Global Search and Filtering

1) Role of global search

• understands the intention (entity vs action),
• results in ≤ 1-2 clicks,
• resistant to input errors and synonyms,
• is secure and respects access rights.

Principles: relevance> completeness, latency <200 ms before the first prompts, transparent filters and sorting.

2) Intent Model

• Entities: game, provider, tournament, report, payment method.
• Actions: "deposit," "withdrawal," "create a report," "invite an affiliate."
• Navigation: "wallet," "profile," "settings."
• Reference: "RTP," "limits," "KYC levels."

UX pattern: mark the result type with a badge (Game/Provider/Action/Help), from above - the most likely type by context (country, role, history).

3) Input points and form factors

Line in the header (desktop) + hotkey '⌘/Ctrl + K' - opens the palette.
Icon/tab bar search (mobile) - full-screen modal.
Contextual mini-searches - Filter the current list inside the catalog.

Placeholder hint: "Find a game, provider or action...."

4) Hints and auto-completion

Appear in 150-250 ms, top N popular/recent.
Results sections: Recent, Popular, Entities, Actions, Help.
Highlight, diacritic and translit (Pragmatic).
Support errors (fuzzy) and synonyms ("output "= "within," "cashout").
Keyboard-UX: arrows/Enter/Tab; on mobile - large targets.

5) Filtering and facets

• For games: provider, mechanics (Megaways, cluster), volatility, RTP, category (Live/Slots/Crash), tags.
• For reports: period, operator, geo, currency, status.
• For promotions/tournaments: dates, bets, participants, prizes.

• Chip filters over the results (selected facets are always visible).
• Multi-select in drop-down lists; counters next to the options.
• Reset All and Save As Preset.
• On the mobile - sheet with facets and the "Show results" button.

6) Advanced syntax (optional)

• `provider:"Pragmatic Play"`, `rtp:>=96`, `volatility:high`
• `type:live`, `tag:"buy bonus"`
• `market:CA currency:CAD`
• Auto-completion prompts for keys/values; poweruser users save time.

7) Relevance

• Match accuracy (title, alias, synonyms).
• Popularity/conversion (kliki→igra/depozit).
• Personal context (history, favorites, recent).
• Accessibility by role/jurisdiction.
• Freshness (new releases/promotions - boost limited time).

Show explanation when reordering ("New releases above within 7 days").

8) Empty and borderline states

No results: offer to remove/loosen filters, show popular in the current category, link to help.
Request too general: ask for clarification (show "categories").
Network errors: offline text, retry, saving the typed string.

Access restrictions: "Not available in your region/role" with CTA "Details."

9) Security, rights and compliance

RBAC/tenancy: index and results - only from available space.
Geo-guard: Hide inaccessible markets/games.
PII minimization: do not display personal data in tips.
Query log (impersonal) for audit and quality improvement.
Filter queries for prohibited keywords (brand guides, legal risks).

10) Performance and architecture

The time to the first prompts ≤ 200 ms, to the first results ≤ 300-400 ms.
INDEX: incremental updates, synonym dictionaries, n-gram/edge-ngrams for prefixes.
Caching popular queries and prompts.
Debounce 150-250 ms, cancel obsolete requests (abort/timeout).
SSR/streaming for the first screen of the catalog; then client filtering.
Virtualize long result lists.
Local indexes (Web Worker) for client filtering of already loaded sets.

11) Availability and localization

Role 'combobox '/' listbox', correct 'aria-', reading the number of results.
Focus management: return to the search bar, Esc - close, Enter - go.
RTL: Mirror the order of the chips and the position of the icons.
Terms and facets are localized, numerical formats and currencies are by region.

12) Telemetry and Metrics

• `search_open`, `search_type` (entity/action), `suggest_impression/click`,
• `query_submit`, `facet_apply/remove`, `preset_save/apply`,
• `result_click` (rank, id), `no_results`, `latency_ms`.

• Query → Result Click-Through (qCTR).
• Time-to-Result (first click/Enter).
• Zero-Result Rate (ZRR) and the proportion of "empty" saved.
• Facet Adoption %, Preset Adoption %.
• Action Success% (for intentions "deposit," "withdrawal").
• Latency p95 prompts and full output.

13) A/B experiments

Order of result sections (action entities vs).
Ranking strategy (popularity vs freshness).
Grid/results sheet, number of tooltips.
Variants of phrases in the placeholder/empty state.
Chip filters on top vs in the side panel (desktop).

Guardrails: INP/CLS, ZRR, complaints/errors, RBAC leaks = 0.

14) Scripted patterns (iGaming)

Find the game: entering the name/provider → tips (games, providers, tags). Facets: category, provider, volatility, RTP. Quick actions: "Demo," "Play."

Payment actions: 'deposit', 'withdrawal' → actions first. Diplink to wallet with context preserved.

Operator reports: 'ngr', 'arpdau' → Reports entity, period/operator/currency facets; presets: "Today," "Last Week."

Help/Policies: 'RTP', 'Limits', 'KYC' → Articles/FAQ, "Chat."

15) Antipatterns

Results without types (the user does not understand what he is opening).
Clues with a delay> 500 ms (user "loses rhythm").
Filters without visible indication of active conditions.
Reset state when "back" (line and facets are lost).
A search that returns entities without rights and with "access denied" by clicking.
Ignoring typos/synonyms, no translit.

16) Implementation checklist (step by step)

1. Directory of entities and actions (ID, aliases, synonyms, rights).
2. Index and ranking: rules, scopes, boosts (freshness/popularity).
3. Tips/AutoComplete: Debut, Highlight, History/Popular.
4. Facets and chips: multi-select, counters, "reset," presets.
5. Blank/errors: texts, alternatives, offline mode.
6. Performance: cache, cancel requests, virtualization, p95 goals.
7. A11u/locale: ARIA roles, keyboard, RTL, terms.
8. Telemetry and A/B: qCTR, ZRR, Time-to-Result, experiments.
9. Security: RBAC in index and issuance, audit logs.
10. Documentation: facet dictionary, sample queries, copyright guiding.

17) The bottom line

1. accurately understands intent and ranks by utility,

2. works instantly (prompts ≤ 200 ms),

3. gives transparent facets and preserves context,

4. respects rights/geo and speaks understandable language,

5. measures quality (qCTR/ZRR/Time-to-Result) and continuously improves through A/B.