Shop analytics · Dashboards · AI access
Shop controlling without a controller.
Your figures sit in the shop, in Google Analytics and in your email tool. I bring them together every night, pseudonymised, and you ask them questions in plain language.
What the package contains
Refreshed every night
A pseudonymised extract of your shop database moves into a separate analytics database overnight. In the morning yesterday’s figures are there.
Six dashboards
Sales, year over year, repurchase and retention, abandoned carts, traffic and email, promotions and margin. In English or German.
Google Analytics alongside
Sessions, users, funnel and purchases from GA4, next to the actual orders instead of in a second browser tab.
Email next to revenue
Contacts, lists, automations and campaigns from your email platform, joined to the orders. ActiveCampaign today, other providers on demand.
Ask instead of click
Through the open MCP standard you ask your figures questions in plain language, in Claude or ChatGPT for example. Each person with their own login and permissions.
What the data says
A written review from me at onboarding. A dashboard shows numbers; this tells you what is in them.
Running and updates
I keep Metabase and the server current, watch the nightly run and deal with it when one does not complete. Included in the monthly rate.
Connect more sources
An ERP cost feed, a second email platform, marketplace revenue. €450 for an existing connector, €900 for a new one, plus €50 a month each.
What the dashboards look like
Six of them, in English or German, filled with your figures every night. Three of them here, with invented sample data.
Is revenue running ahead of last year, and what is driving it?
Revenue and orders over time, against the same month a year earlier. Beside it, which brand and which shipping country actually carry the number.
When do your customers come back, and which ones never do?
How the gaps between two orders are distributed, a median of 66 days in this example. Then the products worth a subscription or a reminder, because people reorder them anyway.
What do your discounts cost, and do they bring back more than they cost?
Every kind of promotion with its orders, revenue, discount given and gift value. Beside it, how many orders happen with no promotion at all.
What it costs
A fixed price for setup, then a monthly rate that is the same for every shop. Nothing here is tiered by revenue: a shop turning over three million needs the same pipeline as one turning over four hundred thousand.
In production since 2026 for an Austrian natural-products retailer, with around 190,000 pseudonymised addresses in the nightly run.
-
Call
€0Free, 30 minutes, no commitment. I look at what your shop runs on and which sources exist, and tell you honestly whether this is worth it.
-
Setup
€3,900 fixedAbout a week. At the end the dashboards run on your figures and you have the first written review. €3,120 with a 12-month commitment.
-
Running it
€299 / monthHosting, the nightly run, all six dashboards, all users, the AI access, updates and monitoring. The same price for every shop.
If you want more than the figures
What separates the steps is my time, and nothing else.
-
With quarterly reviews
€499 / monthRunning it, plus four written reviews a year: what stands out, what it means and what I would look at next.
-
With monthly guidance
€699 / monthPlus a call about the figures every month. Custom dashboards are available and quoted separately, because their size cannot be estimated honestly in advance.
Where the data actually sits
Most shops do not lack figures, they have too many places for them. Revenue in the shop, visitors in Google Analytics, campaigns in the email tool, cost prices in a spreadsheet. Any question touching two of those costs half a morning, and it ends with two people arguing about two numbers.
A controller would bring those together. That post pays for itself in very few shops, and off-the-shelf analytics software does not know your promotions, your lists or your margin. This package is the middle path: the consolidation as engineering, the reading as a service.
It runs on servers in Austria, and the customer data never leaves your own shop server in the first place. For a shop selling into the EU that is not a detail, it is the difference between a data processing discussion and none.
How setup runs
Four steps, about a week. After that it runs by itself.
-
Access
A read-only database user in the shop, SSH access for the nightly export, your Google Analytics property id and a read key for your email platform. Nothing beyond that.
-
Set up pseudonymisation
The export configuration is checked against your database, including tables your extensions added. Anything personal is replaced or dropped before it leaves your server.
-
Pipeline and dashboards
The nightly run goes live, Google Analytics and the email platform are connected, and the six dashboards are adapted to your lists, promotions and categories.
-
First review
You get a written read of the numbers from me: what stands out, why it matters, and what I would look at next if the shop were mine.
Common questions
Is giving an AI access to my shop data a security problem?
The analysis lives in its own database, separate from the shop. Pseudonymisation happens on your shop server before anything leaves it: email addresses are hashed, names and addresses replaced. The analytics server only fetches that finished export, read-only, from a single directory. There is no access to your shop database at all. Nothing can be changed either way: the database account has read permission only. What a person may see through the AI follows the permissions on their own login. What actually goes to the AI vendor is your question and its answer, so figures, not customer data. Whether that is acceptable is your call, the same way it is with your accountant.
Do I have to use Claude or ChatGPT for this?
No. The connection uses MCP, an open standard, over HTTPS. Which assistant connects is your choice, including one hosted in the EU or run by you. Claude and ChatGPT are the ones the vendor has verified; which client fits your setup we settle during onboarding. You can also skip AI entirely: the dashboards stand on their own.
Does this only work with Magento?
The data model underneath is platform-neutral, and five of the six dashboards read nothing else. They work with any shop system that supplies the data, as do Google Analytics, the email connection and the AI access. What is connected today, though, is Magento alone: for Shopware, WooCommerce or Sylius this would be a connector rather than a rebuild, but none of them is built yet. The exception is cart recovery, which depends on a Magento module and would have to be rethought on another platform. Shopify sits on someone else’s servers, where the export works differently in principle. Tell me what your shop runs on, and I will look at it and tell you what it takes.
What happens to my customers’ data?
It stays in the shop. Before the export, email addresses are hashed with a salt, names and addresses are replaced, and fields people habitually type personal details into are emptied. A check on both ends scans the finished export for email addresses; an export that fails it is never loaded. The result is pseudonymised, not anonymous: the hash is attributable to whoever holds the salt. The salt sits on your shop server and on the analytics server, and in no repository.
One Person, From the Question to the Code
I work out what is needed, write the requirement, organise whoever else works on it, and then build it myself. Usually that is three companies, and something is lost at every handover between them. Here there is nobody between your question and the running system.
Related services
If your problem sits somewhere else, these pages take it further.
Which number do you miss every month?
Tell me the question you keep asking and cannot answer today. On the call we find out whether this package answers it.
this page as a PDF to pass on