A Recipe for Green-Field Software Development

[Re-post from Nov, 2008]

Developing Your Product and Your Customers in Parallel

I’ve been a member of a software development team since 1985.  That’s 23 years as a software developer.

Like many developers who have been around for a few years, some of my grey hair can be attributed to having worked on hellish projects, or on projects that failed outright.  Over the years, I’ve gradually added to my mental list of “worst practices”—things that tend to lead to project failure, or at least hide a failing project until it’s too late to turn it into a successful one.

It’s much easier to compile a list of worst practices than it is to pick some “perfect” development process.  But worst practices can lead to best practices simply by avoiding the bad practices.  At a minimum, we should at least avoid making the same mistakes over and over again.

If I had to pick a single worst practice (there are many), it would be this:

Not building the product that the customer really needed

This happens all of the time.  We (the developers) build a product that is bug-free, efficient, scalable, and does exactly what we intended it to do.  We even occasionally get the work done in something close to the amount of time that we said it would take.

But our well-built software still fails—for a variety of reasons.

  • It’s too hard to use
  • Users are unable to use it efficiently/effectively
  • It’s missing one or more critical features
  • Users don’t have a need for the software in the first place
  • It’s too expensive, given what it does

In order to be “successful”, software has to meet a critical need that the user has.  Good software solves a pressing problem.  Great software does so in a way that seems natural to the users.

So what do I mean by software “failure”?  Simply put, “failed” software is: software that doesn’t get used.

What are the consequences of failure?  For internal software projects, it means wasted time and energy that could have been spent on things that the organization does need.  For consulting houses, it means possibly not getting hired back by the client, or seeing your reputation diminished.  For ISVs developing software/services to sell, it means lost revenue or even bankruptcy.

For developers, failure means knowing that you’ve wasted your time, intellect and energy on something that no one is going to use.  That’s no fun.

Our goal then as developers is to build great software.  We want to see users working with our stuff, to see it making their lives better, and to see them excited about it.  That’s the true Holy Grail that many of us work towards.

The Remedy

So how do we develop great software?  We can’t all be Steve Jobs, or hire him into our organization.  So without brilliant insight, how do we figure out what the users truly need?

To understand what users need—truly, madly, deeply—we need to think beyond developing products and start thinking about developing an understanding of our users.

This focus on a Customer Development Process, rather than a Product Development Process, is the point of the book The Four Steps to the Epiphany, by Steven Gary Blank.

Blank’s main thesis is that we should work towards an understanding of our customers and their needs much earlier in the development lifecycle.  We need to fully understand what customers need and whether we can sell them our product before we go too far down the path of building that product.

Blank proposes a very detailed Customer Development Process and talks a little bit about when it should occur, relative to the typical phases of a Product Development Process.  This is a little tricky.  If we wait too long to learn about our customers, we risk building the wrong product.  But if we talk to them too early, before we’ve had time to think a little bit about our vision, we’re not really innovating, but just trying to build what they tell us to build.  That also can lead to failure.

The Product Development Process

Having worked as a software developer for so many years, I’ve seen lots of different software lifecycles.  In my first job, with Control Data Corp in Bloomington, MN, we were a Dept of Defense shop and rigidly followed DOD-STD-2167A—a very rigidly-defined classic waterfall process.  (Ask me about the 4-foot high Software Requirements Specification).

I’ve also done my share of agile development, working in groups that used various agile methods.

My main takeaway on software process is that it’s important to develop iteratively.  For me, that’s the most critically important piece of the agile movement.  Many of the other techniques, like pair programming and test driven development might be important, but don’t seem quite as critical as being able to build your product iteratively.  Short iterations allow agility.

For me, one book that really made a lot of sense in laying out a framework for an iterative lifecycle was The Rational Unified Process Made Easy, by Per Kroll.

