Derives quality and characteristic signals for a location by analyzing travel patterns to and from it, turning travel investment into a behavioral quality signal beyond passive popularity counts.
Patent Overview
- Inventor
- Prabhakar Raghavan
- Assignee
- Google LLC
- Filed
- 2013-08-13
- Granted
- 2016-02-02
- Application Number
- US 13/965,571
The Challenge
Visit Counts Are A Noisy Quality Signal
Counting visits to a location measures convenience as much as quality. A coffee shop in a busy commuting corridor gets many visits because it is on the way, not because it is exceptional. Travel-time investment is a different signal: when users travel far or invest substantial time to reach a location, they are voting that the location is worth the effort. The system needs to mine travel-related visit data to extract these stronger quality signals.
- Proximity Inflates Visit Counts — Locations near population centers or transport hubs get many visits regardless of quality. Pure visit-count rankings reward convenience over excellence.
- Travel Investment Reveals Preference — Users who travel significant distance or time to reach a location are revealing a preference that overcomes the cost. The signal is stronger than passive visits and harder to game.
- Connectedness Is Implicit Quality — When a location is associated with a high-quality second location through user travel patterns (people who visit A also visit B), the connectedness suggests both locations belong to a similar quality tier.
- Need Visit Data Beyond Check-Ins — Pure check-in data is biased toward urban, smartphone-using populations. Visit data from broader sources (GPS traces, transit usage, anonymized location pings) gives more representative coverage.
- Privacy And Aggregation Are Required — Individual travel data is sensitive. The signal has to be derived from aggregated, anonymized travel patterns rather than from per-user trip identification.
Innovation
Travel Data Reveals Location Quality
The system collects aggregated visit data indicating travel of users to a first location from other locations, and travel from the first location to other locations. From this data it derives characteristics of the first location: a connectedness measure, association with second locations of known quality, and an inferred quality score driven by the time and distance users invest to reach it.
- Collect Aggregated Visit Data — Aggregate anonymized travel records: users who visited the first location, where they came from, where they went next. Source data must be anonymized to preserve privacy.
- Compute Inbound Travel Pattern — For the first location, analyze the distribution of origin locations for visiting users. Capture how far they traveled, what areas they came from, what time-of-day patterns emerge.
- Compute Outbound Travel Pattern — Analyze where visitors go after the first location. Patterns reveal what kinds of additional places visitors associate with this one.
- Derive Connectedness Measure — Combine inbound and outbound travel into a connectedness measure between the first location and each second location it is travel-associated with.
- Look Up Second-Location Characteristics — Retrieve known characteristics (quality tier, category, authority signals) of the second locations. These propagate as evidence about the first location through the connectedness.
- Compute Quality Score — Combine travel-investment signals (distance, time) and second-location characteristics into a derived quality score for the first location.
- Apply To Ranking — Use the derived quality signal in location-aware ranking. Locations with strong travel-investment evidence rank higher for queries where quality matters.
Travel Investment As Behavioral Quality
The mechanism turns aggregated travel patterns into a quality signal that is hard to manipulate and orthogonal to visit-count popularity. Users vote with their feet (and their travel time).
Effort To Reach Reveals Preference
When users invest time and distance to reach a location, they are signaling that the location overcomes the cost of getting there. Aggregate travel investment across many users produces a clean quality signal.
- Inbound Travel Pattern — Where visitors come from. Far-origin visitors and broad-origin distributions both suggest the location is a destination, not a convenience stop.
- Outbound Travel Pattern — Where visitors go next. Patterns reveal the category and quality tier the location is associated with by user behavior.
- Connectedness Inheritance — Strong travel-association with known high-quality second locations propagates quality evidence to the first location.
Quality is what people travel for.
<\/section>Technical Foundation
Data And Signals
The framework operates on aggregated, anonymized location data combined with quality signals about other locations.
- Visit Data — Aggregated, anonymized records of user travel patterns. Each record represents an inbound or outbound travel event without identifying individual users.
- Connectedness Measure — Strength of travel association between two locations. Combines inbound and outbound flow magnitudes.
- Second-Location Characteristics — Known properties (quality, category) of locations the first location is travel-associated with. Propagate as evidence.
- Travel-Investment Signal — Aggregate time and distance users invest to reach the first location. Larger investments indicate stronger preference signals.
Key Insight: Travel data is harder to fake than reviews or click data. Manipulating a location's review count is cheap; manipulating aggregate travel patterns to it is dramatically harder because it requires actual human movement at scale. The travel-investment signal is therefore a more robust quality input than other behavioral signals, especially for local-search ranking.
<\/section>The Process
End-To-End Quality Derivation
The pipeline runs offline against aggregated travel data and outputs per-location quality scores consumed by ranking.
- Aggregation — Aggregate travel data across user populations into per-location flow records. Anonymize to preserve privacy.
- Per-Location Pattern Computation — Compute inbound and outbound travel patterns for each location. Identify travel-associated second locations.
- Connectedness Scoring — Score the strength of association between each (first, second) pair.
- Quality Propagation — Propagate known second-location characteristics through connectedness to derive evidence about the first location.
- Travel-Investment Combination — Combine propagated characteristics with raw travel-investment magnitudes to produce the derived quality score.
- Publish To Ranking — Write per-location quality scores to the ranking feature store for use in location-aware retrieval.
What This Means for SEO
What This Means for SEO
Travel-investment signals are difficult to fake and increasingly important in local search. Understanding the mechanism shapes how to think about location quality, audience proximity, and competition with convenience-driven competitors.
- Convenience Is Not Quality — Being on a busy corridor produces high visit counts but weak travel-investment signal. Quality-driven local SEO needs to attract users who choose to travel, not users who happen to pass by.
- Attract Visitors From Distance — Locations that draw visitors from a wide origin distribution earn the travel-investment signal. Marketing that reaches beyond the immediate neighborhood directly feeds this signal.
- Cross-Visit Patterns Compound — When your visitors also visit other high-quality locations (and vice versa), the connectedness propagates quality between you. Being near (in travel-association) other quality businesses lifts your derived quality score.
- Destination Status Is A Real Signal — Being a destination people travel to rather than a stop people pass through is the strongest local-quality signal the system can read. Building destination appeal pays off behaviorally in ways review counts cannot match.
- Reviews Plus Travel Signals Together — The patent's signal is orthogonal to reviews. Both contribute. Locations that have strong reviews AND strong travel-investment patterns dominate over locations with only one.
- Privacy-Aggregated Data Is Real Data — Aggregated travel signals are part of the ranking even when individual location tracking is opt-out. The aggregate-level signal works even under strong privacy protections at the individual level.
- Cross-Visit Marketing Has Local SEO Effects — Partnerships, co-promotions, and packaged offerings with other high-quality nearby businesses produce the travel-association patterns that feed connectedness scoring. Local business networks compound this way.