Problems Leitagu solves

This page describes six problems shoppers in Estonia have when buying furniture: matching what they own, getting it in time, making it fit, knowing what to add, searching many stores, and trusting colours on a screen. For each one it gives a typical scenario, what people do today, what Leitagu does instead, and what is still unsolved. The people in the scenarios are the personas from Who it is for.

Leitagu currently runs on example product data: every product, price, stock level and delivery time in the app is made up and labelled as such. The behaviour described here is real; the catalogue is not yet.

#1. Matching what you already own

Scenario. Toomas has a light oak dining table with an oiled finish and tapered legs. He wants a sideboard (puhvetkapp) in the same light oak, at most 160 cm wide, for under €400.

What people do today.

• Filter each store's site by "oak" and get dozens of results in tones from pale to honey, because stores name colours differently ("tamm", "sonoma oak", "natural").
• Squint at studio photos to guess whether a piece is solid wood (täispuit), veneer (spoon) or wood-look foil (kile, laminaat).
• Take a photo of the table to the store and hold the phone next to the sample.

What Leitagu does.

• Takes a reference in three ways: a description ("hele tammest puhvet"), a piece in a saved room (tapping it shows what Leitagu sees: material, wood, tone, undertone, finish, style, legs, handles, fronts and a colour swatch), or a room photo, which becomes a room with its pieces.
• Compares each attribute on its own and gives partial credit for near values, for example one step lighter, the same finish family, or solid wood against veneer.
• Lets you say what matters: tone and colour, style, or both. An "exact" match makes tone, undertone, finish, colour and material count more and drops weak matches.
• Groups results into "Very close", "Close" and "Other options", and explains each one in a line such as "Same light oak · warm undertone · veneer like yours · Scandinavian style".
• Flags wood-look foil on cards: "Wood-look foil, not real wood".
• Accepts corrections such as "it's walnut, not oak" (this needs the Claude connection) or a tap on the filter chip.

The exact rules are in How it works.

Still unsolved.

• Product attributes in the example catalogue were entered by hand. For real store products, a sync reads them from the store's data fields and product texts with fixed word lists (how fields are mapped), and stores can correct them. Reading attributes from product images is planned, not built.
• Leitagu does not compare images yet, so shape and wood grain pattern play no part. Matching uses attributes and one average colour per piece.
• A colour read from a phone photo is an estimate.
• "Goes with it but isn't the same" matching within one category (a complementary match) is not built. The "what should I add" suggestions do match other categories to the room's look.
• Match quality has not been measured on a test set of real photos and products. A 50-pair test set is planned.

#2. "Can I take it home today?" versus weeks of delivery

Scenario. Kadri gets the keys to her flat in Tartu on Saturday. She needs a chest of drawers (kummut) for the hallway that weekend and does not want to wait for delivery.

What people do today.

• Open product pages one at a time to find stock per store and the delivery time, which often appear only on the product page or at checkout.
• Call the store, or drive there and hope.
• Find out late that a piece is made to order and takes weeks or months. Long waits are common enough that Kaup24 publishes an article on why furniture sometimes takes months to arrive (kaup24.ee↗).

What Leitagu does.

• Asks "How soon do you need it?" with four answers: take it home today (täna kaasa), see it in a store first, delivered within 2 weeks, or doesn't matter. For the first two it asks which city: Tallinn, Tartu, Pärnu, Narva or all.
• Treats the answer as a filter, not a hint: a product that is not in stock in Tartu is not shown for "take it home today in Tartu".
• Shows availability on every card: "Take home today: Tartu (3 pcs)", "On display: Tallinn, Tartu", "Home delivery 4–6 days", or "Not in stock, delivery 6–8 weeks". Delivery over two weeks is highlighted.
• Understands it in text too: "täna kaasa Tartus", "I can take it home today".

Still unsolved.

• All stock, showroom and delivery data is example data. Real per-store stock and delivery times have to come from the stores; the outreach email asks for exactly these fields.
• Stock changes during the day. Leitagu does not yet show when stock was last checked.
• Only four cities are modelled, and delivery to a specific address or region is not.
• You can't reserve a piece through Leitagu.

#3. Will it fit my room?

Scenario. Liis and Marko's living room has a 480 cm back wall with a 150 cm window whose sill is 85 cm from the floor, and a door on the left wall. They want a TV unit (TV-alus) and a bookcase (riiul).

What people do today.

• Measure the wall, write the numbers on a phone, and filter each store by width, if the store has a width filter.
• Forget the window sill height or the space the door needs to open.
• Discover at delivery that the bookcase covers half the window.

What Leitagu does.

• Keeps a model of the room: its size, wall and floor colour, windows (wall, position, width, sill height), doors, and the furniture already there.
• Works out the free stretches of wall. Furniture standing within 60 cm of a wall blocks it, each door keeps its width plus 10 cm on each side clear, and under a window the tallest piece allowed is the sill height minus 3 cm.
• When a room is linked to the chat and you ask for storage furniture, the size question offers a ready answer: "Fits your room: up to N cm", based on the longest free stretch of wall.
• In "what should I add" suggestions, a stretch under a window comes with a height limit, and pieces that shouldn't stand under a window (TV units, bookcases, cabinets) are not suggested there.
• Shows the product in a 3D view of the room, in the first free gap or in place of the piece it replaces.