For a lot of people, when they think about RUP (Rational Unified Process), they think: heavyweight and lots of modeling/diagramming.  But Kroll explains that RUP doesn’t necessarily mean heavy.  He refers to the number of artifacts that you’re required to produce as your level of ceremony.  And he describes RUP as:

An iterative approach with an adaptable level of ceremony

Again, iterative is the important part.  An iterative approach with very little ceremony would basically be an agile methodology.  But sometimes you work on projects that require a bit more ceremony—some project tracking, etc.  On these projects, you can still be iterative, but with more ceremony.  Here’s a picture:

rupmap

RUP talks about general phases of the lifecycle being: Inception, Elaboration, Construction and Transition.  But Kroll is quick to point out that this does not just map to the classic waterfall model.  (Requirements, Analysis/Design, Implementation, Testing).  Rather, you’d typically perform some of each of the classic waterfall activities during each iteration of RUP.  You’d likely be doing a lot more requirements-gathering in your Inception phases, but you’d also be doing some implementation.  And you’d spend a lot of time writing code during Construction phases, but you might also still be tweaking some requirements.

Here’s a nice picture of how RUP typically plays out.  Again, note that the idea of iterations is important:

rupphases

Personally, I like this view of the development lifecycle a lot.  This model is how I think about software development.  Work in an iterative fashion, but be aware of what main phase you’re in and adjust your activities accordingly.

Over the years, I’ve taken lots of notes to answer the question, “what activities should I be doing at each step”?  Kroll has a lot of good information in his book, but I’ve borrowed liberally from other sources.  What I ended up with was a fairly detailed “a la carte” list of activities that you might engage in across the lifecycle.

I say “a la carte” because every project will not engage in every step.  But I think it’s a nice master list to draw from.  And I think that I have the activities listed in a pretty reasonable order.

So here’s my list of high-level activities for the Product Development Lifecycle.  (See below for more detail for each activity):

  • Vision
  • Initial Requirements Gathering  (Inception)
  • GUI Conceptualizing
  • Identify Candidate Architecture
  • Define Non-Functional Tasks
  • Define Iterations  (Project Planning)
  • Write Development Plan
  • Execute Elaboration iterations
  • Write Business Plan
  • Write Test Plan
  • Write Deployment Plan
  • Execute Construction Iterations
  • Initiate Test Plan
  • Deploy Alpha Release
  • Write Customer Support Plan
  • Execute Deployment Iteration
  • Deploy Beta Release
  • Launch Product
  • Begin Planning Next Release

The Customer Development Process

I love the idea of the Customer Development Process, as Blank presents it in The Four Steps to the Epiphany.  He talks about discovering who your customers are, figuring out what they need, and then testing your hypotheses.  In other words—iterate not just on the product, but also on your model of who the customers are.  This includes not just an understanding of the customers and their needs, but also gets into the actual selling proposition and marketing of your product.

Here are the top-level activities in Blank’s customer development process:

  • Customer Discovery
  • Customer Validation
  • Customer Creation
  • Company Building

Lining Things Up

The trick is to figure out how the Product Development Process and the Customer Development Process relate to each other.  If you listed out the various phases of each process side by side, how would they line up?  At what point in your product development process should you start the customer development process?  Or should it be the other way around?

Here’s a first stab at lining the two processes up.  My goal was to come up with a high-level roadmap for doing new (green field) development, which would cover developing both the product and the customers.

Product Development Process
Customer Development Process
Vision
Initial Requirements Gathering
GUI Conceptualizing
Identify Candidate Architecture
Define Non-Functional Tasks
Define Iterations (Project Planning)
Write Development Plan
Execute Elaboration Iterations  >> Customer Discovery
Write Business Plan
Write Test Plan
Write Deployment Plan
Execute Construction Iterations  >> Customer Validation
Initiate Test Plan
Deploy Alpha Release  >> Customer Creation
Write Customer Support Plan
Execute Deployment Iteration
Deploy Beta Release
Launch Product
Begin Planning Next Release  >> Company Building
.

Should I Keep Reading?

The rest of this post expands on each of the activities in the table above, listing details of what happens during each phase.  This is the outline that I use when I’m trying to figure out “what to do next”.

Note again—this process is very much geared towards green-field development, and in a market where you are developing a product or service to sell to end-users.

