S P A R K

S P A R K S P A R K Two experiments that yielded 30% increase in Wireless Broadband acquisitions 5x better at identifying at-churn-risk customers

S P A R K Time in months -> Wireless Broadband Base Movement -> Churn experiment Acquisition experiment Net Positive Movement Leaky Bucket *not to scale and only for the purposes of the demo

S P A R K

S P A R K 1. AI Identify patterns at scale

S P A R K S n o w f l a k e U s e r G r o u p M a r 2 0 2 3 P A G E 6 Slower time to market

S P A R K Feature Store WBB Feature group Weekly Insights WBB Churn model AI/ ML Models Alert & Monitoring Feature engineering Data Lake Manual ingestion & preparation … How we made it faster to experiment? S n o w f l a k e U s e r G r o u p M a r 2 0 2 3 P A G E 7

S P A R K 2. Product Creating delightful customer experiences

S P A R K S P A R K Be Data Centric S n o w f l a k e U s e r G r o u p M a r 2 0 2 3 P A G E 9 Most people relate network back to reliability(n=1320) 92% say being reliable and consistent is most important in relation to network 80% say they consider operators offering highest speed connection "Seems silly to be asking me after I leave. Was not interested in asking me when I said I was leaving. Had ongoing connection issueswhich I went around in circles time after time. Told it would get resolved. Someone would call me back. This happen over a number of years. So, I finally gave up and left after being a loyal customer for many many years" Fair/ Limited Good Signal Strength is Excellent S n o w f l a k e U s e r G r o u p M a r 2 0 2 3 P A G E 9

S P A R K S P A R K Packaging the insights in an easy to consume data product S n o w f l a k e U s e r G r o u p M a r 2 0 2 3 P A G E 1 0

S P A R K 3. Lead Fail Learn Faster

S P A R K S n o w f l a k e U s e r G r o u p M a r 2 0 2 3 P A G E 1 2

S P A R K S P A R K Thank you for your attention manish.kukreja@spark.co.nz S n o w f l a k e U s e r G r o u p M a r 2 0 2 3 P A G E 1 3