Lean Expectations — PoC, Prototype, MVP
For new software to have intrinsic value, it typically introduces innovative approaches to solving a problem. This can come in many forms but is usually improvements on functionality, usability or performance. Sometimes it is addressing new use cases by greeting new market segments that are not currently being addressed by existing software. Whether introducing new approaches or exposure to new markets, there are assumptions made that need to be tested.
If we fail to put our assumptions to the test, we risk investing heavily in building a flawed solution or worse, a solution to the wrong problem. Whether or not these new approaches have fully tested assumptions prior to starting development, much learning should be done along the way.
“Life’s too short to build something that nobody wants. “ — Ash Maurya, Spark59
We live in a rapidly changing world and software can be out of date by the time it is released. Users are smart and rarely wait for us to solve their problems for them. They adopt other tools or devise clever work arounds to fill the void we are attempting to address. The original assumptions that the software was based upon will continue to evolve while our software is being developed.
It is important to be purposeful in the iterative learning process since it can take time to deliver software and the use cases we set out to solve are ever changing. Adopting the tools of Lean Software Development is ideal as it focuses on the following principles: eliminate waste, amplify learning, decide as late as possible, deliver as fast as possible, empower the team, and build integrity in. Transitioning from the traditional Waterfall to the more modern Agile methodology is a great place to start as it fosters many of the lean principles.
Now that we are a part of a Scrum team using Agile methods we can iteratively improve our development practices by introducing more principles of Lean Software Development. We set the stage for being able to keep up with the rate of change of our users and are ready to develop new software. It doesn’t take long to make it appear functional. We worked out the kinks in the workflow and it demoes so well. We built it directly along the happy path. No matter how much time we invest, it can appear perfect until users start to, well, use it. That’s when the frustration begins. We fix the bugs, adapt the workflow, and once again present it as a working product. The cycle repeats continuously and discouragement can begin to set in. Some of this is inevitable. We cannot produce error free code. If we did we probably took too long to do it.
It only seems logical that more programming flaws and design holes exist earlier in the product’s development. This leads to the initial impression of the application to be marred very quickly. So how do we set expectations early in the development to prevent frustration and discouragement?
There are quite a few misconceptions along the way that can be cleared up by setting proper expectations, especially early in the software’s lifecycle. While we may be able to convince our stakeholders of the value of adopting lean principles, the expectation remains that the software produced will be immediately usable and address the use cases the first time.
Software development priorities can evolve as the project matures. One important concept to explore is the inverse relationship between the scope of learning and stability. This sounds like a far stretch but has a simple and logical explanation. True freedom to learn and apply what is learned requires freedom to introduce change. Significant changes undermine stability. Stability is gained by limiting the scope of change in the learning process. We want to always encourage a fast rate of learning but reduce the scope of changes incurred as the product matures.
To strike a balance of this natural tension, we follow the second principle of Lean Software Development, amplify learning. Shorten your learning cycles by frequently involving user input and user testing. This can introduce change in a more predictable manner and over time offer a sense of increased stability to the users. When there is much to be learned, prepare to introduce drastic changes. As the learning from the users results in smaller changes, a more stable product naturally emerges.
One of the ways to help set proper expectations while learning to meet user’s needs in a timely manner can be implemented in an iterative cycle of Proof of Concept (PoC), Prototype, and Minimum Viable Product (MVP). While we may already be familiar with the terms it is helpful to understand and communicate the value of each so that we keep ourselves and others from being frustrated along the way. This provides a framework of terminology that can help shape the conversations about our progress and how much change should be expected.
Proof of Concept
A Proof of Concept (PoC) is basically a stand-alone project that covers a very small scope and is quickly developed or hacked together so that it can test assumptions. Using the scientific method, you devise experiments to test your hypothesis. This can be accomplished in a multitude of ways such as stand alone codebases, web pages, user workflow diagrams, or even spreadsheets with macros. I’ve seen great functional ones built on FileMaker Pro that also served as a Prototype. There are even applications made for this phase. Regardless of which tool you use, you should always know up front what you want to learn and how to measure success or failure. This ultimately guides whether or not to continue investing in this approach or try something new.
There are two goals during the testing of the PoC, to test assumptions and to share the vision. Either you fail quickly and learn from it or you learn that you proved the concept and are now ready to share the vision. Both outcomes are a step forward as stated in the book The Lean Startup, “you will always succeed — in seeing what happens.” Adapting your approach based on your failed experiments during this early phase can prevent years invested in a project that was based on incorrect assumptions. If you do end up with a successful PoC, it is evident when you are no longer spending much of your time convincing others. Let the PoC speak for itself and others will confirm it. This makes raising the funding much easier for both the PoC and subsequent development.