Product Development Process—Details

Here’s the detailed breakdown of the steps involved in the Product Development Process.  Much of this content comes from the books I list at the end of the article, primarily The Rational Unified Process Made Easy and Head First Software Development.  But a lot of it is just my own concept of what you typically do during each phase.

Note that iterations occur in the Elaboration and Construction phases.

Vision

  • Do a short description of a handful of possible products
    • For each, describe:
      • Market or market segment  (which group of users, short description)
        • What is the problem that they experience
        • How painful to the users?  What workarounds are in place?
      • Short description of product that might help them solve this problem
      • Short feature list (<10 items, single sentence)
      • How does this product specifically solve the customer’s problem?
  • Expand on top one or two possible products.
    (For each, expand information to include):

    • Write up SIMs (Specific Internet Market Segment, Walshpg 158..)
      • Lots of details about the targeted user group
    • Other user information
      • How often do you expect each user group to user the product?
      • Why might they stop using the product?
      • Rough guess as to size of this market, e.g. # potential users?
    • Are there other non-primary users?
      (E.g. users who didn’t purchase, or just different group of users)

      • Also describe their “problem” and how product solves it
      • Are they also potential buyers, or just users?
      • What subset of the feature list might they use?
      • What else differentiates them from primary user group?
      • Also write up SIM for this user group
    • Bluesky one or more revenue models.  Answer for each:
      • What would this customer pay for this product?
        • Specify target price or range that might be tolerable
      • One-time purchase or ongoing subscription?  (or a combination)
      • Pricing tiers?  If so, describe
      • Trial period?  If so, describe
      • If >1 user group, repeat for other groups
      • What level of certainty do you have that they will purchase?
      • Why might they not purchase?  List reasons.
      • Possible ways to mitigate non-purchase reasons.
      • Does revenue model depend on continued use of the product?
      • How will you collect money?
      • What is your distribution channel?  (i.e. how do they receive product)
      • If ongoing subscription, describe what happens when they stop paying
        (E.g. limited access)?
    • Revenue model/estimates  (per yr)
      • Make rough estimate of ongoing costs, per user or account
      • List expected revenue per user, single or ongoing, per yr
      • Starting with desired annual revenue, describe # customers required to meet goal
    • Marketing / Sales vision
      • How might product be branded?
      • How is product positioned (e.g. tagline)
      • How might you reach desired users?  (make aware)
      • What is selling proposition (argument to buy)?
    • Competitive analysis
      • Assess and summarize competitive products that do similar/same thing as yours
      • What are their strengths/weaknesses?
      • Where can you improve on them?  To what degree?
      • What are you missing that competitors might have?

Initial Requirements Gathering
(Chap 2 of Head First Software Development)

  • Generate quick list of basic ideas, 1-3 sentences each, Title/Description
  • Ask user questions to flesh out the list
  • Bluesky to generate feature lists (Title/Description)
    • Use other people
    • Okay to spit out non-functional stuff, like specific GUI thoughts, or architecture
  • Build User Stories
    • Describe one thing
    • Language of the customer
    • Title/Description
    • 1-3 sentences
  • Pull out non-customer stories, save for later (design)
  • Refine initial set of user stories, w/customer feedback
  • Provide estimates for each user story
    • Include assumptions
  • Team estimation and revision
  • Clarify assumptions w/users, if necessary
  • Break apart users stories that are >15 days
  • Add up estimates to get estimate for total project

