Most of the people I interact with are unequivocally excited about the opportunity to try agile for the first time. Not.
Most people have doubts. And one of the top questions I often get is "can I seriously use agile for a very large project with multiple work streams, and with cross-impacts from other groups who aren't agile at all?" Yes! Yes, you can. In fact, agilists and the agile approach has a particularly strong set of tools for helping you manage cross-team interactions. These tools not only help you scale your "pure agile" project management approach, but also can help strengthen best practices in your "waterfall" PM approach, if you're working in a hybrid environment. So let's talk about that.
First, a few myths that you might have heard, but I'm sure you agree aren't true.
- Agile doesn't scale. Um, wrong. Agile Scaling has been around for almost as long as agile itself, discussed, for example, this year by Tom Grant, in his blog on Forrester, and supported for ages before that by well-known luminaries Scott Ambler (Agile@Scale) and Dean Leffingwell (Scaled Agile, Inc.). Myths about non-scalable agile are just there to prevent poor, deprived enterprise IT people from enjoying their rightful standing in the world of agile. You're falling prey to agile snobs who prefer to work on small teams for small companies with bean bag chairs, and they are successfully trying to make you feel bad for working in a big company. Sometimes it's hard to be Goliath.
- If you do planning before you start coding, in order to support agile at scale, that's not agile. Wrong again, in my humble opinion, and of course that of many others.
But let's say I concede the point, and agree that every single practical thing that you might do to make agile work at scale makes it not-agile. Let's look at what you can do within a multi-stream project which has agile elements (that would not be possible in a classic waterfall implementation).
- For the work you will be able to manage internal to the scope of your own project, and will therefore be able to handle in an "agile" manner, you have divided up the needed features into small enough pieces that you can see in a very granular way whether it's possible to create parallel work streams who won't step on each other by touching the same parts of the code base simultaneously, or ever.
- Agile release planning calls for partitioning the work across the teams after you understand the architectural direction and the full feature set, but before you even start detailed analysis.
- If you have an insufficient number of people with analysis, development, or testing skills to keep development moving in multiple streams, you know that ahead of time, rather than hiding those details in the "batch" mentality of delivering all requirements before starting all development.
You can plan integration strategy with all cross-impact partners proactively. Effective agile teams will inventory their data interfaces during release planning, and for each interface, determine:
- Is the integration partner already done? In this case, we build to their specification.
- Is the integration partner going to start work after we're completely done? In this case, they will build to our specification, and we can document where we ended up later, so long as it's in time for their process.
- If the partner is building in parallel with us, targeting the same release? Then the next question is, is the integration partner doing their needed new work in an agile or waterfall methodology? For waterfall partners, agile teams need to develop detailed requirements around the interface before starting subsequent work, so that the agile team can mock and stub the partner behavior correctly, even if the partner doesn't have time to collaborate beyond that.
- But better, does the integration partner want to do proactive interface testing with you, and could that be scheduled ahead of the official "Integration Test" period for the release? You and your partner teams can also make proactive arrangements for mutual stubbing or mocking of code, developing test sets reciprocally (you provide test strategies around your code as it interacts with their system; they provide the obverse to you).
You can still do meaningful program-level roll-ups showing progress across the disparate work streams, so long as you agree to use business-friendly metrics like "actual code complete" rather than "points burnt up." And in fact, since the work streams which progress in an agile manner will get to the coding sooner, the program-level dashboard becomes a showcase for the delights of agile over waterfall.
Please don't let the agile-for-tiny-teams-only people get you down. Corporate moguls running ginormous agile programs world-wide, unite! You have nothing to lose but, hm, okay, well, it's actually quite good to be a ginormous corporation. And agile scaling just adds to the benefits.