Lean LaunchPad: Analytics Workshop

  • Published on
    12-Jul-2015

  • View
    3.710

  • Download
    2

Transcript

MixPanelLean LaunchPad Workshop:Defining an Analytics StrategyRyan JungHaas MBA 2014ryan_jung@haas.berkeley.eduWhy Are Analytics Important?Failure to define an analytics strategy can be a fatal error for a startup in 2015.Analytics has changed the landscapeA great analytics strategy is tightly integrated with the overall business strategyWhy You Need an Analytics StrategyLearn faster by creating feedback loopsMore clarity based on behaviorConsensus on future actionThere exists a host of tools to help you with these objectives.3History of Analytics1990s Web counters2000s Click Analytics and SEO2010s Behavioral and Predictive AnalyticsKeys to a Great Analytics StrategyTightly integrated with overall business strategy Iterative processMeasurable set of hypotheses, results, and revisionsStarter: Most companies report not getting enough insights from their data. Why?5The Modern Data-Driven Lean StartupGoal is to optimize a set of business objectives in a logical progression leveraging quantitative and qualitative facts in order to delight customers in a scalable, repeatable fashionMost Important ReportsSegmentation (Cohorting)RetentionFunnelsRevenue TrackingMarketing Campaign EffectivenessPath AnalysisNotifications7Segmentation / CohortingWhat segments are getting what value out of your product?8Value Proposition / Customer SegmentWho is our customer?What problem are we really solving for them?Will they buy from us?How do we reach them?Build customer archetypesAdd properties to define the userUse segmentation to look at differences in customersGood for looking at actions, but need to understand causation to be actionableUsing AnalyticsCut text in half in Using Analytics9Segmentation ExampleLook at aggregated events and then segment by propertiesSee who is doing particular actions and identify trendsWant to segment as far as possiblePoint you to needs and how your product adds valueGoogle AnalyticsAbout 1/3rd of text here, and pick 2-3 points10RetentionWho gets the most value out of your solution?How Churn affects LTVLifetime ValueMonthly ChurnSource: David Skok Matrix PartnersThinking Through RetentionGet > Keep > Grow = Activation > Retention > Engagement Understanding key featuresUnderstanding core users and testing their needsIdentifying most effective channelsRetention ReportsIn-session retentionIn-app retentionKey Question(s)Where do users spend their time in your app? What features are valuable?Are users coming back and using the app repeatedly? Who are users that are more likely to come back?Value PropositionFeatures that are most valuableUsers that get most value out of product ToolAddictionRecurring or Segmented RetentionMixpanel14BIG IDEA:LTV drives CAC which drives channel selectionIncreasing Sales ComplexityLog(Acquisition Cost)CAC < LTVFunnelsHow are users interacting with your solution?Sales FunnelsWhere are we losing customers?How do we know if we are doing well or not well in sales?How can we do better?Core Idea: Track conversion rates between levels of funnel to see where leakage occurs and create strategies to minimize this loss.Is my marketing spend being used efficiently?Funnel ReportsLocalyticsFunnel ReportsKISSMetricsTying funnels to revenuesRevenue = installs x [signups / installs] x [purchases / signups ] x [revenue / purchase]Back-end tells you thisAnalytics tells you thisAnalytics can tell you thisYou control thisThe main point here is that you can break revenue into measureable componentsTie how you earn revenue to what you measureThen understand where you are doing well and not wellThen use your analytics solution to design tests to figure out how to drive more lifetime valueMathematically:Pitfalls to AvoidProblemExplanationSearch vs. Execution MetricsAre we measuring KPIs or are we testing hypotheses? Vanity metricsIf it only goes up and to the right and / or if its not actionable, its a waste of time to measure it.Biased testsBe sure that the hypotheses that you are testing are not set up to confirm your assumptions. Take the approach of trying to disprove your hypothesis.Data overloadMeasuring everything and then mining for insights creates too much noise for most to get any real value from.Move to back of deck21SummaryYou need to be thinking about analytics because your competition probably already isAnalytics is evolving, so keeping up is imperativeAnalytics needs to be tied to your overall business strategy, should be hypothesis-driven, and is an iterative processCase StudiesAirbnbChallenge: Initially wanted to optimize booking flowAllowed them to identify to distinct classes of usersCan better target users and their needsMore info: https://mixpanel.com/case-study/airbnb/Khan AcademyChallenge: increase engagement and the rate at which people learnUsed funnels to optimize search and registration processesStart with a definition for user engagementMore info: https://mixpanel.com/case-study/khanacademy/JawboneChallenge: Assess the viability of Jawbone UPUsed Segmentation reporting to better understand their usersHelps to build customer archetypesFaster iterations and faster time to product-market fitMore info: https://mixpanel.com/case-study/jawbone/Cohort analysisRenewal and upsell ratesReturn on marketing investmentRevenue by Cohort Each Year Builds on a Stronger BaseNote: Excludes inorganic growth. 201120102009200820072006Highly Loyal Customers2007 CohortEarlier Cohorts We meet our customers needs so well that 28Revenue by Cohort Each Year Builds on a Stronger Base20062008 Cohort2009 Cohort2010 Cohort2011 Cohort20112010200920082007Highly Loyal CustomersNote: Excludes inorganic growth. 2007 CohortEarlier Cohorts We meet our customers needs so well that 29Chart11000005000020000LifetimeLTV vs Churn Rate