GUI Conceptualizing

  • Select key use cases that need GUI concept
  • User interaction goals
    • Outline some key goals of user interaction model
      (E.g. discoverability, efficiency, slick animation, etc)
    • Describe user action requirements verbally for specific use cases
  • Assess similar products’ GUIs for similar use cases
    • Pros/cons of each
  • For each use case selected
    • Paper prototype one or more possible GUIs
    • Storyboard any dynamic behavior
    • Brainstorm alternative approaches and paper prototype
  • Identify common sequences of use cases  (e.g. find/edit)
    • Storyboard use case sequences
    • Brainstorm ways to optimize the storyboard/sequence
  • Assess GUI model against several measures of good GUI design
  • Technical review of GUI feasibility
    • Build it, buy it, or freebie (part of tool or freeware)?
    • If build, rough guess as to effort
    • If buy, what is the cost?  One-time & ongoing.
  • Assess technical risk
    • Are there any aspects of GUI that need to be proven out?
  • Hammer looking for nails
    • Make short list of most compelling GUI models in new products
    • Is there anywhere in product that you might use these models?
  • Brainstorm user assistance model (help, wizards, demos, etc.–learning)
  • GUI models from competitors
    • If there are competitive products that compete in same space, summarize their main GUI elements.
    • Assess—is their GUI or user interaction model an asset or a liability?
  • Review
    • Early review internally (not w/customers)
    • Review static screen mockups (e.g. paper prototypes)
    • Look at discoverability–obvious what it’s used for?
    • Aesthetics

Identify Candidate Architecture

  • Identify one possible architecture that could support the product
  • Identify any areas of technical risk
  • Discuss pros/cons/risks
  • [For more details, see Kroll]

Define Non-Functional Tasks

  • Define additional tasks to be completed during inception that are NOT use cases
  • Possible examples include:
    • GUI proof-of-concept, for technical & usability
    • Building candidate architecture  (must do)
    • Technical proof-of-concept
    • Capacity testing for architecture (e.g. scalability)
  • Provide priorities and estimates for all tasks
    (will feed into iteration planning, along w/use cases)

Define Iterations
(Chap 3 of Head First Software Development)

  • Set target date for first (next) release
  • Work w/customer to prioritize user stories
  • Also feed non-functional tasks into process
  • Select subset of user stories to meet target date  (Milestone 1)
    • If the features don’t fit target date, re-prioritize
    • Focus on user stories that are absolutely critical
    • Baseline functionality–smallest set of features so that SW is at all useful to customers
  • Prioritize user stories in Milestone 1.0  (1-5, with 1 as highest priority)
  • Assign user stories to iterations, using 20-day iterations
    • Based on priority
    • Or user stories required to implement other user stories
    • Use velocity of 0.7  (0.7days effort for each real day)
      I.e. Each iteration should contain 14 person-days of estimated work
  • Reassess schedule and adjust schedule and/or content
    • Because of velocity, things likely can’t fit
    • Add iteration(s), change M1 date, or both
  • Get iterations & user stories into tracking spreadsheet
  • Construct burn-down graph for 1st iteration
    • X axis is calendar days
    • Y axis is person-days of work accomplished
    • Draw diagonal line as ideal burn-down rate

Write Development Plan

  • Describe everything that will happen from here on out
  • Include
    • Iterations and contained user stories & tasks
    • Milestones
    • Review points and staff
    • High-level test plan
    • alpha/beta/release plan
    • Staff
    • Schedule
    • Tools
    • Show steps in Product Development process
    • Show steps in Customer Development process
    • Show how the two processes line up
    • Include calendar alignment, if appropriate

Execute Elaboration Iterations
(Chap 4 of Head First Software Development – User stories & tasks)

  • Break down each user story into tasks
    • Title/Description/Estimate
    • Provide estimate for each task
    • E.g. Create class, create GUI prototype, create schema, create SQL scripts
    • Each task should be 0.5 – 5 days
  • Plot where you are on burn-down graph
    • Calendar day vs. new estimate of person-days of work left
  • Update spreadsheet w/tasks & their estimates
    • Task estimates replace user story estimates
    • Also track burn-down in spreadsheet
    • Put stickies on big board, tasks, In Progress/Complete, etc. (pg 116 of Pilone/Miles)
  • Start working on first tasks
    • Move stickies on board when they are In Progress
  • Daily stand-up meetings to track progress
    • First thing in the morning
    • Track progress–what has each person accomplished
    • Update burn-down rate
    • Update tasks
    • What happened yesterday, what is plan for today
    • Talk about any problems
    • 5-15 mins long
  • Add unplanned tasks to iteration, if necessary
    • Add user story, estimate, break into tasks, review w/customer, add to board
    • Use red stickies
    • Update burn-down, showing that you’re off
  • Next iteration
    (Chap 10 of Head First Software Development)

