Productivity day 2013 measuring collaboration quality
1. Seven steps to success in
managing collaboration quality
Jukka Koskenkanto
CEO
Cloudriven Oy
2. Step #1: Enable collaboration
Publish objects that may be
social
Enable communication
about the objects
Connect objects to
processes
Use also other
arrangements to support
collaboration if possible
5. Case: Forecasting B2B sales
Forecast # of weekly deals
Keep the model simple
Create reliable model
Customer
request
6. • Reliable forecasts
• New approaches to sales
work and sales management
Case: Forecasting B2B sales
Customer
needs
7. Step #4: Data analysis
Analysis vs. repeated guess
Coverage
Why people do what they
do?
8. Case: Forecasting B2B sales
Problem
mathematically
Variables
Model
Results and
conclusions
9. Step #5: Present the results
Keep it simple (at least on the first slide)
Use the language your customer understands
Be prepared to meet the expectations of critical
voices on next slides
10. Case: Forecasting B2B sales
Key message 1:
Forecast # of weekly deals with high relaibility
and small variance
Magic formula shared with the sales controller
Example forecast:
75 80 85 90 95 100 105 110 115 120 125
Forecast:100
11. Case: Forecasting B2B sales
Key message 2:
Pay attention to behaviours / actions which will
maximise
Number of created proposals
Number of high quality customer & internal discussions
between the proposal date and the decision date
Change your management system by organizing
time for those actions.
Do not include action ”modify proposal in the
CRM” in your sales management model – keep it as
an indicator
13. Case: Forecasting B2B sales
70 % might be 80 or 90 % by eliminating the affect
of vacation periods from the model
It might be possible to specify which kind of
discussions produce highest rate. What effect does
it have to the model if we
met decision maker,
met influencer,
organized a meeting where another customer told how
we did in the project,
Replied to customer call,
called customer,
sent an email or
sent an email with attachement?
People connect via theobjects to eachotherCo-createdocumentsDiscussabout a blogpostSolve a problemInenterpriseworldsocialobjectsshouldberelated to processesgenerating business value -> processes and actionsrelated to themarerelativelyeasy to manage. Managingcollaborationthat is notrelated to anyprocessgeneratin business value is veryhard and questionable
If youdonotknow, whenwhodidwhatregardingtheobjectyoucannotbeaware ofcausalconnections.In practice: keepthe version history of differentobjects…This is themostsimpliestform of big data withinenterprises
If onehasthe data butaccess is grantedonlybased on IT department’scontrolneeds -> thevalue of the data is minimizedCustomer’susingour software hasextensive data storagethatcanbeaccessed via APIs
Casecustomer: Offersprofessionasservices to companiesVirta CRM userorganization – CRM in use as ourcustomerstend to haveDifferentofferingsDealvalues and salescyclesdiffer
If youunderstandtherootcauses of a problem, whydon’tyoufix it?
Plenty of ”rules of thumbs” drivebusinesses. Somtimesrule of thumb is goodenough and there is no point to carrythecost of buildingrealunderstanding. However I canguaranteethattheanalysisshouldbe made by a professionalmuchmoreoftenthan it iftheprocess in question is producingorconsuming business value (and volumesarebigenough)Doestheanalystunderstandthe business problemwellenough to highlightvariablesthatmightstillimprovetheresults?Doestheanalystunderstandhowthenumbersarecreated? Whichactionsresulted to thosenumbers? Whypeopleactedlikethat?
Time seriesmodeling, whichrequiresdeepunderstanding of modeling -> eg. Autoregression and multicollinearitymaycausesignificantsystematicerrors to thefinalmodel in wronghandsIndependentvariablesexplainswhydependendvariablesgettheirvalues – in modelingprocessunderstanding of thereal life business problem is veryimportant -> avoidegg – chickenproblems, introducethemostrelevantvariablecandidates for themodelTriedeveryaspectthatmightberelevant. Variablesbased on systemgathered data, variablescreatedbased on systemgathered data and variablesfrom outside world (liketrendbarometers)Foundthemodel –regression modelbased on twoexplainingindependentvariables. Testedthemodel in case of multicollinearityor autoregression exists. Got results… Someremarks, withoutputtingcustomerintimacy in danger… Coefficient of determinationmeaningvariablesability to explainthenumber of deals per weekwas 38 % with 1 tradiationalvariable (howmuchstuffyouhave out there). By addinganothervariablewhichreflectssalesrepsbehaviour, thecoefficient of determinationraised to 71 %, whichcouldbestillimprovedbyignoringthe summer vacationweeks in themodel. Whatwastheothervariable? It wasthetotalcount of modifications of proposalsfromwhichthecompanywaited for decisionfrom a customer. Whatdoesthistell to us? Well, salesrepslikeall of us arelazy. If there is no realreasonwewillnotmodifyourproposalsmarked in the CRM (and in someorganizationsthediscipline to addtheproposals is actuallymissingtotally). Themodificationsrepresentquality in customerorinternaldialoguethatwilladdsuchinformation to theproposalthatthesalesrep is reallywilling to usehis/hertime for editing it in the CRM.
Independentvariables, dependendvariables, multicollinearity, autoregression, coefficient of determination. Whatdoesthatmean in practice?
Extendyourunderstandingwider (doeseg. Customer, contact person orsomeotherdemographicvariableaddsomething to themodel -> shouldwecreatenewvariablesbased on the data?)Can somevariablebestilldivided tosuchthathavebettercapability to explainthenumber of deals?Runmorestatisticaltests