Mirakls reads your orders, your reviews and your rivals' shops and ads, then builds 500 synthetic shoppers from your real customers. Try a price change, an offer or an ad on them first. Who buys. Who balks. Why. Before you change anything in your store.
Raise prices 10%. Lift the free-shipping threshold. Run the Black Friday discount. Start selling into Germany. The changes that move your revenue are exactly the ones you can't A/B test: every shopper sees the new price at once, and nobody un-sends a discount.
So they get decided the old way: a gut feeling, a competitor's move, or the loudest voice in the room.
Give us your website and connect your store. Mirakls reads your orders, your reviews, your rivals' shops, reviews and ads, and the countries you really ship to. Then it tells you what it found, before you ask anything.
Synthetic customers shaped by your own orders and analytics: the countries they buy from, their ages, what they spend, what they complain about in your reviews. Each one has a personality, buying habits and a memory of your category.
Group by group, they are the people already landing on your product pages — finally willing to answer questions. Talk to any one of them, or put a question to all 500.
“Should we raise free shipping from €49 to €59?” The same day, not in six weeks: who adds one more item, who pays the shipping, who leaves the cart. Group by group, with the reason each group gives, in their own words.
Not another dashboard to interpret. A verdict with the argument attached: what to do, who it moves, why, and a plain label saying how far you can lean on it.
A chatbot with your files connected is a fine analyst for the questions you think to ask. It can't see your rivals, it writes its own numbers, and it forgets when the chat ends. Here's what changes when the answer comes from a market instead of a model.
Hand us decisions you already made and know the result of: a price rise, a summer sale, ad A against ad B. We run them without seeing the outcome, and we put the same questions to a chatbot with your data. Then you score us both against what actually happened.
We won't print an accuracy number we haven't earned on shops like yours. Mirakls is new, so we prove it the honest way: test us blind on decisions you already know the result of, next to a chatbot, and keep score yourself. Two more things hold every day: the insights from your orders, reviews and rivals are counted by code, not guessed, and every simulated verdict carries a plain label saying how far you can lean on it, with its reasoning shown.
Keep Claude. It's great for the questions you think to ask, about the files you gave it. Mirakls does three things it can't: it reads your rivals (their prices, reviews and running ads), it counts instead of guessing, so the same data gives the same answer every time, and it puts your question to 500 shoppers built from your customers instead of one model's opinion. It also keeps watching: your store is re-read every week. Ask both the same question. Only one of them has done the reading.
It replaces guessing. Keep A/B testing everything an A/B test can reach, like subject lines and button colours. For everything it can't (a price every shopper sees, a free-shipping threshold, a new market), you now have something to test on where you used to have a hunch. Then try the winner in your store however you like.
The 500 shoppers are built for your shop, not pulled from a shared panel. Upload your orders or connect GA4 and they take on your real mix of countries, ages and spend. And it polices itself: if a group is too thin to carry a conclusion, the verdict says so before you act on it.
You can start with just your website. Everything else is optional and makes your shoppers more like your real customers. Store and analytics connections are read-only. An order file is read for its totals and then deleted, and customer names are never stored. Your data sharpens your model and only yours.
Built for e-commerce brands selling in Europe. WooCommerce connects directly. Shopify and any other platform work through a standard order export. Add Google Analytics 4, Google Ads and your Meta Ads Manager export for the full picture. Reviews and rivals' ads are pulled in for you.
Connect your store on Monday and read what your orders, reviews and rivals say the same week. Ask your price question, then the follow-up nobody budgets a second study for. Mirakls is in closed beta with a handful of European e-commerce brands. The next seats go to founders with a decision they'd rather not guess.
Mirakls is opening to a small group of European e-commerce brands. Give us your website and one real decision. We'll do the rest.
We'll start looking at your market and reply within a few days with your first steps.
Pick one you know the ending to. We run it without ever seeing the outcome, and put the same question to a chatbot. Then you put both answers next to what actually happened.
We'll model your market, simulate the decision without seeing your outcome, and send you the side-by-side. You reveal what really happened when you're ready.