What an e-commerce project actually needs, and where the money really goes

An e-commerce project needs five things: a platform, an ordered catalogue, payments, logistics and measurement. It always gets complicated on the second and the fifth — the catalogue and the data — while quotes concentrate almost entirely on the first.

That is why so many projects overrun: not because the platform costs more than expected, but because nobody counted the work behind the product pages.

We are a communication and digital marketing agency in Italy, and we have worked catalogues at both ends of the range — from a few hundred products to the eighty thousand of a supermarket group. What follows is the part that repeats at every size.

1. The catalogue — where the real work is

An e-commerce is a catalogue before it is a website. And a catalogue ready to sell almost never matches what sits in the ERP.

It needs structure, not a list. Categories, attributes, variants, filters. If a customer cannot narrow by size, colour or compatibility, on a wide catalogue they get lost and leave.

It needs descriptions written for the buyer. The manufacturer’s spec sheet answers questions the customer is not asking. If the page talks about features while the search talks about problems, the page intercepts nobody.

It needs photography. The most underestimated cost line there is. On a wide catalogue, image production routinely exceeds the cost of the platform. In no kick-off meeting we have ever attended has anyone said “the risk on this project is the photography”. Then the photography arrives.

It needs a rule for the future. Who adds new products, with which mandatory fields, through which approval. Without that, catalogue quality degrades within a year.

How to size an e-commerce project, in practice: not by the number of pages, but by the number of SKUs and how ordered their data is. Two hundred clean products are a simpler project than fifty with chaotic variants.

We learned this at an unusual scale. In 2018 we worked on the online grocery service of Iper La Grande i, one of Italy’s large supermarket groups: over eighty thousand SKUs, effectively the assortment of an entire hypermarket, two years before the pandemic made food e-commerce obvious. Nothing about that project was hard because of the technology. Everything about it was hard because of the data.

2. The platform — the least dramatic decision of all

Contrary to how it is usually presented, the platform is rarely decisive. The mature options are all adequate; what differs is the cost and maintenance model.

Open source gives you total control and no licence fee, in exchange for owning updates, security and performance. Suited to those with a stable technical partner.

Subscription platforms remove maintenance and impose their limits. Suited to a fast start and a linear catalogue.

A custom build makes sense when the selling model is unusual — configurators, per-customer pricing, B2B logic — and no standard platform covers it without contortion.

The useful question is not “which is best” but: who will maintain it in two years, and what does that person cost?

3. Payments — simple, with two judgement calls

Technically the most solved part: gateways integrate in hours. Two decisions of substance remain.

Which methods to offer. Every added method costs fees and complexity, but removing one your audience uses costs abandoned baskets. Decide by looking at the audience, not the fee schedule.

How much to ask before payment. Forced registration before purchase is the single easiest cause of abandonment to remove. Guest checkout is a switch.

4. Logistics — the part that comes back to you

Shipping costs and return conditions are not operational details: they are part of the offer, and they are decided before opening, not after the first month.

The questions to close: who ships and in how long; what happens if the customer is out; who pays for returns; where returned goods physically go; who answers the customer when the carrier gets it wrong.

That last one matters most, because the customer will write to you — and the quality of that reply weighs on reviews more than any technical choice.

5. Measurement — postpone it and it is gone

Last on the list, and the first thing dropped when a project runs late. It is a mistake you pay for over years: data not collected cannot be recovered, and without data every subsequent decision is an opinion.

The minimum before launch: which products are viewed, which reach the basket, where the purchase breaks, which channel orders come from, what a customer is worth over time.

At scale it becomes a question of architecture. On the Iper catalogue we built the entire measurement layer in Google Tag Manager, without a single line of code from the development team — because the tagging could not depend on manual entry, or it would have survived neither the scale nor the turnover of the people loading products.

What changes across markets

If you sell in more than one country, three decisions move from the “later” pile to the “before opening” pile.

Catalogue structure per market. The same product can belong to different categories, carry different compliance attributes and answer different searches. A single structure translated into four languages is not a multi-market catalogue: it is one catalogue with four surfaces.

Search behaviour is not translatable. People in different countries search the same product with different words, different intent and different seasonality. A category tree optimised for one market can be structurally wrong in another — and this is discoverable in advance, from historical search data, which is a discipline in itself.

Who reviews the final copy. If nobody native to the market reads the product pages before launch, the catalogue will be understandable and unpersuasive. It is the cheapest quality gate available and the one most often skipped.

What it costs

The ranges are wide because the projects differ, but a quote is read the same way every time. What moves the total:

the number of SKUs and the state of the data — variable number one, and almost no quote quantifies it;
how many systems must talk — ERP, warehouse, invoicing, CRM: each integration is a project inside the project;
content production — photography, copy, video: if it is not in the quote, it will arrive later;
the first year of operation — an e-commerce opened and left alone does not sell. A launch cost without a running cost is an incomplete number.

Comparing two quotes honestly means checking they cover the same five areas. Most of the time, the cheaper one covers three.

Frequently asked questions

How long does it take to launch an e-commerce site?
With a ready catalogue and simple integrations, a few weeks. With a catalogue to rebuild and back-office systems to connect, months. The variable is almost always the catalogue, not the development.

Is it better to launch an e-commerce with the full catalogue or part of it?
Start with a narrow, well-executed scope: part of the catalogue, one market, one shipping option. Decisions taken on the first months’ real data are sounder than decisions taken on assumptions, and widening later costs less than correcting.

Does a B2B company need an e-commerce site?
Often yes, but with different logic from consumer retail: per-customer pricing, recurring orders, quote requests, negotiated terms. This is where customisation of the functions matters more than the design.

Does an e-commerce site also need to rank on Google?
Yes, and it must be decided before building it: category structure, URL patterns, handling of variants and out-of-stock products are SEO decisions disguised as technical ones. Fixing them after launch costs far more than doing them right.

Who runs an e-commerce site after launch?
It is the right question to ask any supplier before signing. If the answer is “you do”, that is fine — provided somebody has trained the people and the time required has been quantified, because a shop left alone does not sell.

---

We have designed and run e-commerce projects since 2006, including catalogues in the tens of thousands of SKUs, and we operate one of our own — because some things are only understood when the money being lost is yours.

→ Talk to us about your project