Nixeny Dijital - Mersin / Digital Inequality Index 2026

Digital Inequality Index

A composite measure combining the results of the series' six studies into a single 0-100 score. It makes no new measurement; it weights six existing ones. It is at sector and district level and scores no individual business.

Column

Each of the six components of the index: Presence (25), Visibility (20), Quality (15), Trust (15), Access (15), Future (10). The weights are a value judgement and are set out in the configuration file.

Reach adjustment

Four of the six columns can only be measured on businesses that have a website. A sector where 3% of businesses have a site and those sites are perfect is not digitally strong: its quality reaches almost nobody. So the site-based columns are multiplied by that segment's website share before they enter the index.

Site-conditional value

The value before the reach adjustment: read only over businesses that have a site. It answers “how good are the sites that were measured”. The reach-adjusted value answers “how many businesses does that quality reach”. The two are never mixed in one sentence.

Absolute scale

Each column is a 0-100 quantity with a fixed definition, not a ranking normalised to this year's data. Normalised, someone would be guaranteed a 0 and someone a 100 every year, and the 2027 repeat could not be compared with this one.

Bottom third

The lowest third of the ranked segments on a column. A segment in the bottom third on 4 or more of the six columns is marked as “cumulative disadvantage”.

Cumulative disadvantage

The same segment falling behind on more than one dimension at once. That is exactly the question this report tests: is digital disadvantage cumulative, or does each dimension spread on its own?

Spearman rank correlation

Summarises whether two rankings move together, between −1 and +1. A value near zero means the two measures do not predict each other. In this report the correlations between columns are computed on the site-conditional values.

Gini coefficient (between segments)

Summarises how far the segments' index values spread, between 0 and 1. Note: this is not inequality between businesses. The Review Economy study's Gini of 0.843 measures the distribution across 16,124 businesses; the 0.182 here measures the spread of 60 sector averages.

Coverage

The share of the total weight made up by the columns that could be calculated for a segment. A column that cannot be measured is removed from both numerator and denominator; the remaining share is published as coverage.

Imputation

For the 13 segments with too few audited sites to calculate a site-based column from their own data, the province median was used. Their reach adjustment still reflects their real website share; the alternative — leaving the column blank — would have judged low-site sectors on fewer columns and pulled them artificially upward.

Province-wide index

Calculated by treating every active business as one segment. It is not the average of the sector indices and cannot be read as one.

This index makes no new measurement. It weights the published results of the series’ six field studies and combines them into a single 0-100 score. Every column in the table below names its source and links to that study’s own page: a reader who doubts a number does not have to trust this page, but can read the method at its source.

The weights are not a measurement result but a value judgement. Presence at 25 and Future at 10 is a choice; it is set out in the configuration file, and a different set of weights changes the ranking.

Four of the six columns can only be measured on businesses that have a website. That creates a problem: a sector where three percent of businesses have a site and those sites are flawless is not digitally strong — its quality reaches almost nobody.

So the site-based columns are multiplied by that segment’s website share before they enter the index. The Quality column falls from 67.3 to 16.4, Access from 71.5 to 17.4. The province-wide index is 23.2; the same calculation without the adjustment gives 48.8.

Both numbers are correct and each answers a different question. The site-conditional value says “how good are the sites that were measured”; the reach-adjusted value says “how many businesses does that quality reach”. Nowhere in this report are they used in the same sentence.

Across the 60 ranked sectors the index runs from Logistics company (42.4, 208 businesses) to Auto electrician (8.3, 250 businesses) — a factor of 5.1. The mean of the top five is 3.64 times the mean of the bottom five.

Those two ratios are not the same thing and should not be confused: the first is the ratio of the two extreme sectors, the second the ratio of two five-sector group means. Extremes are always sharper than groups.

What the sectors at the top of the ranking share is that they sell to businesses rather than to consumers: logistics, customs brokerage, software, international freight. That is not a coincidence, and it is the subject of section 07.

A ranking gives one number; a heat map shows where that number came from. Each row below is a sector and each column a measurement. A dark cell says the sector is strong on that column, a cell with a red underline that it is in the bottom third.

Colour is scaled within each column, not across columns. The columns were measured with different instruments — one is a website-ownership rate, one a median hygiene score — and a shared scale would only show that the instruments differ from each other.