    • End of iteration review (pg 342-343 of Pilone/Miles)
    • Verify SW passes all tests
    • Demo/review w/customer
    • Plan next iteration
      • Add new user stories, if required
      • Update priority/estimates for everything
      • Adjust velocity
      • Feed in bug backlog as tasks
      • Priority tradeoffs include bug-fixes vs. new features
    • Follow steps from Pilone/Miles chap 4 (break down into tasks and estimate)

Write Business Plan

  • Goal isn’t to get funding, but to make a case for the product/business
    (at least on paper)
  • Include stuff like
    • Market description
    • Customer description
    • Description of competitors’ products/services
    • Product proposition
      • What is customers’ problem?
      • What is your vision for product/service
      • How does product/service solve their problem?
    • Outline of development plan
    • Outline of Product/Customer Development processes
    • Summarize revenue model
    • Summarize marketing/sales plan
    • Business structure: staffing, organization, etc.
  • Add other typical business plan elements

Write Test Plan

  • What to test, how to test, when to test
  • Plan for test-driven development, using appropriate TDD tools
  • Plan automated/nightly builds and automation of testing
  • See Pilone/Miles, chap 7, among others

Write Deployment Plan

  • How/when will software be deployed?
  • Map out alpha/beta/launch
  • Where does testing fit into the plan?
  • How will customers get the software?
  • How will they pay for it?
  • How do you track customers?

Execute Construction Iterations
(Chap 4 of Head First Software Development – User stories & tasks)

  • Break down each user story into tasks
    • Title/Description/Estimate
    • Provide estimate for each task
    • E.g. Create class, create GUI prototype, create schema, create SQL scripts
    • Each task should be 0.5 – 5 days
  • Plot where you are on burn-down graph
    • Calendar day vs. new estimate of person-days of work left
  • Update spreadsheet w/tasks & their estimates
    • Task estimates replace user story estimates
    • Also track burn-down in spreadsheet
    • Put stickies on big board, tasks, In Progress/Complete, etc. (pg 116 of Pilone/Miles)
  • Start working on first tasks
    • Move stickies on board when they are In Progress
  • Daily stand-up meetings to track progress
    • First thing in the morning
    • Track progress–what has each person accomplished
    • Update burn-down rate
    • Update tasks
    • What happened yesterday, what is plan for today
    • Talk about any problems
    • 5-15 mins long
  • Add unplanned tasks to iteration, if necessary
    • Add user story, estimate, break into tasks, review w/customer, add to board
    • Use red stickies
    • Update burn-down, showing that you’re off
  • Next iteration
    (Chap 10 of Head First Software Development)

    • End of iteration review (pg 342-343 of Pilone/Miles)
    • Verify SW passes all tests
    • Demo/review w/customer
    • Plan next iteration
      • Add new user stories, if required
      • Update priority/estimates for everything
      • Adjust velocity
      • Feed in bug backlog as tasks
      • Priority tradeoffs include bug-fixes vs. new features
    • Follow steps from Pilone/Miles, chap 4 (break down into tasks and estimate)

Initiate Test Plan

  • Begin regular testing
  • Functional, integration, system test (no customer testing yet)

Deploy Alpha Release

  • Outline goals for beta and exit criteria
  • Identify potential customers/users
  • Communicate w/users about goals & feedback mechanism
  • Feeds into appropriate spot in Customer Development process
    • Possibly at end of each construction iteration
  • Gather feedback and feed into development
    • E.g. Changes desired/required?

