When discussing the concept of a part time team member, I find myself constantly trying to quantify the cost of the 50% split against the benefit of the full time team member. So, I decided to go back to a blog post from a few weeks ago (The 50% Team Member) and do a more mathematic treatment of the situation (for those who like to see this in numbers).
So, a caveat or two first — number one, I am not a mathmatician. If you are and want to challenge my assumptions or, even better, back them up more fully, please feel free. I’ll modify this treatment with any valid comments. Second, I still have more research to do to provide support to my assumptions. So, if you can add references and proof, feel free to draw them to my attention. As soon I can add references and proof to my assumptions and statements, I will.
In the meantime, here’s a somewhat detailed treatment of the cost of the 50/50 team member.
If start with a grand total of 52 weeks times 40 hours, we get 2,080 hours per year. If we subtract 10 holidays and 10 vacation days, we start the actual utilization calculation at 1.920 hours (that’s the original 2,080 minus 80 hours of vacation and 80 hours of holidays.
So, our employees begin with a maximum of 36.9 hours per week (1,920 hours / 52 weeks). Now, let’s assume a 90% utilization (I know it’s high, but I want to be conservative). That means that we lose 4 hours per week due to simple administrative overhead.
We’re down to 32.9 hours per week (1,712 hours / 52 weeks). Now, let’s assume that our 50/50 team member is on, of course, two teams. We’ll be nice and assume that they only have to attend part of both team’s Sprint Planning meetings, but they have to attend both daily Scrums every day and both Sprint Reviews. I’ll give on the Retrospective and allow them to alternate between them each Sprint. So that’s 8 hours for planning, 13 days of two daily Scrums (that’s 30 minutes per day), two hours total for both Sprint Review meetings and 90 minutes for a single Sprint Retrospective meeting. Now, since we can hold a maximum of 17 Sprints a year, that’s 18 hours per iteration or 306 hours per year.
We’re down to 27 hours per week (1,406 hours / 52 weeks). If our team member is really 50/50, we can assume that he or she is optimizing this time as much as possible and the context switch occurs at noon (the morning with one team and the afternoon with the other team). Going with a VERY conservative 15% overhead per switch and two switches (that’s the extra overhead our 50/50 team member experiences walking away from one set of tasks to figure out where he or she was in the other set of tasks), we lose an additional 8 hours per week to context switching.
We’re down to 19 hours per week (990 hours / 52 weeks). One last subtraction. The average Scrum team (unless they are working on new product development), experiences a minimum of 25% of their total capacity support during each iteration. In many teams, the number is much higher. We’ll go with 20%, just to be nice. That’s 20% of each full time member. If we assume 20% of a 50% team member (again, to be nice), we’re assuming 4 hours per week.
We’re down to 15 hours per week (782 hours / 52 weeks) So, for each team, our 50/50 team member can spend, perhaps, 8 hours per week.
If we quickly do the same thing for the 100% team member, we see 1712 hours – 238 hours (iteration overhead) – 416 (defects) = 20.3 hours per week.
So, while a full time member can spend roughly 20 hours per week on a team, a 50/50 member can spend only 16 total hours on two teams. On a weekly basis, we’re losing at least 2 hours of productivity. And that’s just the simple math. When we start factoring in the frustration factor (which can reduce our 50/50 team member’s practical utilitization even further, not to mention what it does to everyone else on the team), we can start seeing even uglier numbers.
For example, if we assume a 5% reduction in performance due to team members that are sometimes on the team and sometimes not and, further, assume two Scrum teams of 7 people where one person is 50% on both teams, we see our 50/50 team member’s performance reduced by 5% due to frustration moving back and forth between the teams (now we’re down to 14 hours per week for the 50/50 team member) and 12 other team members’ performance reduced by 5% due to the 50/50 team member not being in the team room at a time when their input was needed and further complicated by the impact of the 50/50 team member’s input being offered after the team made a decision, forcing the team to reconsider a decision. Now, each of the tweleve team members (that’s the other 6 on each team) is reduced by 5%. That’s a total loss of productivity of 13 hours per week. And that’s only a conservative number (5%). Tom DeMarco puts the impact considerably higher in case studies reviewed in his 2001 book, Slack (this is where I want to do more research — starting by re-reading the relevant section of the book — if you can provide detailed references and support, let me know).
Therefore, finally, on an average sized Scrum team of 7 people, where one person is 50/50 split across two teams, we lose a minimum of 2 hours per week (104 hours per year) for starters. Using similar case studies, conservative estimates are actually suggesting that we’re losing, weekly, 3 hours of the 50/50 person and 13 hours for the rest of the team. That’s 16 hours per week or 832 per year (or, to put it another way, we’re wasting 5 man-months to get one person on two teams).
Wouldn’t it be easier to invest in some cross-training?
