Agile Informatie Technologie

en hoe Scrum toe te passen

De Scrum optimale sprint duur

Wanneer een Scrum team wordt gevormd moet de sprint duur worden afgesproken. Naar analogie met de formule van Camp kan die worden berekend. In de praktijk kiest bijna ieder team voor 2 weken. Niet het minimaliseren van de kosten van een mislukte sprint, maar het gevoel van urgentie bij een korte sprint blijkt belangrijker.


Het terugvallen naar de gewoontes van de Waterval methode is het belangrijkste gevaar voor een Scrum team. Door te kiezen voor een sprint duur van 2 weken is de kans daarop kleiner. Twee weken zijn te kort om alle Waterval fasen serieel te doen, er moet wel parallel worden gewerkt. Dat de software ook nog moet zijn getest is een uitdaging die het team dwingt de tests direct voor te bereiden en te automatiseren.

Voor de liefhebber hieronder toch nog de berekening die uitgaat van een minimaal afbreukrisico van een sprint:

What is the optimal sprint duration?

Before the start of the project the sprint duration is set. Scrum prescribes a duration between 2 and 4 weeks. Longer durations are discouraged because:
  1. the danger that within the print the team falls back to the waterfall method
  2. the cooperation with the business will diminish
  3. the moments for reflection, for inspection and adaptation in the retrospective and the sprint review occur less
  4. the loss of productivity and the cost of a prolonged project end date when sprints fail

Calculation of the optimal sprint duration

Q: the calculated sprint duration
D: the expected project duration in team days
S: the fixed costs per sprint (sprint planning + retrospective + review + regression test) in team days
h: the project risk as the expected number of sprints that will fail (can be a fraction like 1.5)
f: the team costs per day
p: the business costs or loss per day when the project is extended

Q= S*f*D/(h*(f+p))  and S*f/((f+p)*D) < h << D/Q

Note that this formula takes into account only the fourth criterium: the loss of productivity and the cost of a prolonged project when sprints fail. If the outcome is more than 4 weeks I suggest you take measures like reducing the fixed costs S.

The derivation

This formula calculates the minimum of the total costs Tc consisting of:
  • the fixed costs for all sprints: S*f*D/Q in € or $ en
  • the costs of failed sprints : h*Q*(f +p) in € or $

Tc = SfD/Q + hQ(f +p)

dTc/dQ = d/dQ[SfD/Q + hQ(f +p)] = 0

-SfD/Q2 + h(f +p) = 0

hQ2(f +p) = SfD

Q2 = SfD/(h(f +p))

Q = sqrt(SfD/(h(f +p))) en S*f / ((f+p)*D) < h << D/Q because Q must be smaller than the project duration and h must be much smaller than the number of sprints in the project.

Use the calculator below to experiment:

Enter the values for your calculation:
D =
S =
h =
f =
p =

result

Examples:

D = 200 days, S = 3 days, h = 1, f = €10000, p = €5000 ==> Q = 20 days.
D = 100 days, S = 3 days, h = 1, f = €10000, p = €5000 ==> Q = 14 days.
D = 100 days, S = 3 days, h = 2, f = €10000, p = €5000 ==> Q = 10 days.
D = 200 days, S = 3 days, h = 3, f = €10000, p = €10000 ==> Q = 10 days.
D = 200 days, S = 6 days, h = 1, f = €10000, p = €5000 ==> Q = 28 days!

Note the last example. When the for instance the regression test time takes several days the optimal sprint duration becomes unacceptable long. Automatic regression testing is highly desired for Scrum.

The sprint duration is fixed in Scrum. There are no valid excuses to extend a sprint. So changing the sprint duration is not done. That's why it is so important to choose the right duration up front.

The formula is derived by me inspired by the optimal batch size formula known as the formula of Camp.

© 2006-2018 | Created by Erik Verheul