Scaling a marketing intelligence platform - Aleph.bet event #scaling
Download as PPTX, PDF2 likes362 views
This document discusses scaling a marketing intelligence platform. It summarizes some of the challenges in scaling the platform and how they were addressed, including onboarding large amounts of data using a custom language to map taxonomies and process data in Hadoop, rebuilding data by replaying and manipulating it from the beginning, and making processes more self-service through quality assurance and configuration UIs. It also discusses building trust by improving research deliveries through integrating machine learning models with Hadoop and automatically refreshing models. Other topics covered include providing actionable insights, integrating with third parties, setting clients up for success through discovery, expectations, training and support, and ongoing usage analysis.
Scaling a marketing intelligence platform - Aleph.bet event #scaling
- 2. Helping marketers optimize their spend Across: Channels | Devices Online + Offline | External Factors Product
- 3. Named leader in report. Processes 100TB of data daily 10B daily data points 100+ data sources
- 4. Product Processes Organization High touch What to scale?
- 5. Product offering completeness Partial Full Client complexity Low High Low Mid High Low Medium High Convertro Yours? Touch Know your clients *AOL aims for Fortune 500
- 6. How did we hack scalability challenges?
- 8. Onboard Data • CSL (Convertro Specific Language) for taxonomy mapping Hadoop Parser • Real Time feedback
- 9. Rebuild Data • Replays and manipulates data from the beginning of time • Triggers: – Implementation Error – Bugs – Organization / LOB change – Belated data streams
- 10. Self Servicize* • Data quality assurance UI • Tag manager / builder • Integrations UI • Share link – snapshotting a page with a configuration *proactively simplify a process to be self service
- 11. Challenge #2:
- 12. Build Trust • Boost research deliveries: – Crawl – Walk – Run – Integrating R with Hadoop • Models automatically refreshed • Model comparison
- 13. Action is valuable • Automatic actionable insights • Integrations with 3rd party • Suggestive budget planner
- 14. Challenge #3:
- 15. Set a path to success • Client discovery questionnaire • Work through the hierarchy • Solution engineering • Setting expectations + timelines to get value day 1 • Training • Never ends
- 16. Usage analysis • Analyzing: features, pages, reports, integrations • Clients forum, surveys • Periodic calls
- 17. Yaniv Shalev yaniv@convertro.com t: yanivshalev li: il.linkedin.com/pub/yaniv-shalev • Understand your client pains • Add value or reduce cost • Fail Faster • Keep monitoring scalability debt
Editor's Notes
- #2: This is what data driven is all about
- #3: ROI analysis across channels Which channel works the best So Convertro is all about analytics and actionablitliy Take a better example: Jcpenny adevertising in TV, Catalogds Google , facebook they need understand the efficiency Hence they should expect to see something like this … We’re not going to talk why dashboards are needed right? Make Faster, Smarter, Data-driven Decisions Or why it’s so hard: Taking to the limit : Data Structures, Retrieval time, Flexibility , UI
- #5: We’re as tech people has the tendency to solve things with tech Product driven – MONITOR ALL THE TIME What/Where to scale – VERTEXES OF THE TRIANGLE Technology is changing and frankly this diagram is high level and pretty updated Each of you have to decide what it most significant for scale It changes – you can start with a lean product, then you move to differetn client, then you understand that you have to invest more with having professional service and processes Monitor what is the right mix Dev -> Customer services -> self service “Lean” Oriented Thinking Technology enough for scaling
- #6: Change manage through the hierarchy The more open the feedback loop: * No built on experiments * You need to convince people all the time You’re are building on other executer Not well defined KPIs The more complex is the client Organization Line of businesses Politics The growth hacking is to from high to low -
- #7: why? Big clients want you to get intimiate with their data, ow they will consider you as not matured Very hard to for them to R&D resources -> custom work tailored made @scale Very optimed
- #8: Or MESS CUSTOMIZATION why? Big clients want you to get intimiate with their data, ow they will consider you as not matured Very hard to for them to R&D resources -> custom work tailored made @scale Very optimed Mass custommization just emphsies
- #9: TO BE ABLE TO ONBOARD
- #10: LETS SAY A CLIENT IS CHANGING ITS UTM PARAMS
- #12: This is what data driven is all about
- #13: Intimacy with your client data
- #14: KICK ASS IN SELLING BETTER THEN DEEP LEARNING/ LOGISTIC / DESCION TREE
- #15: ALL THE KPIS ARE TALKING ABOUT RETAINING CLIENTS
- #16: CMO, SPECICALIST, HEAD OF BI, ADVANCE ANALYTICS, CFO …. . SOLUTION ARCH – ADDRESS NEEDS BUT NOT GETTING A SITUATION YOU’RE NOT ALIGNING WITH YOUR PLATFORM
- #17: * Add and example
- #18: IN EARLY STAGE IT’S REALLY IMPORTANT TO MAKE THE RIGHT DECISIONS – ADD VALUE OR REDUCE COST SCALABILITY DEBT IS NOT JUST TECHOLOGY