Write Customer Support Plan

  • How will customers be supported after software is deployed?
  • List main goal(s)
  • Decide on tools, e.g.
    • External issue tracking
    • FAQ sheets for tech support staff
  • Decide on process(es), e.g.
    • Communication w/customer
    • Looking for answer in FAQ
    • How to communicate answer, if known
    • Get closure
    • 2nd tier–investigate, look for workaround
    • Interaction between staff (e.g. support/development)
    • 3rd tier–reported bug, e.g. give bug # and have traceback mechanism
  • Escalation process

Execute Deployment Iteration

  • Working on deployment tasks, e.g.
    • Installs/uninstalls
    • Distribution/delivery
    • Payment
    • Customer tracking

Deploy Beta Release

  • Outline goals for beta and exit criteria
  • Identify potential customers/users
  • Communicate w/users about goals & feedback mechanism
  • Feeds into appropriate spot in Customer Development process
    • Likely after Deployment Iteration
  • Gather feedback and feed into development
    • E.g. Changes desired/required?
  • Beta should not last indefinitely

Launch Product

  • Announce product
  • Launch it
  • Throw a big party

Begin Planning Next Release

  • Post-mortem for Product/Customer Development processes
  • Lessons learned
  • Changes to process
  • Assess organize outstanding bugs
    • Prioritize, estimate
  • Organize list of possible future features
    • Plan mechanism for using existing customers to get feedback on priorities
    • Prioritize, estimate
  • Decide on schedule for next release
  • Map out iterations
    • Each iteration either bug-fixing or new development
    • Bug-fixing phase likely first
  • Goal: All bugs fixed prior to next release
  • Continued refinement of plan, based on ongoing customer feedback
  • Back to start of Product/Customer Development processes for next release
    (follow same processes, including updating relevant plans)

Customer Development Process—Details

Here’s the detailed breakdown of the steps involved in the Customer Development Process.  This comes directly from Four Steps to the Epiphany.

Customer Discovery

  • State hypotheses
    • Write briefs, state assumptions about product, customers, pricing, demand, competitors
  • Test problem hypotheses
    • Test in front of potential customers
  • Test product concept
    • Test product features in front of customers
    • Solves their problem?
    • Also test business model
      • Must-have?
      • Pricing
      • Distribution
  • Verify
    • You understand the customer’s problems
    • Your product solves these problems
    • Customers will pay for the product
    • You have in mind a profitable business model
  • Repeat if necessary

Customer Validation

  • Get Ready to Sell
    • Articulate value proposition
    • Prep sales materials & collateral plan
    • Develop distribution channel plan
    • Develop sales roadmap
    • Hire sales closer
    • Synch up Product/Customer Dev teams on features/dates
    • Formalize advisory board
  • Sell to Visionary Customers
    • Sell unfinished product
    • Answer all sales roadmap questions
  • Develop Positioning
    • Initial positioning
    • Articulate belief about product and its place in the market
  • Verify
    • Enough orders to prove we can sell?
    • Profitable sales model?
    • Profitable business model?
    • Can you scale the business?
  • Repeat, if necessary

Customer Creation

  • Get Ready to Launch
    • Market Type Questionnaire
    • Choose Market Type
    • Choose 1st Year Objectives
  • Position company & product
    • Select PR agency
    • Positioning audits
    • Positioning to market type
  • Launch company & product
    • Select launch type
    • Select customer audience
    • Select the messengers
    • Craft the messages
    • Message context
  • Create demand
    • Demand creation strategy
    • Demand creation measurements
    • Iterate or exit

Company Building

  • Reach mainstream customers
    • Change “earlyvangelists” into mainstream customers
    • Manage sales growth by market type
  • Review management / Create mission culture
    • Review management
    • Develop “mission-centric” culture
  • Transition to functional departments
    • Set department mission statement
    • Set department roles by market type
  • Build fast-response departments
    • Implement mission-centric management
    • Create an “information culture”
    • Build a “leadership culture”

Closing Thoughts

The key takeaway for me from Blank’s book was as follows:

It’s not enough to build a great product.  You must also build a product that the customers need and that they will buy.  It’s also critical to find a workable business model to sell to the right customers at the right price.

If you haven’t yet read Four Steps to the Epiphany, I’d really encourage you to go out and read it.  Then you can start thinking about how to develop your customers, as well as your product.

Sources

I’ve used the following sources, in varying degrees, for this post:

Advertisements

One thought on “A Recipe for Green-Field Software Development

  1. Pingback: Software (and Product) Development | Fred's Blog

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s