Data Innovation Summit 2019
Data Innovation Summit 2019
- https://datainnovationsummit.com/presentations2019
- https://is.gd/dis19slides
- https://is.gd/dis19videos
Thursday
9:20 P Keynote: Building a win-win data environment for analytics success
- AI makes prediction cheap
- data becomes more valuable
- need to have clear data management & strategy
- central data repository (unified warehouse)
- ability to scale is crucial
- not accidental architecture/ecosystem
- study your data
10:30 M1 How to get mind-blowing ROI from your Analytics and Data Science teams
- problems:
- focus on big & actionable problems
- infrastructure
- start by MVP
- people
- have strategy for targeting top talent
- engaged people & shared ownership
12:00 M2 Agile Analytics – Building a data driven culture
- pull analytics instead of push
- every role should have analytics support
- Laloux culture model
13:30 M2 Doing data science at iZettle - What worked and what didn’t
- https://medium.com/izettle-engineering/how-we-do-machine-learning-at-izettle-6ff596585253
- centralized data team (without separate backlog) with members embedded in product teams
- focusing on high.prob.of succ/value, starting with weight on success
- always have a customer
- all data in one environment
- easily accessible data quality reports (dashboards even for managers)
14:30 M1 Algorithmic attribution modelling
- shapley values
- pouzitie v prinosoch pre zisky z reklamy (cileni, hw, …) ?
- https://github.com/slundberg/shap
16:30 M3 Handling personally identifiable information in a modern data platform
- anonymize data on ingestion and create mapping table (containing also 'removed' metadata)
- multiple mapping tables (prod, dev, diff access roles)
- adding fuzzy analytical fields
17:00 M1 How to introduce Data Science in big companies
- to introduce data science to a company, you need:
- business problem to solve
- meet with like-minded people regularly (mastermind method)
- define success measure
- taste your food while cooking (MVP)
- prepare model
- keep stakeholders engaged by showing data insight regularly, keeping up the momentum
- serve insight while hot
- process:
- build > show > use > feedback
- secret ingredient: excite people
Friday
9:45 P Keynote: Responsibility in the Data Age
- make ethics first class concern
- use ethical checklists from the internet
- book Ethics and Data Science
10:15 M6 Accelerating Analytics in a New Era of Data
- what makes competition competitive? The data.
- massive amounts:
-
- 60% time spent preparing
- 10% data gets analyzed
- => shallow reports
- systems are not designed for todays needs
-
- solutions: MPP DBs, MR
- alternative: GPU-accelerated DBs
- high throuhgput
- can analyze all data
- Sqream, MapD/OmniSci
10:45 M6 What problems should be solved with AI in Banking
- think different - don't be afraid to rethink instead of incremental evolution
- start small and learn fast
- we don't have time to do incremental changes
- work on BIG stuff (not just improving but creating the new)
- always think 10x
11:15 M2 Customer centric data science
- book Customer Centricity
- have strong customer identity
- store all interaction
- consent to use
- avoid personal data
- examples:
- collaborative filtering: recommending
- segmentation: long term behaviour (customer churn rate, lifetime…), e.g. prob. of alive customer given his recency vs frequency
11:45 M1 Model Industrialization in ING Bank
- analytics > innovation > new business models
- customer interaction based on analytics
- engage more data scientists instead of data analytics (use models instead of just manual analysis)
- empower analysts wit hthe tools od data scientists (see also auto-sklearn, mlflow.org etc.)
12:15 P Chairman's Closing Remarks
- learn by doing
- balanced innovation
- empower people with tech
- have in idea
- be stubborn
- don't be afraid to speak to people