What is Search Prefix?
Search prefix (or prefix search) returns matches where a field begins with the user’s input—e.g., typing “air m” surfaces “air max 270.” It’s the core of autocomplete and as-you-type experiences.
How It Works (quick)
- Indexing options:
- Edge n-grams (e.g., “a”, “ai”, “air”…).
- Prefix trees/FST/tries specialized for completion.
- Prefix fields separate from full-text fields.
- Querying: Apply prefix to specific fields (title, brand, category); gate by min length (e.g., 2–3 chars).
- Ranking: Blend prefix score with popularity, recentness, and business signals; de-dup near-identical suggestions.
Why It Matters in E-commerce
- Speed to intent: Fewer keystrokes to viable results and categories.
- Error prevention: Early guidance reduces typos and dead-end queries.
- Merchandising: Surface collections, brands, and policies as users type.
Best Practices
- Separate completion index: Keep prefix logic out of main ranking fields.
- Min/Max guards: Trigger after 2–3 chars; limit suggestions (5–10).
- Locale aware: Normalize accents/diacritics; handle hyphen variants (gore-tex/gore tex).
- No fuzz on codes: Disable fuzziness on SKU/MPN prefixes; offer a dedicated SKU mode.
- Diversity: Mix results: products, categories, brands, help topics.
- Telemetry: Track keystrokes/query, suggestion CTR, reformulations.
Challenges
- Over-inclusive n-grams, multilingual tokenization, popularity bias drowning niche items.
Examples
- Typing “magi” → “MagSafe charger,” “magnesium bottle,” “gift cards (help).”
- Typing “air” → brand/category suggestions + top products with phrase hits.
Summary
Prefix search makes autocomplete feel instant and helpful. Use a dedicated completion index, sensible triggers, localization, and diverse suggestions.
