A semantically fitting product can still be unsuitable.
Constraints are conditions that a result must actually meet. They prevent a meaningful-looking product from being shown while the size, connection, price, stock or customer access is incorrect.
The difference between preference and condition
A customer can have a soft preference for a particular material, but set a hard requirement for size or connection. Both signals affect search results, but in a different way. A preference can guide the ranking; a constraint excludes unsuitable products.
This distinction must be made consciously. If every trait becomes hard, too few results quickly arise. If each property remains soft, products that the customer cannot use appear.
-Common constraints
- Size and variant availability: the product must be available in the requested version.
- Price limit: “under 100 euro” is an explicit upper limit.
- Technical compatibility: connection, voltage, model series or platform must be correct.
- Stock and delivery time: Availability can determine the purchase decision.
- Market and customer rights: assortment, contract and visibility sometimes vary by country or account.
- Exclusions: words like “without”, “not” and “none” change the desired set.
The source data must be reliable enough to apply a constraint hard. A missing value is not automatically the same as “not enough”.
From natural language to filter
Query understanding first recognizes the condition. After that, it is linked to a structured product field. “60 cm wide”, for example, belongs to a width field with a known unit, not to a free text search to the number 60.
Normalization is important. Centimeters and millimeters must be correctly converted. Price limits should use the right currency and market. Sizes may have a different meaning per category. The search layer only applies a hard filter when the link is sufficiently secure.
What do you do with missing data?
Products without value can be excluded, lower placed or treated separately. The right choice depends on risk and assortment. Electric compatibility requires caution. With a soft style preference, an unknown value can still be shown as an alternative.
Make this choice explicit by attribute. This prevents missing product data from being interpreted tacitly as appropriate.
Contradictory conditions
A query may contain requirements that are not available together, such as a very specific combination of size, brand and price. A good system does not just release all constraints. It may indicate that nothing exactly satisfies and checked showing which condition differs with an alternative.
That transparency helps the customer decide. A result that exceeds the price limit undetected or has another connection is damaged.
Constraints and ranking
Hard constraints determine the allowed candidate set. Ranking determines the order within that set. This prevents a high semantic score from putting an unsuitable product at the top.
Some conditions may be gradual. A desired delivery time of one day, for example, can get a strong boost, while delivery in two days is still acceptable. Establish such rules as preferences and use harsh exclusion only where it is functionally needed.
Control and explanation
Log which constraints have been recognized, what values have been used and why candidates are excluded. That makes error analysis possible when a search query unexpectedly gives few results.
Findoviq combines constraints with entity recognition, product data and hybrid retrieval. Analytics shows which conditions often lead to zero-result searches. This allows a team to distinguish between a search problem, incomplete data and a real assortment deficit.
Discuss your search questions
Do you want to know how this approach fits your assortment, product data and customer behavior? Together, we look at which query types have priority and where exact, semantic and business signals need to complement each other.