla and Yenişehir as hundreds of times larger, and what we measured would not be trade but where the district boundaries happen to run.

This report’s strongest-looking finding is also its most easily misread: as you move away from the core, the share of businesses with their own website falls. The rank correlation is -0.539 and the gap between the first and last ring is 17.5 points. That is a relationship.

The same pattern is produced by the composition of sectors alone, even if distance had no effect at all. In the centre there are lawyers, accountants and customs brokers; in the countryside car washes and tyre shops. Those two groups already differ in website ownership. Separating how much each carries would need the same businesses on two different dates, or a comparison holding sector constant; this study has neither.

A single cross-section cannot measure growth.

Whether the clusters have grown over time, or whether the core has shifted, cannot be tested with this data. A repeat of the study will produce the panel data that can measure it.

The sample is limited to 72 business types.

A cluster of a sector outside the scope does not appear on the map at all. An empty cell does not mean “there is no trade there”; it means “none of these 72 types is there”.

Dense areas may be saturated.

Places Text Search returns at most 60 results per query. In very dense areas the coverage is cut at that limit, which understates density — so the densest cells may be denser than what is written here.

Coordinate precision varies from business to business.

Businesses inside a shopping centre or an office block can fall on a single centre point. That pulls the nearest-neighbour distances down and makes differences under 100 metres uninterpretable.

The clusters have no names.

The clusters carry no neighbourhood or bazaar name. The largest are formed by chaining and represent not a bazaar but the unbroken trade fabric of a district centre; any naming must be verified by hand.

A convex hull is not a density denominator.

Google's district field misfiles a few businesses per district — a mobile roadside-assistance listing registered to Akdeniz appears 150 km west along the coast. A single point like that multiplies the hull tenfold; it adds one cell to the footprint. The hull appears in this report only as an indicator of spread.

District shares are not economic shares.

Because the same query effort was spent in every district, these numbers measure shares of the businesses in the sample, not the economic size of the districts.

The distance-to-digital relationship is not causation.

That the share of businesses with their own site falls as you move away from the centre does not show that distance lowers digital presence. The composition of sectors produces the same pattern: lawyers and customs brokers in the centre, car washes in the countryside.

These six rules are not a matter of style. Each is here because breaking it changes what the number means.

1. 01Density is always given with an explicit denominator; the province's area is never used.
2. 02Do not say “Mersin's commercial centre”; say “the densest trade block in the sample”.
3. 03The relationship between distance and digital presence is a relationship; no causation is claimed.
4. 04When districts are compared the base is that district's business footprint; it is neither its administrative area nor its convex hull.
5. 05Differences in distance under 100 metres are not interpreted; a Places coordinate is not building-precise.
6. 06The clusters carry no neighbourhood or bazaar name; any naming must be verified by hand.

1. 01UniverseThe data set of the Mersin / State of Digital 2026 study: the 19,615 local businesses visible as active on Google and carrying a coordinate. There is no record without a coordinate; coverage is 100%.
2. 02ProjectionA local plane projection (plate carrée), origin 36.668 / 34.292. All distances are in metres. Published distances are calculated with the haversine formula; the grid and the clustering use plane distance.
3. 03The gridA square grid 500 m on a side. The density denominator is the area of the occupied cells alone; Mersin's area enters no calculation. The hotspot threshold is the top 1% of businesses per cell: 138 businesses.
4. 04Clustering150 m single-link clustering. Single-business clusters are not counted as “noise” but reported as clusters of their own. Results for half the radius (75 m) and twice it (300 m) are published in Figure 12.
5. 05Sector thresholdSector comparisons are limited to the 68 business types in the study's own “headline” tier. Sectors with fewer points appear in the distribution figures but never become a headline at either end.
6. 06District assignmentThe administrative_area_level_2 field in Google's address components. The 130 records that could not be assigned were not estimated; they were kept in a bucket of their own, outside the district tables.

1. 01Google Places API (New) business profilesThe primary data source; name, address, coordinates, primary type (primaryType), business status and the website field. Fieldwork 2026·08·25. No extra API call was made for this report; every map was produced from the existing coordinates.developers.google.com
2. 02Nixeny Dijital · Mersin / State of Digital 2026The same universe, the same field date. The 19,615-business base, the district assignment and the definition of “own website” this report uses come from that study. There, location was an auxiliary field; here it is the subject itself.nixeny.com
3. 03External geodata: noneThis is not a source but the absence of one, and it stands here because it changes how the report is read. Population, district area, coastline and the road network do not enter this study. The silhouette of Mersin on the maps was not drawn; it is a by-product of where the businesses are.
4. 04The formulas usedGini: 2·Σ(i·xᵢ)/(n·Σx) − (n+1)/n over the ascending vector. Lorenz: 101 cumulative points in 1% steps. Clark-Evans: the ratio of the observed mean nearest-neighbour distance to the value expected under complete spatial randomness inside the sector's own convex hull. Moran's I: queen contiguity over the occupied cells. Principal axis: the first principal component of the coordinates. Clustering: 150 m single-link, plane distance.

No area, distance or density calculation in this report was compared against an external map layer. An axis that coincides with the coastline is a finding as long as it is not verified against coastline data; verify it and it becomes a circle. All that is shared with Mersin / State of Digital 2026 is the universe and the field date.

Conducted and published by Nixeny Dijital, Mersin. Data collection, projection, spatial statistics, map production and report design are Nixeny’s own. The universe and the digital presence classification are shared with the Mersin / State of Digital 2026 study. The maps are produced on the server for this page; no map service and no external layer is used.

The method, the raw data sets, the grid file and the verification records are shared on request. For corrections and questions: info@nixeny.com