It is important to not expect a PoC to be used long term by users for more than testing due to little effort placed in software architecture and user management. A major question that can be answered is if a new standalone product is needed or if it should it be included as a new feature inside another application to extend the definition of the respective MVP. This is more difficult to answer if it was built in the codebase of another project and quite complex to extract if that is what is necessary. PoCs can also sit on the shelf for a couple months or even a couple years before the timing is right to invest further.
PoCs are ideal for new projects and major features. New features can include user interface design and workflows. You can test this by creating mockups of the interface and get user feedback. This is commonly skipped and much development time is wasted in integrating untested user interfaces. While professional designers have specialized tools for this, it can be done in presentation software like Apple’s Keynote. Every hour of a designer’s time can save a day of a developer’s time, and days spent here can save weeks. Diversifying the team to include this role or even contracting this out can help scale the effectiveness of the team.
There are times where the concept does not have to be proven. This can be the case with small incremental feature improvements or small features. Creating a PoC outside of an existing project for these can be counter productive. Even when this is the case be careful not to invest too much time without user testing these new features or improvements.
The key in developing a PoC is to use the tools you know and focus on the easiest way to get it done quickly. This is common in hackathons. Resist the urge to optimize performance or test too many assumptions concurrently or your PoC will take way too long. You can always circle back and combine other successful PoCs into a new one. Try to keep the scope small enough of each to complete in a couple of weeks using 1–2 people. This doesn’t have to be done by a developer. Many times non developers are quite effective at creating PoCs. The prerequisites are motivation and passion to solve the problem. This can also allow developers to stay focused on their other projects.
The PoC typically has the smallest scope and has limited usability but is crucial in testing assumptions with minimal investment. Be prepared to start over a few times and introduce a lot of change because there is the most to be learned here. This phase is immensely helpful in sharing the vision while obtaining funding.
A Prototype can be thought of as the step between a PoC and MVP. Once the concept has been proven in a limited scope, you can take a little more time proving whether or not the concept works in a more integrated implementation. Unless the prototype was specifically user experience related, this usually focuses on functionality rather than design or even usability.
The Prototype is typically not a refactor of the PoC but a rewrite because of the shortcuts taken previously and applying what was learned. This should be more thoroughly user tested to further understand the bottlenecks of things such as software architecture and user management. This typically has a lot of bugs that have yet to be addressed and labeled as Alpha or Beta. While using Semantic Versioning, it can be signified as a release less than 1.0.
There are use cases such as internal software, that is not customer facing, that can get away with using prototypes in production. In these cases, the software is built to complement the services offered by the company or organization rather than built for the customer’s direct consumption. Updates come as inspiration hits and frequently after firefighting sessions. Releases aren’t usually planned and appear erratic. A common sign of using prototypes in production can be internal staff being frustrated by bugs and downtime that prevent them from delivering their services.
Prototypes are ideal for taking the next step of a PoC. While the concept has already been proven there may be uncertainty as to how it is integrated and architected. This provides a safer environment for peer feedback which lead to clearer separation of concerns, refactoring large portions of the code and architecture improvements to address performance bottlenecks. Avoid pre-optimization as most of the time it would be nullified by design changes. The Prototype minimizes the impact of the above as there should be fewer dependencies compared to directly developing the MVP by skipping the Prototype phase.
There are times when the separation of concerns are already clearly known and the architecture is predetermined. This can be the case when implementing an existing feature from another software project. In this case, a Prototype can feel like a waste of time. However, it is easier to promote a Prototype than it is to iteratively refactor and improve the code if this phase is skipped. It is also easier to exceed expectations this way. Introducing a refactor when this phase is skipped is much harder to sell to the development team and stakeholders.
If the PoC took a couple of weeks, the Prototype would typically take a couple of months of 1–3 developers. A complex Prototype can take much longer. The time is more proportional to complexity than it is to the number of developers. In fact many times this is fastest by the “lone wolf” developer but lacks the advantages of peer feedback.
Adding more developers rarely speeds up the development process as it complicates team dynamics and the complexity of code. The Ringelmann Effect applies here. It states that a smaller team can be more productive than a larger one. As the code is properly modularized it makes it easier for more team members and even multiple teams to contribute without slowing the process down as much because they are working on separate but integrated projects. This is discussed in a bit more depth in the MVP section.
Much insight is gained during testing of the Prototype as to what the separation of concerns are and what architecture should be used in the MVP. Showcasing the progress made and learning of the Prototype helps set expectations for the MVP. Be prepared for a lot of changes here which help minimize wasted development time if refactoring code integrated in the MVP.