Still unsolved.

• The room model has two walls, the back wall and the left wall. The other two walls are not modelled.
• Radiators, sockets, skirting boards and sloped ceilings are not modelled yet (planned).
• A room made from a photo has estimated measurements. The app asks you to check them.
• The size suggestion uses the longest free stretch, not the exact spot you may have in mind.
• The 3D view is labelled "Illustration, not to scale".
• Real store data sometimes lacks dimensions or has them only in the product name, which makes size filtering unreliable for those products.

#4. "I have space but don't know what to add"

Scenario. Anna's bedroom has a bed, a rug and a plant. Something is missing, but she doesn't know what, and she can spend at most €200 per piece.

What people do today.

• Browse Pinterest or store category pages for ideas.
• Guess what would fit, then measure, then search.
• Buy something that turns out too big or in the wrong tone.

What Leitagu does.

• Offers "What could go here?" (Mida siia lisada?) whenever a room is linked to the chat, and also understands "what should I add?" in text.
• Looks at the pieces a room of that type usually has, skips what is already there, and checks where they would fit on free wall or floor.
• Reads the room's look (colour, tone and wood from the main storage piece or table, and the most common style) and searches for products that match it.
• Keeps your own limits, such as budget, stores and availability.
• Shows up to four ideas, each with a reason and up to eight products. Browsing an idea places each product in that free spot in the 3D room.

If Anna starts from Leitagu's bedroom template (380 × 360 cm, a window on the left wall with the sill at 90 cm, the door on the back wall), the ideas are a chest of drawers ("There is 138 cm of free wall on the left wall. A chest of drawers up to 128 cm fits there."), a side table under the window (up to 55 cm wide and 87 cm tall) and a shelf up to 92 cm wide. Each idea appears only if Leitagu finds matching products under her €200 limit.

Still unsolved.

• Suggestions cover only the eight categories in the catalogue: sideboards, chests of drawers, cabinets, TV units, shelves and bookcases, and dining, coffee and side tables. Lighting, textiles, beds and sofas are not covered.
• The rules are fixed lists per room type, not learned from how people furnish.
• There is no whole-room or whole-flat budget planner (planned), and you can't dismiss an idea so it stops coming back.

#5. Products scattered across many stores

Scenario. Kadri wants to compare white chests of drawers, at most 120 cm wide, across every store she knows: big interior stores, international chains and smaller shops.

What people do today.

• Open ten tabs. Each store has different filters, and some can't filter by width at all.
• Translate between colour names and between where each site lists dimensions.
• Lose track of which piece was in which store.

What Leitagu does.

• Keeps one catalogue in which every product is described with the same fixed vocabulary of attributes, whatever the store called it.
• Ranks results from all stores together, by match, with one set of filters.
• Understands store preferences in text, for example "only Aatrium" or "everything except JYSK".
• Links "Open in store" straight to the store's own product page.

Still unsolved.

• No real store data is in Leitagu yet. The example catalogue uses real store names so the prototype feels realistic, but no store has supplied or approved any data. Stores can now connect a Shopify store, a product feed or a CSV file in the partner app; none has yet.
• Getting data differs by store: feeds are the preferred route, some sites publish structured data, and some block automated access.
• Legal limits apply to copying catalogues and showing product images. A product whose data includes an image link shows that image, loaded from the store's website; example products have none and show a 3D rendering. How to show stores' images is still to be settled with them.
• Second-hand marketplaces are out of scope for now.

#6. Screen colour versus reality

Scenario. Toomas orders a "natural oak" sideboard online. On screen it looked right next to a photo of his table. At home it is noticeably more orange.

What people do today.

• Order wood samples or swatches, where the store offers them.
• Visit the store with a photo or a drawer front.
• Return the piece, if the store allows it, and pay for the trouble.

What Leitagu does.

• Shows a note on every results list: "Colours on screen differ from real life. For an exact tone, see the piece in a store."
• Offers the "See it in a store first" filter, and shows on each card which cities have the piece on display.
• Describes tone (six steps from white-washed to very dark) and undertone (warm, reddish, neutral, cool) in words, so you can check them against the real piece, not only against pixels.
• Flags wood-look foil, which can photograph like real wood.
• Gives a photo tip: put a sheet of white paper in the photo and shoot in daylight. If Claude sees the paper, it corrects the colour reading and says so.
• Has a day and evening light switch in the 3D room, with the note "Warm evening light changes wood tones."

Still unsolved.

• There is no calibrated colour measurement. Each piece is one average colour, compared with a simple colour-difference formula.
• Store photos are taken in different light and some colour variants reuse the same photo.
• Ordering swatches is not part of Leitagu.
• The colours in the 3D view are approximate too.

#Summary