Ho How to to me measur ure pr produ ducti ctivity ity in in softw twar are de development nt Traian Kaiser, GoTo Prague Conference, November 2011 2
Traian n Kaiser er Director Agile Project Management & PMO (XING AG) 13 years of Internet-Business Companies: IBM, Yahoo, XING, … MBA, Agilist, People Manager
~140 Employees in product development 50 50 Releases a year
Braint ainteaser (ge get started) 6
Wh What makes the the difference?
vs.
vs.
Produ duct ctivity ity (overall all) 10
what coun ounts ts Out Outpu put Surplus Producti uctivity ty Value Inpu put What is is ne neede ded
2008 2009 2010 Revenue (Output) 34 Mio. € 44 Mio. € 54 Mio. € Out Outpu put Surplus Producti uctivity ty Value Inpu put 2008 2009 2010 Cost (Input) 23 Mio. € 33 Mio. € 38 Mio. €
60 45 Productivity lack 30 15 Input Output Output same productivity 0 2008 2009 2010 Overall Figures 2008 2009 2010 Revenue (Output) 34 Mio. € 44 Mio. € 54 Mio. € Cost (Input) 23 Mio. € 33 Mio. € 38 Mio. € „Surplus Value“ 48% 33% 42% Revenue with 48% SV 34 Mio. € 49 Mio. € 56 Mio. €
Produ duct ctivity ity in in softw twar are de developme ment nt 14
Value of developed and distributed software solutions for business requirements (including quality and time). Outpu Out put Producti uctivity ty Inpu put in SW W de devel elopment Cost of created and distributed software solutions trough people, services and other resources needed.
„Do the right things .“ Effect ectiveness Effi ficiency „Do the things right “
Business Models Time to market Do o the the right th things Market knowledge Domain expertise Innovations Coordination- & communication efforts System cost Service Costs Do o the the th things right Staff costs Development efforts Maintanance efforts
“Software development is a creative and complex process. Effectiveness and efficiency can not be optimized independently of each other and the interactions are not detectable in their entirety.” Traian Kaiser
Ex Examp ample 1 1 – Discovery Process Balancing effectiveness vs. efficiency by investing time & money to find out if our user benefit from what we plan to do
Ex Examp ample 2 2 – Staff ffin ing Performance differences between mediocre contributors and „ artists “ can be huge. In addition, working team patterns and team productivity is another game to play.
Examp Ex ample 3 3 – Opti timiz mization Prob oblems Proc ocess- Capaci city Reor eorgan anizatio ion Highest Productivity opti op timiz izatio ion increase Lowest Cost Slack? Best Service Time per ticket Cost Tickets per Team Member *Real project example by David J. Anderson from the Book „ Kanban “
How can we Ho we me measur ure pr produ ducti ctivity ty in in softw twar are de developme ment nt 22
Redu educe coor ordin inati tion on and nd comm ommuni unicati tion effor orts ts Valu lue add addin ing g efforts ts to “the right things” Redu educe „ Waste “ (eff fforts ts wit without value)
7 Types of Waste to avoid… … and how to measure them… Parti tiall ally Done one Work ork (Blocked Cards/Tickets/Stories) No Not used ed Document mentation n Ex Extr tra a Featur ures s (Feature- / Module usage metrics) Task Swi witchin hing g (Work In Progress) Hand ndoff ffs s (Cycle Time) Delays Del s (Lead Time) De Defect cts s (Failure Load, Code Metrics, Test Coverage, Bugs per Line of Code, … )
Communication & coordination to avoid … … are they measureable?… Insuffi ficent nt or or e excessive e pr process s / frame mework Uncl clear ar or or inade adequ quate division of or organi nisation units, , role oles and nd respon onsibilit ilities Too many depen dependencies of all ki kind nd Value Stream Mapping Performance to Commit / Deadlines Productive Time Analysis Throughput / Velocity Variance
Value adding efforts to increase… Add ddit itonal al Sold Sold / used ed pr products cts Milestone /Project KPIs Team KPIs User Impact Indicators Custom omer Sati tisfact ction User Ratings Number of Customer Complaints / 1000 Customer Net Promoter Score
Be Bewar are: : Effi fficiency me metr trics s can be be dysfun unct ctional al As Assumed med effi fici ciency me metr trics cs Working hours Utilization Lines of Code Rel elative me metr trics cs & & me metr tric com ompari rison Velocity Me Metr trics as as repl eplacement nt for or tr trust All Metrics
What doe does XING G me measur ure 28
Custom omer Sati tisfact ction Net Promoter Score Number of Customer Complaints / 1000 Customer User Test in our own lab User Ratings (platform & mobile apps)
Key Perf erform orman ance inde ndexes by by dedi dedicated Team Revenues Customer Base Activity rates Other key indexes who drive our business (vary by Team)
Qu Quali lity ty Me Metr trics Code Metrics (just started) Automated test coverage Bugs per Team / Overall
Cycl cle e Time me & L Lead d Time
Visuali alizati tion on is is a a powerf rfull ull instr trum ument nt Inventory Started Designed Coded Complete 240 220 Points 200 180 ory Poi 160 140 Cycle Time 120 100 Stor 80 WIP 60 40 20 0 *Example by David J. Anderson
Con oncl clusion 35
Productivity metrics can be dysfunctional. Consider Recognize Used correctly as software dependencies feedback channel, certain development as between metrics could provide a creative and effectiveness important information. complex process and efficiency Coupling effectiveness and efficiency metrics is essential.
Thank you ou Ques Questions? Traian Kaise iser r Director Agile Project Management & PMO XING-Profile: xing.to/traian See what‘s up in our Lean Software Development Group: http://www.xing.com/net/leansoftwaredevelopment/ The Professional Network www.xing.com XING – Powering Relationships 37
Recommend
More recommend
