Risk-Aware Product Development (a.k.a. Scrum)
"There's no predictability/commitment in Agile/Scrum"
Over the years I've heard my share of these kinds of statements from various levels of executives:
"When my guys run a product development release I really want to know what I will get at the end so I can make business plans accordingly"
"In the old days when we ran projects we knew the timeline, we knew the scope, there were no surprises. These days it seems like the inmates are running the asylum. Product Management keeps telling me in this Agile way of doing things there is no committed scope, no plan, only 'Responding to change'. I can't help but think they are using it as an excuse for thinking short-term"
"When an important new/existing customer asks for a feature I really want to know whether I can commit to him that it will be delivered in the next release so I can get him to commit to a purchase"
"When you talk to people here, I want you to be very clear on the fact that agile will not take away the predictability they have so that they will not oppose the agile initiative we are trying to kick off"
This post is dedicated to you guys out there trying to make sense of the confusion surrounding predictability and Scrum/Agile (Or as I often refer to it these days Risk-aware Product Development), whether you consider yourself one of the people quoted above or one of the people working for them trying to bridge the desire to work in an agile nirvana-like flow and dealing with those pesky executives stakeholders and customers bent on interrupting that flow.
Back to the roots
Let's make it clear up front. I totally empathize with these executives. They are either already victims of "Bad Scrum" or are basing their concerns off of a lot of "Bad Scrum" going on out there. It is such a shame though because in reality Agile product development using Scrum actually maximizes the predictability you can have in your environment. Wait, That was a complex sentence. Why didn't I just say "Agile/Scrum provides great predictability" and get it over with? Because that won't actually be true...
Let's start from the beginning. Let's talk about the contexts where agile approaches are required. One of the first things I discuss with executives is the concept of uncertainty and specifically Stacey's Uncertainty Matrix.
Types of Uncertainty in Product Development
I explain the concept of Requirement/Business uncertainty as the chance that you are aiming at the wrong target and Technology uncertainty as the chance that you'll have execution problems getting there. I then ask the people in the room to classify the projects they are currently working on and typically get a picture somewhat like the one above with different projects having different uncertainty profiles. At this point, it is typically easy to get to the understanding that when you have low uncertainty predictability and success is a matter of careful and disciplined execution of greatly crafted plans.
Dealing with Technology Uncertainty - The Waterfall Passive/Buffered Risk Management Style
When uncertainty starts to mount, it is a different ballgame. Teams facing technology uncertainty with big batch waterfallish approaches only get REAL predictability (meaning the one that stays valid all the way through the release of high-quality product with the committed scope) through huge buffers. These buffers serve to hide those horrible big bang integration hells that are the source of statements like "We are totally on track all the way through development but when we reach the code freeze period all hell breaks loose and we don't trust a word we're told about when the product will be released".
Dealing with Technology Uncertainty - The Agile/Scrum Active Risk Management Style
Agile teams in this context should be able to provide predictability by weighing the amount of effort required to achieve the business expected outcome, taking into account the amount of uncertainty, and make some buffered plans at the high level before going into low-level iterative product development aimed towards achieving those high-level commitments. You not only get predictability of "If we said we will deliver this feature in this release then that is what we will do" you also get visibility of progress towards meeting that commitment by seeing working product frequently and getting reports based on real integrated progress. You use this Transparency to Inspect and Adapt where you are in order to maximize your ability to deliver to your goals. Again, this doesn't mean full predictability - it means maximizing the possible predictability in your complex context.
Dealing with Requirements/Business Uncertainty
Here, the definition of success is actually a bit different than what executives are typically used to. Success doesn't necessarily mean delivering according to commitments. It means hitting the target if it is indeed a real target and alternatively learning as quickly and cheaply as possible if we are aiming at the wrong target. This is not some theory we're talking about. Data from more than 50,000 projects suggests 50% of features developed are actually not used. (See Chaos Report 2013).
Is Predictability what we want?
So aiming at the right target is not a trivial task in Product Development, especially if you are in a more innovative space where it is not even clear that there is a market need for your product, but also when you are working on internal IT projects where this kind of uncertainty seems irrelevant... In the Lean Startup movement, we typically talk about the "Build it and they will come" fallacy. Actually, we are NOT sure they will come, so we want to be very careful with what we build without knowing. Predictability might be a dangerous thing to wish for in this environment.
The Risk Burndown Exercise
An exercise I often use to get this point across is to ask people in the room to draw a chart of the amount of risk/uncertainty in their projects from initiation all the way to the moment all risk/uncertainty has been exposed. the X axis reflects the time or stages in their "Stage-gate process". The Y axis is "remaining risk/uncertainty" in either the Business/Requirements/Technology domains. One answer I typically get is a line curving up at some point only to go down later on. That is impossible. The risk is there. If the line curves up what you mean is that you actually just become aware of the risk at that point. I guess this is the best indication of the fallacy of "predictability" people have. Others start with the maximum risk and then go down quite quickly - telling me that they learn everything they need to learn at the point of "Requirements/PRD/Design" and from then on it is a matter of execution. I then typically ask whether they find surprises/defects later on in the cycle and whether they actually know all of their features are in use. This typically gets them closer to the epiphany... Others get it by this point and draw a chart where a lot of the risk is there active until you finally get to have your software/product in the hands of real users. At this point, the discussion very quickly gets to the main way to reduce the time of active risk - cutting projects/products into smaller pieces so they can get to be "Working product in the hands of users" faster. (Which is, by the way, one of the key differentiating factors of successful projects regardless of the methodology according to the Chaos Report)
A reverse correlation between Business uncertainty and the need for Predictability?
Luckily enough it seems like Predictability is typically required when there is a strong business need on the other end of the line (if there is a partner/customer adamant on having that feature it kind of reduces the business uncertainty of whether it is a real need...). So effective classification of projects into the right uncertainty/predictability profiles will typically help satisfy the business executives without creating unrealistic expectations from the product development group. There's still the need to understand you are sometimes running small "startups" or "venture capital" inside your product development portfolio, which might be tough for an execution-oriented IT/product development executive to fathom. Geoffrey Moore talks a lot about these kinds of struggles in "Escape Velocity". Which is highly recommended reading by the way.
So, with all this in mind, what can you do?
My advice to executives is to make sure they and their teams understand the uncertainty profile of each of their projects and act accordingly:
- When uncertainty is low, use whatever kind of process to deliver the desired outcomes with high predictability.
- When dealing with technology uncertainty and medium/low requirements uncertainty (Sustaining Innovation) and it is desirable to be predictable, use effective agile release planning approaches to setup a reasonable plan with maneuvering space to account for some requirements/execution uncertainty using either a time or scope buffer (a set of features that are considered stretch / extra-weight to be jettisoned in case of problems).
- When dealing with serious business uncertainty (Disruptive Product Innovation space?) make sure that a fast iterative approach like Scrum done professionally with a real feedback loop that enables Empiricism is used to minimize "building it before knowing whether they will come". Learning needs to be emphasized over predictability in these environments.