Turning a shopper's sentence into product search constraints
Turning a shopper's sentence into search constraints
"A quiet dishwasher for a small flat under £400" holds a budget, a size limit, a noise preference and a few unspoken needs. Here is how to split a shopper's sentence into hard and soft constraints, handle units and exclusions, and know when to ask a follow-up question.
Shoppers don't type filters. They type sentences. "A quiet dishwasher for a small flat under £400" reads as one simple request, but it holds a budget, a size limit, a noise preference and a few needs the shopper never said out loud. Getting from that sentence to the right products means pulling it apart carefully. This guide walks through that example step by step.
Step one: list everything the sentence says
Start by writing down every requirement, stated or implied. Don't judge them yet.
Product type: dishwasher.
Price: under £400.
Noise: quiet.
Size: suits a small flat.
Market: the £ sign suggests a UK shopper, so UK retailers and UK plug and pricing.
Step two: sort hard from soft constraints
A hard constraint rules a product out. A soft constraint changes the order. Mixing them up is the most common mistake. Treat a preference as a hard filter and you may return nothing. Treat a hard limit as a preference and you show products the shopper can't use or can't afford.
Requirement
Type
How to apply it
Dishwasher
Hard
Category filter or core of the query
Under £400
Hard (with care)
Price range filter, maximum 400 GBP
UK market
Hard
Market set to GB
Small flat
Soft, possibly hard
Rank slimline and compact models higher; ask if space is fixed
Quiet
Soft
Rank by lower noise level where the data is available
Budget is marked "with care" because shoppers often say "under £400" when they would consider £410 for the right product. A good pattern is to filter strictly, then mention near misses separately: "There's also one at £419 that is notably quieter."
Step three: turn vague words into units
"Quiet" and "small" are not specs. To rank by them you need measurable fields, and you need to choose a reading you can explain.
Quiet
Dishwasher noise is usually listed in decibels (dB). Lower is quieter. Rank by the noise level field and treat products with no listed noise level as unknown, not as quiet.
Small flat
Dishwashers commonly come in full-size and slimline widths, with compact countertop models too. Width in centimetres is the key field. Rank narrower models higher.
Under £400
Compare price amounts in GBP. Remember stored prices are snapshots and may have changed, so items near the limit should be confirmed at the retailer.
Write down the reading you chose. If your assistant explains "I've favoured slimline models because you mentioned a small flat", the shopper can correct it in one line.
Step four: find the implicit needs
Some needs are never stated because they seem obvious to the shopper. A small flat suggests a few likely ones:
Freestanding or integrated? Many flats have fitted kitchens, so an integrated model might be needed. This is worth asking.
Capacity: fewer place settings is usually fine for a smaller household.
Quiet also matters more in an open-plan space, which reinforces ranking by noise.
Implicit needs should usually be soft. You are inferring them, so let them shape the ranking but not remove products.
Step five: handle exclusions
Shoppers often add exclusions in follow-ups: "not that brand", "no countertop ones", "nothing refurbished". These are hard constraints and should persist for the rest of the conversation. A common failure is forgetting an exclusion two messages later. Keep a running list of constraints, not just the latest message.
Step six: know when to ask
Asking too many questions is annoying. Asking none leads to wrong answers. A useful test: ask only when the answer would change which products are possible, not just their order.
AskFreestanding or integrated? The two are not interchangeable, so the answer removes products.
AskDo you have a fixed space, such as a 45 cm gap? That turns size from soft to hard.
Don't askHow quiet exactly? Ranking by noise level handles it.
Don't askWhich retailer? Show several and let the shopper choose.
Often the best move is to show a first set of results and ask one question alongside them: "Here are three slimline models under £400, quietest first. Is your space for a freestanding or built-in machine?"
Putting it into a search
With Simlir, the natural-language part of the request goes into the query and the hard constraints become filters. Using the REST search endpoint, the example might look like this:
Code exampletext
GET https://simlir.com/api/v1/search?q=quiet+slimline+dishwasher&gl=gb&limit=10
plus a price range filter with a maximum of 400
Each search costs 1 credit. Results come back as structured JSON with specifications, a dated price snapshot, a retailer link and a relevance score, so your assistant can then apply the soft ranking on noise and width. Treat the limit as a ceiling, not a promise: a tightly constrained search can return fewer results, or none. See the API docs for the exact parameter names.
What is the difference between hard and soft search constraints?+−
A hard constraint rules a product out, such as a budget, the market or a product type. A soft constraint only changes the order, such as a preference for quieter or narrower models.
What if a product has no noise level listed?+−
Treat it as unknown. Don't rank it as quiet, and don't remove it unless noise is a hard limit. Tell the shopper the figure isn't listed.
Should the budget be a strict filter?+−
Usually yes for the main results, because a shopper's stated limit deserves respect. You can mention a close option just above it separately, clearly labelled.
How do I keep constraints across a conversation?+−
Keep a running set of constraints and update it with each message, rather than rebuilding from the latest message alone. Exclusions in particular should carry forward.
