Analytics – Balancing ‘Advanced’ and ‘Actionable’
summary
Panelists describe concrete failures and fixes: a homegrown collaborative filtering recommender launched quickly but surfaced absurd diaper recommendations until rules were added; price‑optimization spit out nonsensical premium pricing so constraints were layered on; perishables forecasting targets a ~2% uplift using iterative models rather than off‑the‑shelf deep nets and teams compare techniques in Python/R and sometimes LSTM/CNN prototypes; productionization often fails when models must undergo painful SQL conversion, turning deployment into a new project.
AI-written summary · about
chapters (10)
- 0:00Balancing advanced and actionable
- 1:15Actionability gap in analytics
- 7:54Cutting-edge vs practical outcomes
- 15:30Tangible results vs advanced techniques
- 22:49Analytics gone wrong
- 32:18Align analytics with business
- 36:36Disconnect between advanced and actionable
- 42:00Analytics ecosystem overview
- 47:06Insights as continuous journey
- 49:40Closing remarks
related talks
26mHow to develop useful and unbiased AI
22mReal-world applications of machine learning: Beyond the hype
34mForget about AI and do mathematical modeling instead!
19mLearning from eight years of data science mistakes
02mScale By the Bay 2021 intro video by Adarsa Sivaprasad
36mML: Predictive analytics in Ad Engineering
23mBig data and big insights
01h 17mLightning Talks