The 60 sectors were split into five equal-count bands by index — twelve sectors in each. What varies is not the number of sectors but the number of businesses behind them: the twelve sectors in the bottom band hold 4,518 businesses, 23.0% of every business included in the study.

A lack of digital capacity is not a small minority’s problem: the bottom band holds twice as many businesses as the top one.

This study exists for one question: is digital disadvantage cumulative? If a sector falls behind on one dimension, is it behind on the others too?

The answer is no. The mean Spearman rank correlation across the fifteen pairs of six columns is 0.108 — near zero. The strongest relationship in the matrix is between Quality and Future (0.777), which is no surprise: both measure the technical state of the same home page. The second strongest is negative: Presence ↔ Trust -0.657.

That means there is no single axis called “digitalisation”. Sectors sit in different places on different dimensions, and a position on one says nothing about the next.

Even with independent columns, some segments can sit at the bottom on several dimensions at once. 19 sectors and 4 districts are in the bottom third on at least four of the six columns. Those sectors represent 7,073 businesses.

This list should be read with the finding above rather than against it: because the columns are largely independent, these segments are not an example of a rule but the tail where several independent weaknesses happened to land together.

One row in the matrix in section 06 behaves unlike the others: the Presence column is negatively related to four of the other five. The strongest is Presence ↔ Trust (-0.657), the most explanatory Presence ↔ Visibility (-0.332).

The reason is not that a website costs a business its reviews. Sectors with high website ownership mostly sell business-to-business — logistics, customs brokerage, software, international freight — and their customers do not leave Google reviews. Sectors with high review counts face the consumer: cafés, hairdressers, restaurants, dessert shops. Most of those have no site.

Two separate digital economies, with two separate gaps. That is also why a single “digitalisation” prescription does not fit both.

Between districts the index runs from Yenişehir (33.7) to Gülnar (10.5) — a factor of 3.2. The mean of the top three is 2.83 times the mean of the bottom three.

Against the three spatial variables the Trade Geography study measured, the relationship comes out strong: Spearman -0.765 between the index and median distance to the commercial core, 0.787 against business density, -0.900 against the share of isolated businesses.

This is a relationship, not a claim of causation. And it rests on thirteen points: even a coefficient that looks strong has to be read together with that sample size.

A composite index gives the impression that one thing is being measured — “digital capability”, and some sectors have less of it. This report’s own test says that is not the case.

The mean rank correlation between columns is 0.108 and the second strongest relationship in the matrix is negative. There is no single axis of “digitalisation”; there are six separate measurements and they move largely independently. The index itself also shows how much reducing those six to one number costs — which is why all six columns are published separately.

The index is at sector and district level. A sector scoring 8.3 is not a valid judgement about every business in it: the distribution beneath the sector is never shown in this study and no single business is scored. It does not say “the worst sector”; it says “the sector at the bottom of the index”.

The same holds for the geography section. Distance and the index move together; the sentence “distance lowers digitalisation” does not follow from this data. Causation was not tested, and cannot be tested on a sample of thirteen districts.

A single cross-section.

2026 data. The index's real value is in comparison: because the columns are defined on an absolute scale, next year's same calculation shows the change directly. On its own, this year says what today's distribution is.

The weights are a value judgement.

Presence at 25 and Future at 10 is not a measurement result but a preference. It is set out in the configuration file and can be changed and re-run; a different set of weights changes the ranking.

The site-based columns rest on the home page alone.

Quality, Trust, Access and Future all come from home-page measurements. What happens on inner pages is not seen; all four columns are therefore lower bounds.

There are 13 districts.

Every correlation at district level rests on thirteen points and is low-powered. Even a coefficient that looks strong must be read together with that sample size.

The reach adjustment rewards website ownership twice.

Website ownership is both its own column (Presence) and the multiplier on four others. The site-conditional index is the variant with that effect removed, and it is given beside every table.

Imputed values were used in 13 segments.

For segments with too few sites to calculate a site-based column from their own data, the province median was used. Those rows are shown faintly in the heat map and the affected columns are listed by name in the method section.

The source studies' limitations apply.

The sample is not a census, the sector assignment is a heuristic, and automated checks catch only part of the real barriers. This index removes none of those limits.

