Analytics for Developers

A talk I gave at Kansas City Developer Conference 2013.

Transcript

1. Analytics for Developers and Developing for Analytics
2. About Me 2006-2011: Educated Accounting & Finance 2011-Present: Reeducated Marketing & Operations Twitter: @trent_hauck Work: @AlightAnalytics Other: Contribute (now and then) to Pandas & StatsModels
3. Two Parts Analytics (more) Development (less)
4. Why should you care?
5. “In God we trust; all others must bring data.”
6. To do analytics you need x
7. Where x is data collection...
8. Site Analytics Should be a 1st Class Citizen of Development
9. Collect More Than You Need Now
10. Now Some GA Code <script type="text/javascript"> var _gaq = _gaq || []; _gaq.push(['_setAccount', 'UA-31465642-1']); _gaq.push(['_setDomainName', 'trenthauck.com']); _gaq.push(['_setAllowLinker', true]); _gaq.push(['_trackPageview']); (function() { var ga = document.createElement('script'); ga.type = 'text/javascript'; ga.async = true; ga.src = ('https:' == document.location.protocol ? 'https://ssl' : 'http://www') + '.google-analytics.com/ga.js'; var s = document.getElementsByTagName('script')[0]; s.parentNode.insertBefore(ga, s); })(); </script>
11. Next Steps Events _gaq.push(['_trackEvent', 'Cat', 'Act', ‘Label’]); Custom Variables _gaq.push(['_setCustomVar', 1, ‘key’, ‘value’, 1])
12. Where x is data analysis...
13. Differences in Data Small Data == Math Problem Big Data == Engineering Problem
14. The Math Problem
15. Descriptive Stats (please compute these) Max, Min Quartiles Mean Variance Mode
16. Web Stats are Easy A user converts or not... what are the chances of that? p or q (=1-p) 3 users convert or not... what are the changes of that? p^3 or (p^2)q or p(q^2) or q^3
17. Hypothetical Worlds Trials = 100, Size = 100, p = .1
18. Back to real world Stats: p-bar = .08, SE = 0.027
19. So then AB Testing 500 Trials A B p 0.1 0.2 SE 0.01 0.01 95% CI .1 +/- .02 .2 +/- 0.02
20. The Engineering Problem
21. Build Data Pipelines • Repeatable Flows of Data • Handles Initial Analysis For You • Literate Programming
22. Programming For Data Analysis • Scripting good for Discovery • Larger Jobs need Types • Mapping high dimensional space to lower dimensional space... then add
23. Where x is visualization....
24. Visualization Types •Distributions •Comparisons •Time Series •Other (Match Domain)
25. Distributions Single Variable: Histograms Multiple Variables: Scatter plot
26. Comparisons Categorical Variables
27. TimeSeries X Axis is Time
28. Match Domain with Analysis
29. Where x is storytelling...
30. Storytelling
31. 3 Temporal Stages 1. What happened 2. What is happening 3. What will happen (Plus a tease)
32. Start With the Simple Stuff Friday Saturday Sunday 40º 42º (Why do I live in KC)º
33. Build Up to Complex Idea
34. Thanks... Questions?