The 80/20 rule for Search Dictionaries: 80 percent of the value comes from the first 50 mappings sourced from real zero-result queries, and 20 percent of the value comes from the next 200 mappings sourced through anecdote. Mine the data, then iterate monthly. Skipping the data step produces mappings that look thorough and miss what users actually search for.
- Pull the zero-result search log
Setup, Search, Search Insights (or Reports, Search Term Reports). Sort by frequency, descending. The top 50 queries with no results are your initial mapping candidates.
- Map each top zero-result query to existing terms
For each query, find the term it should match (acct -> account, qty -> quantity, mgr -> manager). Add the mapping as a symmetric list in Setup, Search, Search Dictionaries.
- Confirm standard dictionary coverage first
Some terms already exist in the standard dictionary. Adding a duplicate in custom does not hurt but wastes admin time. Check standard before authoring.
- Add per-language dictionaries for multi-language orgs
English admins regularly forget the non-English dictionaries. If the org operates in French, German, Spanish, etc., the same mappings need translation work per language.
- Wait one week and re-pull the zero-result log
The added mappings should reduce zero-result rates on the queries you addressed. The log confirms whether the mappings work or need refinement.
- Add per-object scope if mappings create cross-object noise
Some mappings should apply to specific objects only (a custom acronym used on Account but not on Contact). Use dictionary scope to restrict.
- Schedule a monthly mapping review
Pull the zero-result log monthly. Add 5 to 15 new mappings. The cadence is what keeps the dictionary aligned with how users actually search.
Per-language dictionaries. Each language needs its own mappings; English mappings do not affect French searches.
Mappings are symmetric by default; every term in the list is interchangeable with every other.
Restrict a mapping to specific objects when the term has different meanings on different objects.
Standard is Salesforce-managed and read-only; custom is admin-managed. Custom wins on conflict.
Zero-result log mining is the highest-ROI source; anecdote is the lowest-ROI source.
- Over-broad mappings (account + user in one list) confuse results across object types. Keep mappings tight to a single concept.
- Mappings are symmetric by default; "acct" -> "account" automatically also makes "account" find "acct" matches. Plan accordingly.
- Multi-language orgs need per-language dictionaries. The English admin who forgets this leaves the non-English search experience untouched.
- Standard dictionary coverage is wider than most admins expect. Check standard before authoring custom mappings to avoid duplicating work.
- Mappings that are not reviewed go stale. Quarterly review with the zero-result log keeps the dictionary aligned with how users actually search.