I scraped 103 public prices to build an open dataset of website creation costs in France (CC-BY, DOI)

How Much Does a Website Cost?

I Published 103 Verified Price Points as an Open Dataset

“How much does a website cost?” is one of the most-searched business questions in every language.

And yet, almost every answer online has the same problem: broad marketing ranges, no sample, no dates, no sources, and no way to verify the numbers.

So I did what any data-minded person would do.

I collected 103 public, verifiable website pricing data points and published the whole thing as an open dataset — CSV, CC-BY 4.0, with a DOI.

This post explains:

• how the dataset was built;
• what the pricing data shows;
• what I learned from publishing a small open dataset properly in 2026.

The Problem With Website Pricing Content

Search for “website cost” in almost any market and you’ll find dozens of near-identical articles.

The pattern is always the same:

• huge price ranges;
• no methodology;
• no sample size;
• no collection date;
• no downloadable data;
• no way to check the sources.

For the French market — my playground, since I run a small web agency in Metz — the typical answer is:

“A website costs between €500 and €10,000.”

That is technically true.

It is also almost useless.

The missing piece was simple: nobody was publishing the actual data behind the claims.

That was the gap I wanted to fill.

Methodology: 103 Public and Verifiable Price Points

I only collected prices that anyone can verify publicly.

No private quotes.
No client invoices.
No proprietary agency data.
No “trust me bro” numbers.

The dataset includes pricing data from:

• Official pricing pages of website builders such as Wix, Squarespace, Shopify, SiteW, e-monsite, Hostinger and Webflow;
• French freelance day-rate barometers from Malt and Codeur.com;
• France Num’s pricing grid, an official reference from the French Ministry for the Economy’s SME digitalization program;
• Market guides published in 2025–2026 with explicit pricing ranges.

Each row in the CSV includes:

• service category;
• provider type;
• pricing model;
• minimum price;
• maximum price;
• typical price;
• currency;
• region;
• source URL;
• collection date.

Collection date: June 11, 2026
Total rows: 103
License: CC-BY 4.0

What the Data Shows

Here are the aggregate results, using the median of typical prices per segment.

Project type	DIY builder	Freelancer	Agency
Brochure site / vitrine	~€16/month	€1,450	€3,500
Brochure site range	—	€500–€8,000	€500–€15,000
E-commerce	~€41/month	€3,000	€11,000
E-commerce range	€5–€2,100/month	€1,500–€5,000	€3,000–€30,000
Custom web app	—	—	€10,000
Custom web app range	—	—	€4,000–€50,000

Two Findings That Stood Out

1. The Regional Price Gap Exists — But It Is Smaller Than People Think

One common belief in France is that Paris is dramatically more expensive than the rest of the country.

The data does show a gap, but not the cartoon version.

For example, experienced back-end developers charge:

• €592/day in Paris
• €513/day in Marseille

That is roughly a 13% difference.

So yes, Paris is more expensive.

But the data does not support the idea that “Paris costs double” as a general rule.

2. Total Cost of Ownership Changes the DIY vs Agency Debate

Sticker prices are misleading.

A DIY website builder looks cheap because the monthly fee is low.

Over 36 months, the estimated cost is approximately:

Option	36-month cost
DIY builder	~€576
Freelancer build	~€4,000
Agency build	~€6,100

On cash out, DIY wins.

But that ignores the hidden cost: your own time.

A DIY website can easily take 15 to 30 days of work once you include:

• setup;
• design decisions;
• content writing;
• SEO basics;
• mobile adjustments;
• integrations;
• revisions;
• learning curve;
• maintenance.

Once you value that time, the ranking often changes.

That total-cost-of-ownership calculation was the part readers found most useful.

Publishing the Dataset Properly

Releasing a CSV on your own website is easy.

Making it citable, reusable and discoverable takes more thought.

Here is the publishing stack I used.

1. Open License: CC-BY 4.0

I chose CC-BY 4.0 because it allows reuse while requiring attribution.

That is the right trade-off when you want other people to cite, reuse or build on your dataset.

2. DOI via Zenodo

I uploaded the dataset to Zenodo and generated a DOI:

10.5281/zenodo.20690911

This turns the CSV into a permanent, citable object.

If you publish a dataset without a DOI, you are leaving citations on the table.

3. Mirrors Where Data People Actually Look

I also mirrored the dataset on:

• data.gouv.fr
• Hugging Face Datasets

The goal was not just to publish the data.

The goal was to put it where researchers, builders, journalists and AI systems are more likely to discover it.

4. Schema.org Dataset Markup

On the landing page, I added server-side rendered Dataset JSON-LD with:

• identifier for the DOI;
• license;
• distribution with the direct CSV link;
• sameAs links to every mirror.

That made the dataset eligible for Google Dataset rich results.

And it worked: the page received the Dataset rich result within days.

5. Interactive Calculator Built on Top of the Dataset

A static pricing table is useful.

But an interactive calculator is stronger.

If a page only gives numbers, an AI summary can replace it.

If the page calculates a personalized estimate, users still have a reason to visit.

That is why the dataset is paired with a calculator rather than just a downloadable CSV.

Takeaways for Anyone Publishing a Small Dataset

You do not need a massive dataset to create something useful.

You need clean, verifiable, well-documented data.

The recipe is simple:

• Public pricing pages + freelance barometers + one institutional reference can create a legitimate market dataset.
• Small is fine. 103 sourced rows beat 10,000 unverifiable rows.
• A DOI, open license and mirrors can turn a blog CSV into a citable reference.
• Every price needs a collection date. Without a date, pricing data becomes noise.
• An interactive tool gives the page more long-term value than a static article.

Full Dataset and Calculator

The full barometer includes:

• charts;
• methodology;
• interactive calculator;
• downloadable CSV;
• source documentation.

You can find it here:

Baromètre des prix de création de site web — France 2026

The article is in French, and the dataset column names are documented in English on Zenodo.

Questions about the methodology, dataset structure or doing something similar in another market?

Happy to compare notes in the comments.