Merges search results from a local search engine and a universal search engine into a single ranked page, emphasizing certain business entities in the combined output while preserving cross-source rank quality.
Patent Overview
- Inventor
- Amit Singhal
- Assignee
- Google LLC
- Filed
- 2010-05-04
- Granted
- 2013-03-05
- Application Number
- US 12/773,724
The Challenge
Combining Local And Universal Results
Many queries straddle local and global intent. A search for a brand name might want both the brand’s national homepage and the user’s nearest physical location. Pure local search misses the universal context; pure universal search misses the local relevance. The system needs a merge that combines both, ranks them together, and emphasizes the business entities that the merged page needs to surface.
- Local And Universal Have Different Score Spaces — Local rankings are computed against a small local index; universal rankings are computed against the entire web. Their raw scores are not directly comparable, so a merge has to normalize.
- Business Entity Emphasis Is The Goal — The merge is not just about combining lists; it is about giving certain business entities prominence in the final page. The merge logic has to identify those entities and ensure their representation.
- Need A Unified Ordering — The user expects a single ranked list, not two parallel columns. The merge must produce one ordering that interleaves local and universal results sensibly.
- Duplicates Across Engines Must Be Handled — The same business may appear in both the local results and the universal results. Showing it twice is wasteful. The merge needs deduplication that picks the better representation.
- Page Real Estate Is Limited — Merging two full result sets into one page forces choices about which results from each source make it onto the visible page. Both sources contribute, but space caps participation.
Innovation
Combine, Normalize, Emphasize, Render
The system runs the query against both the local search engine and the universal search engine in parallel. It normalizes scores across the two sources, identifies business entities that should be emphasized, deduplicates overlap, and produces a single ordered result set that combines both inputs while highlighting the chosen business entities.
- Issue Query To Both Engines — The query is sent in parallel to the local search engine and the universal search engine. Each produces its own ranked list with native scores.
- Normalize Scores Across Sources — Raw scores from the two engines are not comparable. The merge normalizes them to a common scale using calibration data or relative rank position.
- Identify Business Entities — Scan both result sets for documents that represent business entities. The merge gives these entities priority placement in the final ordering.
- Deduplicate Overlap — When the same business or document appears in both result sets, keep the better representation (typically the local result if it carries richer metadata; otherwise the higher-ranked one).
- Apply Emphasis Logic — Identified business entities are placed at positions that emphasize them in the final page: top of the merged list, in a featured-result module, or in dedicated slots.
- Merge The Remaining Results — Non-emphasized results from both sources are merged into the rest of the page by normalized score, producing a single ranked tail.
- Render Unified Page — The user sees one ranked page that mixes local and universal results, with business entities prominent. From the user’s perspective there is no visible source split.
Unified Output From Two Indexes
The patent’s contribution is doing the merge intelligently rather than as a simple concatenation. Normalization, deduplication, and business-entity emphasis together produce a result page that respects both sources while serving a coherent user experience.
Sources Are Inputs, Page Is The Output
The user does not need to know there were two search engines. They just need the best combined result page, where local relevance and universal relevance both contribute proportionally.
- Score Normalization — Brings the two engines’ scores into a common scale. Without normalization, one source would systematically dominate or be dominated.
- Entity Emphasis — Business entities get priority placement so the merged page highlights them appropriately. This is what makes the merge useful for commercial intent.
- Deduplication — Overlap between sources is resolved by keeping the better representation. The page does not show the same business twice.
Technical Foundation
Merge Inputs And Outputs
The merge takes two ranked lists and produces one ranked list, with explicit entity emphasis and deduplication logic.
- Local Result Set — Top-k results from the local search engine, each with a native local score and rich local metadata (business name, address, hours, distance).
- Universal Result Set — Top-k results from the universal search engine, each with a native universal score and standard document metadata.
- Score Normalization Function — Brings local and universal scores into a comparable scale. Can use calibration constants, rank-position remapping, or learned conversions.
- Entity Emphasis Rule — Which business entities get priority placement and where they sit in the merged page. Driven by query type and entity confidence.
- Deduplication Logic — When the same entity appears in both sources, decides which version to keep and where to place it. Typically prefers the local version when local metadata is richer.
Key Insight: Treating the merge as more than concatenation is what makes the universal search experience work. Normalization, emphasis, and deduplication each address a failure mode of naive merging. The combination is what lets Google’s search results page intermix local entities, web pages, news, images, and other verticals without obvious source seams.
<\/section>The Process
End-To-End Merging
Both engines run in parallel; their outputs feed the merge stage; the merged output goes to the renderer.
- Parallel Retrieval — Issue the query to both engines. Wait for both result sets to return.
- Score Normalization — Bring local and universal scores into a common scale using the configured normalization function.
- Entity Identification — Scan both result sets for business entities. Tag each entity with its emphasis level.
- Deduplication — Identify duplicate entities or documents across the two sources. Keep the better representation per the deduplication rule.
- Emphasis Placement — Place emphasized entities at the predetermined emphasis positions (top of list, featured module, dedicated slot).
- Merge Remaining By Normalized Score — Order the remaining (non-emphasized, non-duplicate) results by normalized score and append to the page.
- Render To User — Output the unified page to the renderer for display.
What This Means for SEO
What This Means for SEO
Universal search merging is why a single SERP can show your Google Business Profile, your website, your news article, and competing entities all on one page. Knowing the merge logic shapes how to think about cross-format SEO.
- Local And Universal Both Matter For Mixed-Intent Queries — Brand queries, service queries, and many commercial queries trigger universal search. Optimizing only your website or only your Google Business Profile leaves half the SERP on the table.
- Business Entity Markup Earns Emphasis Slots — Pages with strong entity signals (LocalBusiness schema, consistent NAP, knowledge-graph presence) are more likely to be identified as business entities and placed in emphasis positions in the merge.
- Google Business Profile Is Part Of The Universal Index — Your GBP listing participates in the universal search merge. Local Pack appearances stem directly from this mechanism. Treat GBP optimization as part of SERP strategy, not a separate channel.
- Duplicate Across Properties Is Resolved — If you appear in both local and universal results for the same query, the merge keeps the better representation. Aim for both to be strong; the merge will pick the right one.
- Score Normalization Equalizes Sources — The merge normalizes between local and universal scores, so a high-local-confidence result can beat a higher-raw-universal-score result. Optimize for source-native quality, knowing the merge will compare across sources.
- Featured Modules Take Real Estate — Emphasized business entities can occupy multiple result slots through featured modules (knowledge panels, Local Pack, product carousels). These modules compress the ranking space for non-featured competitors.
- Cross-Format Optimization Pays — Sites that own multiple formats (web pages, local listings, news, images, video) participate in the merge across formats. A single piece of content covering one format misses cross-format ranking opportunities.