How To Share An Agile Development Team
But all too often, methodologies focus on a project environment, where the team is focused predominantly on a shared goal. Where the team is largely dedicated to the project.
In reality, this is often not the case.
In reality, development teams are frequently required to develop and support multiple products. Multiple products with multiple product owners. And particularly in 'business as usual' ongoing development.
So how do you share an agile development team?
Operating as a resource pool is often not ideal. Everyone throws their requests over the wall. Those that shout loudest get the team's attention. Or maybe the bigger products get the team's attention, at the expense of smaller products that never get to the top of the list, and never will.
Splitting the team by product sounds great. But sometimes the team's too small for this approach to be practical. Or it leaves too many individal developers, causing problems with cover and you can't exactly collaborate with one developer per product!
So what can you do?
We have a similar situation in lots of our teams. Broadly speaking we are solving it like this:
- There is one Product Owner per product. The Product Owner maintains a separate Product Backlog for each product.
- The development team acts as one team.
- The Sprint Budget (number of man-hours available for a Sprint/iteration) is allocated to each product based on our recharges, e.g. 60%-20%-20%. If you don't recharge, you could agree this at a more senior level as a general rule of thumb. Each product has a known % of the budget for each Sprint.
- Use Cases/User Stories/features are broken down into Tasks and estimated by the development team during Sprint Planning.
-
Each Product Owner can only include Tasks from their Product Backlog up to their allocated % of the Sprint Budget.
This approach means the development team can act as one team. There is knowledge sharing and cover when someone is off, because in this case the overall Sprint Budget is reduced but everyone still gets their usual % share of the available hours.
It also means team members don't have to juggle their time between products on a 0.x FTE basis, which is awkward at best and just plain impossible when the fractions are too small or odd numbers.
Instead the Tasks allocated in the Sprint are already appropriate to the Sprint Budget per product, meaning team members can focus on delivering the Tasks in the Sprint, not worrying about how to split their time.
See also:
10 Key Principles of Agile Software Development
10 Good Reasons to do Agile Development
Top 10 Agile Development web sites
18 July 2007 18:14
Hi Kelly,
Could not help to leave the comment as the topic is quite familiar to me. We tried to address the challenge of accurate iteration planning, when the same team members sharing their time between multiple products, in our Yoxel SW agile product management solution. Although in a slightly different manner.
In our case for each developer the tool allows you to set % of time he will be spending (or usually spends) on tasks from each product. This time sharing could look like 50% on ProductA, 30% on ProductB, and 20% on Other. Quite similar to what you're saying, right?
Now when you start a planning session for your next iteration for ProductA the 50% factor is used to adjust perfect time estimates (entered by the developer) for his tasks. This way you actually get realistic estimate in the context of this particular product/iteration and time sharing scheme.
The same thing happens for ProductB, with 30% time spent on this product the realistic estimate will be that much longer than the perfect one. (9hr will become ~27hr)
'Other' means time spent on critical or unplanned requests. You can think of it as 1-[person's individual velocity] :)
Nice post, I am surprised I do not see this topic discussed anywhere else. From my experience this is a quite common situation when developers participate in multiple projects for different products.
-Alexey
26 November 2007 01:44
The way you describe in your post is roughly how we go about things, too. The problem we've run into is how to determine the order that we work on the various stories. In our domain, we have stories against a legacy system that tends to blow out our time estimates from time to time. That means we occasionally have to drop less important stories off the bottom of our board. You don't mention an absolute ranking across backlogs, but would that be a solution to this issue in your eyes? If not, how would you handle this?