08.07 Moneyworks Data Analysis
- 1. Moneyworks Data Analysis Recommendations A. Continuerefining existing systems 1. Spend more time & resources increasing timestamp consistency & accuracy. 2. Establish data analysis/data mining weekly needs to inform future app upgrades. B. Maintain as much homogeneity between testers as possible. 1. Onboard successfully. a) Establishtechnical difficulties point of contactin initial email. b) Stress needto keepapp on during the whole day. 2. Stay interactive. a) Starbucks check-in:awesome!Fun suggestions keeppeople engaged. b) Ask with eachnew rollout what people have been thinking.
- 2. Data References On multiple accounts, user and system data did not always match up. MW was set out to measure activity every 15 seconds. It cannot be determined whether there is a specific system glitch or user behavior breaking this parameter but there are spans as large as 3 hours 29 minutes (Jonathan Morales, 7/30) and small as 4 seconds (Timothy Prentice, 08/05, 23:17:02 to 23:17:06) between measurements. When assessing user/system accuracy, small time intervals are very important. Jonathan Erickson’s times seemed the most consistently small (every 9-10 seconds), while even users without possible user behavior interference show inconsistent measurement times (Sammy Puga ranges from 9 seconds to 2 minutes and 15 seconds). A possible explanation could be with Android’s confidence levels on user activity. Further analysis is warranted. As data analyses continue, analysts can focus on specifities. Being the first analysis, MW can gather a general sense of how user and system data are generally merging. Going forward, MW can decide if it wants to focus further on studying accuracy of specific times (breakfast, lunch, weekends, Mondays) and/or activities (still, still/walking, driving). Boiling down specific data times/user activities will enable a good foundation for future MW user trend reports. Of the six users (Jonathan Erickson, Jonathan Morales, Mark Shepherd, Michael Argano, Sammy Puga, and Timothy Prentice), data cannot be analyzed for two. While MW is successfully receiving Michael Argano’s system data, there is either a technical difficulty or the user is not submitting his user logs. Mark Shepherd is just the opposite: the user data is coming in fine but there is a dirth of system data. I recommend having a check-in after 3 days with every new tester to make sure everything is functioning properly. Next Steps 1. Clean up analysis sheets to minimize excess system timestamps 2. Follow up with Mark Shepherd and Michael Argano to see if we can help somehow with data collection. 3. Ask all testers to keep app running in the background throughout the day. 4. Dig into the data! Study confidence levels with MW status versus actual status.