A guide for e-shop owners
An e-shop that runs itself
If you already run an e-shop, you know how much work surrounds it. Copy, posts, translations, campaigns, evaluation - every month from scratch. This guide shows what the shop now handles on its own, where the limit sits and how much work falls away. And why you do not need a second invoice for marketing - we set up the accounts and the campaigns, and agents evaluate them.
30 minutes online. We go through your shop, I show you how the same thing runs in Ombra D'oro, and we work out what would pay off for you and what would not.
The starting point
The platform is not the problem. The work around it is.
Shoptet and Upgates do their job well: products, cart, payment gateway, done in a week. Nobody has to defend them for that. But a shop does not live off its shop window - it lives off someone writing descriptions daily, translating them, thinking up social posts, watching which keyword is burning money for nothing, telling a customer her size is back in stock, and once a month making sense of all of it.
That work does not disappear because you pay for a platform. Either you do it in the evenings, or you pay people to do it. And it is precisely that work which can now largely be handed to the system.
So the question is not whether you have "the right" e-shop. It is how much of that monthly work the shop can take over itself - and that can be measured in the hours it gives you back.
One ordinary week
What happened in the shop without anyone starting it
This is not a demo of what could be done. It is the schedule running in Ombra D'oro.
Google Ads handed over yesterday
A script inside the ad account sends the shop the spend, the clicks and above all the search terms - what people actually typed into Google. The data lands in the shop database, not in someone else's interface.
A new product translated itself into Czech
Name, description and page address. The translation applies itself and the old address lives on as a redirect so no links break. The Czech customer is not waiting for someone to find the time.
The stock watchdog sent an email
A week earlier a customer asked to be told when a sold-out size returned. The moment stock went from zero to one, she got the message - whether because a delivery arrived or because someone cancelled an order.
A reminder to those who were waiting on the price
An item in her favourites got cheaper than on the day she saved it - the system can compare because it remembered the price at that moment. One message is sent, not a series, and at night it waits until morning.
An approved post went out to Instagram and Facebook
The copy and the graphic were prepared by a model reading the real catalogue, and the owner approved it with one click. Sending is just a queue - and if the network rejects it, the post is not retried blindly.
A draft article joined the queue
The model looks at what is being written about in fashion, compares it with the last ten published articles so it does not repeat itself, and writes a draft into one of five sections. It does not publish it. It waits.
A customer asked what to wear to a wedding
Dora, the shop stylist, answered. She has no encyclopedia; she has a listing of the real catalogue with prices and stock, so she will not offer a piece that is not available in that size. And when the customer is signed in, she knows her taste as well.
Across that whole week a person did one thing: read the drafts and clicked approve. The rest is operation that does not ask.
The build
Five layers it has to have
Automation is not a feature you switch on. It is a consequence of the shop owning its data and something being able to read it.
A catalogue that knows itself
Size × colour variants, cost prices, size charts, photos tied to shades. Not to look prettier - without it the system cannot say what is genuinely in stock or how much you make on what.
instead of: a manual check before every campaign
Personalisation from actual behaviour
The shop remembers what you looked at, what you kept, what you returned and in what size. From that it builds a taste fingerprint that drives the selection, the reminders and what the stylist tells you.
instead of: "recommended" based on what other people bought
Content the shop reads
Articles, descriptions and posts come out of what is in the shop - the catalogue, the stock, the sales. The model does not invent facts, it reads them. Change a price and the next text already has it right.
instead of: a copywriter asking you for the details
Marketing over the same data
Ad spend and search terms flow back into the shop, where the orders already are. Only there can you join up that this term cost 40 € and brought nothing - while that one brought three orders for 12 €.
instead of: two interfaces that know nothing of each other
Evaluation that comes to you
Not another chart you have to open. An adviser goes through revenue, sizes running low, abandoned carts and recent customer questions, and writes what follows from it - including what it would need for a sharper estimate.
instead of: a monthly report nobody finishes reading
The difference
Personalisation that is not "you might also like"
Every platform has a recommendation block based on what other people bought. That is statistics about strangers. Real personalisation rests on what you know about the customer - and that is more than it seems.
Ombra D'oro builds a taste fingerprint from every visit: what the customer kept weighs most, what she favourited less, what she merely viewed less again. A returned item is information too. Alongside that the shop tracks which sizes actually fitted her in which cut - so on the next piece it pre-fills the right one and notes that M fitted last time.
