What a custom e-shop costs in Greece

An e-shop quote in Greece is a single number on the last page of a proposal, and it hides almost everything that decides what the store will cost. Two quotes for the same shop can be far apart and both be honest, because they are quoting different jobs. This piece is about where the money in a custom store actually goes, and how to tell which job a quote describes.

Begin with the part buyers see least: the catalogue model. Before a page is designed, someone has to decide how products are described — which attributes exist, which values are allowed, how variants work, which fields drive filters and which drive search pages. On a small shop this is a day. On a catalogue of thousands of products whose specifications currently live inside free text, it is a real piece of work, and it is the work that everything else is built on. A quote that does not mention it has either assumed the catalogue is clean or assumed nobody will ask.

Design and the storefront come next and are the most visible line. What moves the cost here is not the number of pages, because a store has a small number of page types repeated thousands of times: home, category, product, search, cart, checkout, account. It is how many of those are genuinely designed rather than inherited, and how much motion, film and photography the brand wants. A store that opens on a film per brand, as SIDEMILE's does, costs more to make than one that opens on a grid, and it should.

The checkout is a small number of screens and a large share of the engineering. Every payment method is an integration with its own contract, test environment and failure modes. Instalments, wallets, bank transfer, cash on delivery, invoices with tax details, VAT per product and shipping rules by weight, value or pickup point each add a branch. None of them is difficult alone; the cost is in making all of them agree, and in testing what happens when a payment fails halfway.

Integrations are the line with the widest range, because they depend on systems outside the agency's control. Connecting stock and orders to an ERP, couriers, a marketplace feed or an invoicing provider takes a few days when the other side has a documented API and someone who answers, and weeks when it does not. This is the main reason two honest quotes differ: one agency has seen your systems, the other has guessed.

Migration is the line most often left out, and the one that does the most damage when it is. Moving products and photographs is the easy part. Moving customers so they can sign in again, past orders under the numbers customers already know, and every old address so the search positions carry over is a project of its own. At Topharmakeio.gr that meant 9,804 products, 12,509 customers, 5,125 past orders and 3,175 redirected addresses, plus fixing the products that had been filed under the wrong brand along the way. A quote that says migration in one line has not looked at the old database yet.

Content and photography are the line that gets assigned to the client in a footnote. A thousand products need a thousand descriptions and photographs that match each other; categories and brands need copy of their own; every landing page needs a title and a heading that is not a duplicate of the next one. Where specifications are clean, descriptions can be generated from the facts of each product rather than invented, which is how Rellas has a written description for every watch and ring. Where photographs are missing or inconsistent, someone has to shoot them. Either way, a quote should say who does it, because if nobody does, the launch date moves.

Channels multiply the work quietly. A store that sells only to the public is one price list and one kind of account. Add partner shops, a wholesale price list applied automatically to approved resellers, a page that tells riders which shop near them stocks a product, a Google Shopping feed and marketplaces, and the same catalogue is now serving several audiences with different rules. SIDEMILE runs all of those from one catalogue and one dashboard. That is more work than a single storefront, and much less than running three systems side by side.

Then the work that makes the store findable: structured data, sitemaps, landing pages built from real combinations in the catalogue, and the redirects above. It costs less when it is designed into the catalogue model from the start, and a great deal more when it is bolted on after launch.

That is year one. Year three is a different calculation, and it is the one worth comparing. A store has running costs whatever it is built on: hosting, the engine or platform, payment fees, monitoring, the changes the business asks for, and the updates that nobody asked for but somebody must make when a provider changes its API or a security advisory lands. On a rented platform those arrive as a subscription and a stack of apps. On an open-source store they arrive as a maintenance retainer for keeping plugins compatible. On a store written once for one business, they arrive late and all at once, usually as a proposal to rebuild.

This is where the model we use changes the arithmetic. On NEXTCOMMERCE the storefront, the design, the content and the data belong to the business, and the engine is licensed as a service that we host, run, patch and back up. The year-three cost is a known licence plus the changes the business actually wants, rather than an unknown bill for keeping the store alive. Nobody pays to rebuild the checkout because a provider changed something.

So the useful way to read any quote, ours included, is line by line. Is there a catalogue model, and has anyone looked at the current data? Which payment methods are included, and are the merchant accounts already in place? Which integrations, and has the agency seen their documentation? What exactly is migrated, and are order numbers and old addresses on the list? What does the store cost to run in its third year, and who pays when something outside it changes?

And the limit we state on every page about this, because it saves some readers money. Below roughly a quarter of a million euros a year, with a small catalogue, no ERP and one courier, a rented platform usually wins the arithmetic, and we will say so rather than quote. Custom pays off when the store has become an operation, and at that point the expensive mistake is not the build. It is choosing a store whose third year nobody priced.