{"_id":"59e70b9fe027ae002e7d2f36","category":{"_id":"59e70b9ee027ae002e7d2edb","version":"59e70b9ee027ae002e7d2ed2","project":"5496d393f52a630b00519cdd","__v":0,"sync":{"url":"","isSync":false},"reference":false,"createdAt":"2014-12-21T19:01:45.031Z","from_sync":false,"order":9,"slug":"appendix","title":"Appendix"},"user":"5496d353f52a630b00519cdc","project":"5496d393f52a630b00519cdd","parentDoc":null,"version":{"_id":"59e70b9ee027ae002e7d2ed2","project":"5496d393f52a630b00519cdd","__v":2,"createdAt":"2017-10-18T08:06:54.462Z","releaseDate":"2017-10-18T08:06:54.462Z","categories":["59e70b9ee027ae002e7d2ed3","59e70b9ee027ae002e7d2ed4","59e70b9ee027ae002e7d2ed5","59e70b9ee027ae002e7d2ed6","59e70b9ee027ae002e7d2ed7","59e70b9ee027ae002e7d2ed8","59e70b9ee027ae002e7d2ed9","59e70b9ee027ae002e7d2eda","59e70b9ee027ae002e7d2edb","59e70b9ee027ae002e7d2edc","59e70b9ee027ae002e7d2edd","59e70b9ee027ae002e7d2ede","59e70b9ee027ae002e7d2edf","5b8661ccdd19310003a3fa0b"],"is_deprecated":false,"is_hidden":false,"is_beta":true,"is_stable":true,"codename":"","version_clean":"2.0.10","version":"2.0.10"},"githubsync":"","__v":0,"updates":[],"next":{"pages":[],"description":""},"createdAt":"2016-08-18T12:18:13.941Z","link_external":false,"link_url":"","sync_unique":"","hidden":false,"api":{"results":{"codes":[]},"settings":"","auth":"required","params":[],"url":""},"isReference":false,"order":2,"body":"In general, users’ Profile-Insights are inferred from home/work locations, visited venues, and, if enabled, apps-usage. Profile-insights are divided into the following main groups:\n  * Demographics; Gender, Household Income, Ethnicity, and such\n  * Interests; Ski, Pets, Fitness, Dining, Grocery, and such\n  * Intents/In-the-market-for... ; a car, furniture, electronics, and such.\n  * Brands; GAP, McDonald's, Old Navy, ,and such\n\nEach insight is assigned with a Confidence-Score: High, Medium, Low; which could be then mapped into Probability-Scores:\n  * Low: 60%\n  * Medium: 75%\n  * High: 90%\n\n**Demographics** \nScores are the results of Placer’s internal algorithms ‘trained’ on ground-truth.\n\n**\nInterests and Brands**\nAffinity is based on:\n  * Frequency of visits to selected and specific venue types.\n  * Population distribution along with user’s percentile within that distribution. For example, visiting fitness-gym 3 time/year may be assigned a ‘Low’ confidence score, while a single visit to ski resort may get an ‘High’ score.\n\n**\nIntents/In-The-Market for**\nAffinity here is similar to Interests and Brands, except it is more time-sensitive, with lesser emphasis on ‘older’ visits.","excerpt":"","slug":"using-insights-confidence-scores","type":"basic","title":"User Profile Insights - Confidence Scores"}

User Profile Insights - Confidence Scores


In general, users’ Profile-Insights are inferred from home/work locations, visited venues, and, if enabled, apps-usage. Profile-insights are divided into the following main groups: * Demographics; Gender, Household Income, Ethnicity, and such * Interests; Ski, Pets, Fitness, Dining, Grocery, and such * Intents/In-the-market-for... ; a car, furniture, electronics, and such. * Brands; GAP, McDonald's, Old Navy, ,and such Each insight is assigned with a Confidence-Score: High, Medium, Low; which could be then mapped into Probability-Scores: * Low: 60% * Medium: 75% * High: 90% **Demographics** Scores are the results of Placer’s internal algorithms ‘trained’ on ground-truth. ** Interests and Brands** Affinity is based on: * Frequency of visits to selected and specific venue types. * Population distribution along with user’s percentile within that distribution. For example, visiting fitness-gym 3 time/year may be assigned a ‘Low’ confidence score, while a single visit to ski resort may get an ‘High’ score. ** Intents/In-The-Market for** Affinity here is similar to Interests and Brands, except it is more time-sensitive, with lesser emphasis on ‘older’ visits.