Head-N-Tail Analysis to Increase Engagement #bbuzz
summary
Head‑n‑tail analysis claims to lift engagement by surfacing at least one relevant result for low‑traffic queries and gives concrete numbers: 100 unique queries with 100 clicks where 10 queries get 80 clicks and the other 90 get 20 clicks yields a potential ~70% lift if tails each get one click. Implementation uses Lucidworks Fusion 4.0 with Solr signals, Spark aggregations, a token and phrase spell correction job and a built‑in head‑and‑tail job that outputs rewrites and synonym or reindex actions; reports show examples like 35k vs 2.2k searches and 0.9 vs 0.04 conversion rates and require A/B testing to validate.
AI-written summary · about
chapters (15)
- 0:00Introduction to head-tail analysis
- 1:48Understanding Pareto principle
- 3:54Head-tail distribution in queries
- 5:18Query performance metrics
- 6:21Classifying poor-performing queries
- 8:26Fixing misspellings with head-tail
- 10:12Index-side misspellings correction
- 11:36Built-in spell correction plugin
- 12:25Generating head-tail reports
- 13:09Rewriting query suggestions
- 14:57Case study: MacBook Pro search
- 15:59Improving conversions with rewrites
- 16:41Bridging product catalog gaps
- 17:26A/B testing query models
- 18:12Q&A: Handling spam and slang
related talks
40mSmarter Search with Spark, Solr and Machine Learning
42mImproving Search @scale with efficient query experimentation
20mAn approach to internal search query analysis
14mVector Search Benchmark[eting]
46mDiscovery & Analytics with Hadoop, Mahout & Solr
39mLive-Hack: Analyzing 7 years of Buzzwords at Scale #bbuzz
45mUsing Search Data to Power Your Marketing Strategy
21mReporting Systems Built on Top of Hadoop & Lucene