How the customer sees it
Five picture questions in a style guide, and at the end she gets the name of her type. "Golden Shadow." "Quiet Elegance." There are sixteen and they are fixed, so she remembers hers and passes it on. Underneath, the selection simply shifts to match.
The same thing feeds a loyalty club with five tiers, where each tier changes the cashback, the free-shipping threshold and the returns window - and it feeds the stylist, who knows who she is writing to.
None of this requires anyone to sit at a screen. It requires the data to belong to the shop. On a rented platform your data is in someone else's format, and what can be read out of it is decided by someone else.
Content
A model that reads your shop writes better than one that reads the internet
The usual approach is to open a chatbot and type "write me a post about the autumn collection". The model does not know your stock, your prices, or what you wrote last week. It invents something plausible and then you correct it - which is extra work, not saved work.
The reverse approach: the shop gives the model tools with which it can ask. What the sales overview looks like. Which pieces sell most. Where content is missing. What the brand rules are. What has already gone out, so it does not repeat itself. The model then writes from reality rather than imagination - and you are not correcting facts, only tone.
Concretely in Ombra D'oro
Seventeen such tools, split across sales and stock, content, marketing and social. Exactly two can write - and both only create a draft. Only a person can publish it. Access is tied to the shop admin accounts, so removing an account removes it instantly, and personal data cannot get out through that gate even by accident: the model sees customers only as pseudonyms.
A side effect worth mentioning: content written this way is also easier to find. Not because of keywords, but because it talks about real products at real prices - and that is exactly what AI search engines look for when they assemble an answer.
Marketing
We set up the accounts and the campaigns. Agents evaluate them.
With the shop you do not get an empty ad account and a link to a manual. You get campaigns set up and running - and a system that then evaluates them against orders on its own.
An e-shop and marketing are not two orders. We do not deliver you a shop and leave you to find an agency for it. Accounts, campaigns, measurement and evaluation are part of the same delivery - and when we hand the shop over, the advertising is already running.
1 · The setup and campaigns we do
This is one-off work and we do it - not the system. We set up the accounts, the measurement and the campaigns themselves: structure, keywords, copy, negatives, budget. Most shops have half of this list done and do not know the other half is missing - and without it the advertising runs blind and the data automation is supposed to stand on never comes into existence.
2 · The agent prepares the campaign, you approve the budget
You do not get a list of keywords to rubber-stamp. You get the whole setup along with why it looks like this: campaign structure, keywords including negatives, ad copy, extensions, budget and bidding strategy - and with every decision, a sentence saying why it is that way.
It also includes what agencies say reluctantly: expectations up front. Roughly what a click should cost, how many clicks it takes to reach a first order, when it is to be evaluated and until when it is left alone. Without that you cannot say after a month whether it worked - you can only say "it still needs time".
What has to come first
In Ombra D'oro the setup came as one package - keywords, copy, negatives, extensions and a plan. Before launch, though, the measurement had to be sorted: the conversion tag was missing from the site, and without it the advertising would have run blind. That is the most expensive mistake you can make in Google Ads, and it is the most common one.
I will say this as the owner, not as the supplier: the campaigns for Ombra D'oro are set up such that I have barely clicked anything in the Google Ads interface myself. Not because I could not - because there was no need. On a consultation I will open that account and you will see from the inside what it looks like.
Matúš Benčo, Unoneto
3 · Reports that come to you
The campaign is not evaluated in Google Ads. Data from the account arrives daily in the e-shop, where the orders already are - and it is compared against the shop's actual orders, not the conversions Google counted for itself. Only there can you say whether the advertising made money.
The difference from an agency's monthly report is what it actually says. Not "we had 4,300 impressions and a 3.1% CTR", but "these three terms spent 40 € and brought nothing, let us add them to negatives" and "this is being searched for and you do not have it in stock". You can also ask outside the schedule, in an ordinary sentence.
Saying it straight
We are not a marketing agency and do not want to be one.
What an agency delivers - account setup, campaign management, copy, negatives and reports - we can deliver as a tool that runs on your side and asks your data. You pay for the setup once. After that you are not paying a monthly retainer to someone who opens the interface now and then.
We do not promise perfect results - nobody who has actually run advertising does. We promise something more verifiable: that you will see what you paid for and what it brought, in orders rather than in a chart.
The limit
What AI in an e-shop does not do - and will not
This is the most important part of the whole guide. Anyone who promises you otherwise either did not build it or does not have to run it.
A person always decides
- Money. Prices, discounts, campaign budgets. The model may propose and justify; a person has to set it.
