AI product recommendations and search that understand your customers.
We set up search on your eshop that understands Greek, Greeklish, English, typos and queries like “running shoes under €80”, and product recommendations built from your real orders. Only products in stock are suggested, and your reports show what each recommendation sold.
Eshop search
papoutsia trexima under 80
Understood:
- Running shoes
- Up to €80
- In stock
14 products match
- Road 5, men's€69.90
- Runner Air, women's€74.90
- Trail Light, unisex€79.00
Frequently bought together
- Running socks, 3 pairs€12.90
- Cushioned insoles€18.50
From the orders of customers who bought the same shoes.
WHERE ESHOP SEARCH LOSES YOU SALES
A search in Greeklish that finds no products “No products found” for something you stock
Your product titles are in Greek. One customer types «papoutsia» in Latin letters, another «παπουτσια» without accents, a visitor from abroad types “running shoes”. The search finds nothing for any of them, and they buy elsewhere.
Search finds the product with or without accents, in Greeklish and with typos.
One search, step by stepA customer looks for a gift and the eshop shows socks Customers search in plain words
They type “gift for an 8-year-old” or “fridge for a 60 cm cabinet”. The eshop's search looks for those words in product titles and shows whatever happens to contain them.
Customers write in plain words, and search understands what they're after.
A list of related products with other washing machines and one sold out “Related products” that don't help
Under a washing machine, the eshop shows four other washing machines because they're in the same category, or one that sold out last week. The stand and the hose the customer would have bought with it never appear.
Under a washing machine, they see its alternatives and what goes with it: a stand, a hose.
A phone whose only email is the order confirmation After the purchase, silence
The only email after an order is the confirmation. A customer who bought an espresso machine never learns you also stock its descaler, and buys it elsewhere when it runs out.
After the purchase, suggestions for consumables and spares go to anyone who hasn't opted out.
ESHOP SEARCH AND AI RECOMMENDATIONS: WHAT WE SET UP
It's all set up on the data you already have: the catalogue, stock and orders in WooCommerce, Shopify or Odoo. It's one of the AI automations we set up.
Search in Greek, Greeklish and English
It finds the product with or without Greek accents, in Greeklish, with typos, in the singular or the plural, and an English query finds products whose titles are in Greek. The words your customers use for the same thing we agree with you.
Search in plain words
A customer types “running shoes under €80”, and the search picks out the category, the price and anything else they ask for. It shows only what matches and is in stock.
A no-results page that helps
Instead of an empty page, the customer sees the closest matches, such as the same shoe in another size, and can ask to be notified. You see what customers searched for and didn't find.
Similar items and “bought together”
Under a washing machine, the customer sees its alternatives and what they'll need with it: a stand, a hose, detergent. “Bought together” comes from your real orders and from rules you approve.
Picks for each customer
From what they viewed and bought, for those who accepted cookies or are logged in to their account. A new visitor gets suggestions from the product they're looking at and from their cart.
Suggestions by email after a purchase
Consumables, spare parts and anything bought again, around the time it usually runs out. They go to customers who didn't say no at checkout, and every email has an unsubscribe link.
ONE SEARCH, STEP BY STEP
What happens from the moment a customer types into the search box. Use the switch to see what happens when what they want is out of stock.
- The customer
- The search
- The catalogue
- Your reports
Results back to the customer
- A customer on their phone types «papoutsia trail 44»: Greek words in Latin letters.
- The search understands they want trail running shoes in size 44. Greeklish doesn't confuse it.
- The catalogue has 9 trail models with size 44 in stock. Anything not available in their size doesn't appear at all.
- The customer puts a pair in the cart. Next to it they see “Frequently bought together”: trail socks, from the orders of customers who bought the same shoes.
- They add the socks too and complete the order.
- The order records where each product came from: the shoes from search, the socks from the suggestion. Your reports show what each one sold.
- There are no trail shoes left in size 44. A plain search would show “No products found”.
- The customer sees the trail shoes available in 43 and 45, and leaves their email to hear when 44 is back.
- On Monday, in the list of what customers searched for and didn't find, you see “trail 44” was searched 23 times that week. You order it from your supplier.
- As soon as size 44 is back in stock, the customer who left their email gets a notification.
WHERE EACH RECOMMENDATION COMES FROM
Each recommendation rests on different data. Most of them need nothing from the visitor.
Scroll sideways to see all columns
| Recommendation | Where it appears | What it's built from | What it needs from the visitor |
|---|---|---|---|
| Search | Search box and results page | Titles, attributes, categories, synonyms and stock | Nothing |
| Similar items | Product page | The product's category, attributes and price | Nothing |
| Frequently bought together | Product page and cart | All customers' orders, without names, and your rules | Nothing |
| For you | Homepage and categories | What they viewed and added to the cart on earlier visits | Accepted cookies; otherwise they see best sellers |
| From your purchases | Customer account | Their previous orders | Logged in to their account |
| Email after a purchase | Email, when what they bought usually runs out | What they bought and when | Didn't say no at checkout; every email has an unsubscribe link |
Every recommendation shows only what's in stock and follows your rules: what is never suggested and what comes first.