These six rules are not a stylistic preference. In a composite index, the wrong framing turns six separate measurements into a single judgement — and this report’s whole finding is that those six measurements do not say the same thing.

1. 01The index is at sector and district level; no business is scored.
2. 02Say “the sector at the bottom of the index” rather than “the worst sector”; the score is a composite of measurements, not a judgement.
3. 03Reach-adjusted and site-conditional values are never mixed in one sentence.
4. 04Every table names the source study of its columns; the index makes no new measurement.
5. 05The number of segments is small (13 districts); district comparisons come with a small-sample caveat.
6. 06The source studies' limitations apply to this index too.

1. 01SourcesThe published results of the series' six studies. No new measurement was made for this report and no request was sent to any site. The universe is the 19,615 active businesses in the 2026·08·25 dataset.
2. 02Columns and weightsPresence 25, Visibility 20, Quality 15, Trust 15, Access 15, Future 10 — 100 in total. Each column is an absolute 0-100 quantity; it is not normalised to this year's data.
3. 03Reach adjustmentThe four site-based columns are multiplied by the segment's website share before entering the index. The unmultiplied (site-conditional) value is also published in every table.
4. 04Ranking thresholdOnly segments with at least 100 businesses are ranked; a site-based column needs at least 15 audited sites. 60 of the 72 sectors and all 13 districts passed that threshold.
5. 05Imputed valuesFor the 13 segments where a site-based column could not be calculated from their own data, the province median was used; the reach adjustment still reflects their real website share. The alternative — leaving the column blank — would have judged low-site sectors on fewer columns and pulled them artificially upward.
6. 06Cumulative disadvantage testThe Spearman rank correlation of the 15 pairs of columns was computed on the site-conditional values. Had reach-adjusted values been used, the correlation would have come out of arithmetic, because all of them carry the same website-share multiplier. A segment in the bottom third on 4 or more columns was counted as cumulatively disadvantaged.

The correlations between columns were computed on the site-conditional values. Had reach-adjusted values been used, the relationship between them would have come out of arithmetic rather than the world — because all six carry the same website-share multiplier — and section 06’s finding would have come out the other way round.

This report’s sources are not external datasets but the series’ own studies. All of them were run on the same universe and the same fieldwork date, and all of them are published; every column’s number stands in its own report with its denominator beside it.

1. 01Nixeny Dijital · Mersin / State of Digital 2026The source of the Presence and Quality columns. Presence: the share of businesses with an independent website of their own, over 19,615 businesses. Quality: the median Digital Score of the 4,167 audited sites.nixeny.com
2. 02Nixeny Dijital · Mersin / Review Economy 2026The source of the Visibility column: the share of businesses with at least 10 Google reviews, over every business.nixeny.com
3. 03Nixeny Dijital · Mersin / Cyber Hygiene & KVKK 2026The source of the Trust column: the median Cyber Hygiene Score of the audited sites — encrypted connection, security headers, cookie and consent configuration.nixeny.com
4. 04Nixeny Dijital · Mersin / Accessibility 2026The source of the Access column: the median Accessibility Score of the audited sites — alt text, form labels, structure, link text, media.nixeny.com
5. 05Nixeny Dijital · Mersin / AI Readiness 2026The source of the Future column: the median AI Visibility Score of the audited sites — access permission, structured data, content depth.nixeny.com
6. 06Nixeny Dijital · Mersin / Trade Geography 2026Not a column of the index. It is the source of the geography comparison in section 08: the districts' median distance to the commercial core, business density per square kilometre and the share of isolated businesses all come from that study.nixeny.com
7. 07Weights and composition formulaIndex = the weighted mean of the columns. Site-based columns are first multiplied by the website-ownership share. A column that cannot be measured is removed from both numerator and denominator, and the remaining weight share is published as coverage. The weights are set out in `config/digital-index.json`; to see how a different set of weights changes the result, change the file and re-run the command.

The absolute value of this index cannot be compared with another province’s index — unless it was measured by the same method, with the same weights and the same column definitions. The whole series is on the Research page.

Conducted and published by Nixeny Dijital, Mersin. The composition of the index, its weighting and the report design were carried out by Nixeny. This study makes no new measurement; it combines the published results of the series’ six field reports and names the source of every column.

The weights file, the intermediate calculations and the full segment-level tables are shared on request: info@nixeny.com