- Publishing. An article or a post is created as a draft and waits in the queue. Approval is one click - but it is a person's click.
- Communication in a crisis. A complaint, a late order, an unhappy customer. An automation can send a confirmation, not an explanation.
- Changing the rules. What the brand says and does not say, who it sells to, what it cannot afford. That is your decision, which the system reads - not one it works out for itself.
A practical rule follows from that limit: automation is not worth switching on where a mistake would cost more than the time it saves. A product translation applies itself in Ombra D'oro, because the worst that happens is an awkward word that gets fixed. Sending a campaign never starts by itself.
A common question
Can part of this be done on the shop I already have?
Yes - and it is fair to say straight away which part. The marketing layer can be built alongside your shop. Personalisation cannot; it has to live inside it.
Marketing and content
Collecting Google Ads data including search terms, evaluating spend against orders, drafting posts and articles from your catalogue via a feed, a social queue with approval, and a monthly analysis with recommendations. Your shop stays exactly where it is.
Everything that knows the customer
The taste fingerprint, size memory, the stock watchdog, price-linked reminders, loyalty tiers that change shipping and returns, a stylist who knows who she is writing to. That cannot be glued on from outside - it has to sit in the stock, the checkout and the customer account.
So the sensible route is usually to start with the marketing layer over your current shop. It pays back quickly, breaks nothing, and after a few months there is something to decide on - including whether a shop of your own is worth it at all. If it is not, you have lost nothing and your marketing keeps running.
The numbers
How much work falls away
Not saved revenue - saved roles. These are the positions a shop like this would otherwise need in order to do the same thing by hand.
| The role you would otherwise need | What it would do | Average monthly cost |
|---|---|---|
| A person for content and the catalogue | product descriptions, 4 – 6 articles a month, translations of the whole catalogue | from 2 000 € |
| Marketing - a person or an agency | campaign setup and management, 20 – 30 posts including graphics | from 2 000 € |
| Evaluation and reporting | weekly review of terms and spend, monthly analysis with recommendations | from 500 € |
| Total | from 4 500 € |
This is a model calculation, not a measured saving. The figures are the average total monthly cost of a position - what such a person costs an employer, not their take-home pay; for an agency it is their monthly invoice. Your result depends on how much of it you actually do today. If you do none of it, you will not save money - you will gain work that currently does not happen in the shop at all. That is a difference you can count too.
Over a year that is tens of thousands of euros - from 54 000 € at this scope. This makes sense if you intend to actually earn money with the shop, not merely have one.
The route
How a shop like that gets built
Not all at once. Each phase has to make sense on its own, otherwise the project stalls halfway and you have nothing.
-
A shop that sells
Catalogue with variants, checkout, shipping, payments, comparison-site feeds. The dull but unskippable part - everything else reads from it.
-
Your own data and admin
Stock, orders, customers, carts including abandoned ones and where they came from. From this point the shop has something to read.
-
Personalisation and the club
Taste fingerprint, sizes, style guide, loyalty tiers. The first layer the customer genuinely feels.
-
The AI layer and automations
Tools over your own data, content drafts, translations, reminders, publishing to social. This is where time starts being saved.
-
Marketing: accounts, campaigns and evaluation
We set up and launch the ad accounts and the campaigns, their data flows back into the shop, and agents evaluate it against orders. This phase is part of the delivery, not an add-on for later.
Administration demo
Or click through it yourself
No need to call us first. We will send you a personal link to the administration dressed up as your own company - no sign-up, no password.
From your company ID we pull the name, registered address and business activities from the Slovak business register - your company appears in the header, on invoices and in e-mail signatures. Deals, contacts, invoices and figures are made up, so that you see the system running rather than empty tables. Everything is clickable.
It is not an e-shop, it is the administration around one: deals and contacts, invoices and reminders, cold outreach, e-mail campaigns, the social post queue, articles and statistics. You will also see what this whole guide is about - drafts that appeared on their own and wait for a human to approve them.
Nothing leaves the demo: e-mails, campaigns and posts are never sent, only recorded so you can see what would have gone out. The demo is valid for 30 days. For a company verified in the register the link arrives within minutes, otherwise once we approve the request, usually within 24 hours. The administration is currently in Slovak only.
I would be glad to show you live
Thirty minutes online, no product pitch. We open Ombra D'oro and I show you how it really runs from the inside - then we look at your shop and work out what would make sense for you, what can be added to what you already have, and what would not be worth it at all.
