Feature Driven Development

De Luca on FDD Newsletter Issue 2

How to Reduce the Risk of Fixed-Price Estimates

---

In FDD the first three processes are one-time and form the project startup phase, while the construction phase is incremental. Between them is a key milestone. The FDD startup phase contains predictive metrics to project future effort based on early available indicators. Explaining this to the client allows the initial project estimates to be tested at the start of the project and the scope and duration to be negotiated as a part of the startup phase. At the end of this low-cost entry to the project, the project manager is in a strong position to proceed on a fixed-price basis. In these risk-averse times, the ability to accurately estimate development duration early and with high-confidence is more important than ever.

---

Feature Driven Development (FDD) comprises five processes that separate into two distinct project phases. The first three processes - Develop an Overall Model, Develop Features List and Plan By Feature are essentially "one time" processes and they form a project startup phase. The fourth and fifth processes - Design By Feature and Build By Feature - are incremental and comprise the project construction phase.

At the end of the startup phase, that is, after the Plan By Feature process has completed, you can set a big fat milestone in place that separates the startup phase from the construction phase. At this milestone (and with great knowledge) you estimate, price and negotiate the construction phase. This is a risk reduction strategy for you and your client and you explain this to the client before the fact. That is, before you start the project (in fact, I explain it as a part of the bid - it is a competitive advantage).

Let's look at why by considering just one of the predictive metrics in Feature Driven Development (FDD).

The FDD startup phase contains predictive metrics to project future effort based on early available indicators. One of the most useful and important predictive metrics comes from the Develop an Overall Model process. This metric says that for every two weeks of time you spend doing the modeling in the Develop an Overall Model process, there is six months of project construction phase time. That is, one week of Develop an Overall Model modeling is about three months of construction phase time, two weeks of modeling is six months of construct, three weeks of modeling is nine months of construct, and four weeks of modeling is about a year of construction phase time.

Therefore, if a client comes to you and says this is a six month project then you know you should spend about ten days in the Develop an Overall Model process (two weeks is ten working days). What this means is, for example, that by day seven (or thereabouts) you will know if you're in the ballpark or not. Using the six month example, if around day seven it is clear there is another week or so left in the modeling, then you can have the discussion with the client right then. Remember you have explained this predictive metric to the client before you started and the client has been in the room right with you as you've been doing the modeling (and the requirements and requirements analysis that goes along with object modeling). You can say something like "well, it's day seven and it's clear there's about another week and a half left in the modeling. This is a nine to twelve month project. What do you want to do?" This is a great position to be in as the project manager. The client has to either cut scope or "kaching!" (translation: ring the cash register). This approach takes all the confrontation out of it. There is no us and them, there's just "all of us." Together, you and the client have the knowledge to make an informed decision about scope and its very real implications.

Now remember that this is day seven of a six month project and you are able to make this informed decision that early (a predictive metric based on early available indicators). You haven't even cut any code yet.

Visibility. Early available indicators. Predictive metrics. These are all gifts for project managers and FDD has many of them - because it comes from the experiences of practitioners.

Having completed the startup phase you are in a great position to estimate a fixed-price construction phase at a nice big fat project milestone that separates the startup phase from the construction phase. You have all the knowledge gained from the first three FDD processes. That is, you've completed all the modeling and requirements analysis that goes along with it, produced the features list - which itself is a complete statement of the requirements, and then planned the entire construction phase accounting for dependencies, class and functional area complexities and pervasiveness, and for the skills of the developers on your team.

This is a terrific base of knowledge from which to estimate a fixed-price construction phase. Given the predictive metric from the Develop an Overall Model process (six months of construction time for every two weeks of modeling), and the fact most projects are a year or less these days, the startup phase is therefore typically a month or less. Often, the startup phase is within two to four weeks.

To size the startup phase you take the modeling time and add about a week to cover the time to do Develop a Features List and Plan By Feature (it's important to remember that the predictive metric based on modeling time in Develop an Overall Model works in both directions; either to predict construction time based on completed modeling time or, as in this case, to predict modeling time based on initial estimates of overall project size). If the modeling time is four weeks, then add two weeks for Develop a Features List and Plan By Feature. Only drop this extra time below one week if the modeling time is less than one week.

You sell the two to four week startup phase as a high value, low cost, low risk entry for a project, after which you can promise a fixed-price construction phase with a corresponding commitment plan.

Clients like this!

You can choose to sell the startup phase as fixed-price itself or as time and materials. The predictive metric lets you get the construction phase sizing right and explaining this to the client at the start, allows you to vary the startup phase as you go based on actual time taken to do modeling. Remember that the startup phase is small - a week here or there is not a big deal relative to the size of the overall project. Explain the process to the client before you start and buy yourselves (you and the client) the flexibility to adjust the startup phase "in flight."

There are several other things I do within the startup phase that are not directly addressed in the FDD processes. I'll write about them in a future newsletter.

With thanks to the review team: Phil Bradley, Gavin Baker. The inital observation of the relationship between modeling time and project time comes from Peter Coad's 20 years experience as one of the world's best object modelers.