Setting Up Search Fields, Ranking, Stop Words, and Spell Check

Updated

Set up the factors that determine which results appear first in Site Search. By the end, you will have a practical baseline for Search Fields, Ranking, Stop Words, and Spell Check.

Prerequisites

Before you start, you need:

  • a Site Search account
  • a Search App with data already ingested
  • the fields you want to use already indexed by connector, crawler, or API
  • a rough idea of which fields matter most for your search experience

Why These Settings Matter

These settings determine whether the right items surface for the right query. If the wrong items appear first, users often assume search is broken even when matching content exists.

  • Search Fields decide which fields Site Search checks for query matching.
  • Ranking prioritizes certain fields or patterns over others.
  • Stop Words remove low-value common words from query matching.
  • Spell Check helps recover no-result queries caused by typos.

Search Fields Baseline for New Setups

A common starting point is to use title, description, and content/body as Search Fields. Add other fields only when your relevance goals require them.

Note: Avoid non-tokenized string-only fields in Search Fields when you want broad text matching. These fields are usually exact-match oriented and can reduce recall.

Tip: During onboarding, keep the baseline simple. Include the fields that should drive relevance first, then expand after you validate real queries.

Configure Search Fields

  1. Open your Search App in Site Search.
  2. Go to Configurations > Basic Configurations > Search Fields.
  3. Review the prepopulated fields.
  4. Add or remove fields based on your relevance goals.
  5. Click Save Draft, then Publish.
  6. Validate the configuration in Preview.

Field Type Guidance

Use tokenized text fields when you want broad keyword matching. If you need word-variant matching, verify your searchable fields use a language-appropriate text pipeline with stemming support.

For API-driven schema setups, use copy-field patterns so a tokenized, stemmed target field is available for search. If stemming introduces risk or complexity in your implementation, validate that behavior before depending on it in onboarding guidance.

Ranking Setup Strategy

For new setups, start with a simple ranking approach:

  1. turn on Smart Ranking if it is available for your account
  2. add manual boosts only where they are easy to explain and validate
  3. start with title, then description, then content

After you have meaningful usage data, revisit those boosts using analytics and query examples.

Note: Keep field-value boosting out of the initial setup path unless you have a clear, validated use case. For most new implementations, Search Field and Function strategies are the more important starting point.

Manage Stop Words

  1. Go to Configurations > Basic Configurations > Stop Words.
  2. Add words one at a time or upload a newline-delimited .txt file. Put one stop word on each line and don't include a header row. For example: 
    a
an
the
of
for
  3. Click Save Draft, then Publish.

Note: Stop Words matter during initial setup because they reduce noise from common query terms before you begin deeper tuning.

Tip: Download the SearchStax recommended English stopwords starter list (.csv), then review and trim the list before using it in production.

Note: Stop Words apply only to tokenized text fields. They do not affect non-tokenized string fields.

Spell Check Setup Pattern

Spell Check runs when a query returns no results and then retries with the closest match from the configured source.

  1. Go to Configurations > Basic Configurations > Spell Check.
  2. Enable Spell Check.
  3. Choose a method:
    • Dictionary-Driven
    • Data-Driven
    • Data+Dictionary-Driven
  4. For data-driven Spell Check, verify the source field is populated for each active language.
  5. For dictionary-driven Spell Check, add words manually or upload a plain text .txt file with one correct word per line. Don't include a header row. For example: 
    catalog
invoice
dashboard
enrollment
  6. Click Save Draft, then Publish.

Spell Check runs only when the original query returns no results. To validate dictionary-driven Spell Check, test a deliberate misspelling that doesn't already return results.

Tip: For developer-led setups, a dedicated spell-check target field is often easier to validate than relying on a loosely defined content field.

Troubleshooting Low-Quality Results

If results still feel wrong after a baseline setup, inspect the request and scoring behavior before changing multiple settings at once.

  • Use debugQuery=true when validating Search API behavior.
  • Check whether the wrong fields are dominating score.
  • Confirm whether the issue comes from Search Fields, Ranking, or Data Filters.

Note: Aggressive synonym expansion can increase query size and complexity. Keep synonym groups focused and validate frequent queries after synonym changes.

Validate in Preview

  1. Open Preview.
  2. Run a query set with known expected results.
  3. Confirm terms matching boosted fields rank higher.
  4. Confirm common stop words do not overwhelm intent.
  5. Confirm misspelled queries recover as expected when Spell Check is enabled.
Site Search Preview showing a corrected misspelled query and ranked results.

What's Next

Next, explore Configuring Advanced Search Features after your baseline setup is stable.

Articles in this section