HOW YOU KNOW THEY SELL
Search and every recommendation slot are measured separately. The order records, for each product, how it got into the cart: from search, from “Frequently bought together” or from an email. So GA4 shows you what each one brought in sales, not in clicks.
When the eshop has enough traffic, we also run a test: some visitors see the eshop without the new recommendations, and we compare the two groups' orders. That way the difference is measured, not assumed.
All of this rests on proper sales tracking, with a cookie banner and Consent Mode. How we set it up is on the page Google Ads & sales tracking.
Where each sale came from
One customer, two orders
- Search “espresso machine”click on the first result
- Espresso machine in the cart€149.00 · from search
- Descaler from “Bought together”€13.70 · from the cart suggestion
- Email: “Time for a new descaler?”seven weeks later, with an unsubscribe link
- New order from the email2 descalers · €27.40
- In your reports€149.00 from search, €13.70 from the suggestion, €27.40 from the email
PRODUCT RECOMMENDATIONS THAT GROW THE BASKET
A customer who finds what they want buys, and one who sees the right add-on buys more. AI product recommendations and search do exactly that.
Product recommendations, only from what's in stock
Similar and complementary products inside the eshop, only from what's in stock and by your shop's rules. Even a brand-new visitor with no history gets suggestions from the product they're viewing and their cart.
Customers describe it in their own words and find it
Search combines the words of the query with its meaning and puts the most relevant results first. The product gets found even when the customer doesn't know what you call it.
Search tested in Greek
We use questions with a known right answer to find which search method and model work properly in Greek. Then we run the same tests on what your own customers type.
Right categories, right “similar items”
AI reads the title and photo and puts every product in its place. That's what “similar items” and the filters that lead customers to the checkout are built on.
Want to see what your eshop could be recommending? Tell us what you sell.
HOW WE SET UP SEARCH ON YOUR ESHOP
As a guide, for an eshop with a few thousand products:
What they search for and buy
Week 1. We read your customers' searches, where they're recorded, and the last year's orders. We see what's bought together and what customers look for without finding it.
Catalogue and rules
Week 2. We fill in missing attributes and write the synonyms, in Greek and English. We agree the rules: what is never suggested, what comes first and when the email goes out.
Testing
Week 3. Search and recommendations run on a test copy of the eshop. You try your own searches, with Greeklish and typos, and we fix whatever you don't like.
Live, with measurement
Week 4. They go live on the eshop together with the measurement. Every month we look at what each slot sold and what customers searched for without finding.
QUESTIONS
What do product recommendations do on an eshop?
They show customers, at the right moment, products they need: alternatives on the product page, what's bought together in the cart, and consumables by email after the purchase. Together with search that understands Greek, Greeklish, English and typos, customers find what they're looking for and add more to the cart.
Does the search understand Greeklish and typos?
Yes. «papoutsia», «παπουτσια» and «παπούτσια» find the same products, and so do singular and plural. If you sell to English speakers too, “running shoes” finds them as well, even when the product titles are in Greek.
My eshop has few orders. Will it work?
Yes. Search and “similar items” are built from the catalogue, so they work from day one. For “frequently bought together”, until enough orders come in, the AI suggests which catalogue products go together, and you approve the ones you want.
Will it suggest products that are out of stock?
No. Search and recommendations read the stock in the eshop or Odoo and show only what can be bought. You can also set what should never be suggested, such as products you only order on request.
Do visitors have to consent?
Not for search, similar items or “bought together”, because they use nothing about the visitor. Suggestions from earlier visits work only if the visitor accepted cookies; otherwise they see best sellers. We also give you what your privacy policy needs to say.
Does it work with WooCommerce, Shopify or Odoo?
Yes, with whichever you have. WooCommerce picks “related products” by itself from shared categories and tags, in Odoo suggested products are set by hand on each product, and on Shopify automatic suggestions based on product descriptions work only on English storefronts. We fill them from your orders and catalogue. If you're building the eshop now, see eshop development.
Can we work in English?
Yes. Calls, proposals and handover are in English or Greek, and the synonyms and rules can be written in both languages, so the search serves Greek and English-speaking customers alike.
How much does it cost?
It depends on the eshop, the size of the catalogue and where you want recommendations. With that in hand we write you a fixed price.
RELATED PAGES
AI automation
Supplier invoices into Odoo, product descriptions and a chatbot that answers from your documents, checked by a person.
AI product data
Clean titles and attributes in one format. Filters, search and “similar items” are built on them.
Google Ads & sales tracking
Tracking that shows which ad, which search and which recommendation brought each order.
AI chatbot
Answers your customers only from your terms and catalogue, and hands over to a person what it doesn't know.
TELL US WHAT YOUR CUSTOMERS SEARCH FOR.
And which eshop you run. We'll show you what your search finds today and what it could suggest.