Minimum Viable Product
The Minimum Viable Product (MVP) is typically the minimal functionality required to meet the root needs of the user. Many times this sacrifices the non-essential but desired features. While the name infers being stripped down to the bare essentials, it includes much more than the previous phases. The development team should not make the decisions of the included functions and features in isolation. Stakeholders were important before but are invaluable in this phase. Much more planning is required developing the MVP.
Clearly designed software architecture and user interfaces are crucial in this phase. Don’t stop learning at the Prototype phase. Continue to let users test your designs and interfaces along the way.
Once the MVP is released with the minimal feature set, new major features can then be added using the natural progression of the PoC, Prototype, then MVP. Starting with the PoC to test that the feature will meet the needs. Ideally this is done externally at least as a module to ensure a clean separation of concerns. If it is too integrated, you run the risk of blurring lines and undermining the approach. Next, rewrite the feature in a more integrated way, as the Prototype, but not necessarily in its final form to be released. The Prototype integration can help decide whether it’s most appropriate home is included in the original product, needs to spin off on its own, or requires a whole new platform to be developed. Finally, integrating the feature into the original product is expanding the definition of the product’s minimum viable-ness.
This MVP phase can have an indefinite timeframe due to its iterative nature but expect as much as a year for a stable release using a small team of about 2–4 developers. This is most dependent on the scope of the software more than the number of developers. According to Brook’s Law introduced in The Mythical Man-Month, throwing more developers at the problem does not always speed up the process and has increasingly diminishing returns. With Modular Programming, where the codebase is organized by functionality as modules, more team members can be added. Multiple teams can work in tandem on separate modules which can improve timelines but requires ample communication and planning, in turn increasing the inherent complexity of team dynamics. Clear examples of this are with front-end user interface and back-end business and data logic. User interface design as well as software architecture design can also be further abstracted to speed up implementation.
While much of the same iterative learning can be made by starting with the MVP, opportunities can be missed by skipping the PoC and Prototype phases. PoCs encourage teams to be more receptive to raw feedback. Sensitivity to criticism tends to be directly proportionate to the amount of effort invested. Prototypes play a similar role but the critics tend to be peer developers. There is something liberating about sharing a prototype with a team for input in place of defending your approach. Directly building features that miss the mark in the MVP come at a higher price including emotions. If the goal is reducing the investment of the learning cycle, PoCs and Prototypes have value.
The MVP should naturally have a higher priority of stability over learning in comparison to the PoC and Prototype phases. Try not to let that be an excuse for failing to learn.
Life after MVP
There are quite a few online articles that discuss other approaches that include phases after MVP. After creating dozens of PoC’s and Prototypes through the years, half a dozen have made it to the MVP phase and brought to market in one form or another including patent applications on a few and one awarded. The MVP phase is usually where I hand the project to a full Scrum team and either fulfill the role of Product Owner or slide over to being a stakeholder when we find the right person to fulfill the Product Owner role.
“The reason to build a new team to pursue an idea is that you believe you can accelerate through the Build-Measure-Learn feedback loop faster than anyone else can.”
— The Lean Startup
Depending on your definition of what is considered minimally viable, you can keep adding features to meet the purpose of the application. Most software has a limited shelf life of only a couple years before the competition over takes it or a new version includes new features that can be integrated through this lifecycle. To stay relevant, it seems strategic to reassess the entire product by then. The lifecycle of new features can help the team know when it is time for that rebirth. If each iteration is developed in modules, code can be reused so that not all of the pieces have to be rewritten.
Example Application Lifecycle
One of the applications that has used the PoC, Prototype, MVP development lifecycle in multiple phases is translationCore (tC). While there were other applications in the domain to fulfill similar roles to check Bible translations, this one had new innovative approaches that addressed new use cases and even new customer segments. We decided that this phased approach would help us test our assumptions and help manage expectations around it.
First, a simple PoC was developed in a matter of weeks. It was a stand-alone web page, quickly hacked together by a single developer. It was used to test the assumptions and explore limits. The PoC and its findings were used to test assumptions, cast a vision, raise funding and build a future user base. Once we achieved enough buy-in and learned from user testing, we moved on to the Prototype.
At first it seemed logical to integrate it into the existing application we had that was already on the market, translationStudio. This application used many of the same resources and had some common functionality. However, we knew that the existing application was designed with a complete different set of assumptions. It had already been heavily modified through the years and would need a major refactor or likely a rewrite to handle this new functionality. We further identified that there might even be a different user base and risked making the original product less effective for the existing users.
We decided to build a new platform. Since the prototype needed to be more stable and robust than the PoC, nearly half a year was spent on it. With our testing and learning from the Prototype, we started rewriting the app piece by piece working towards our MVP. Due to the modularity of the Prototype, we were able to do so in iterative phases allowing some portions of legacy prototype code to stay around until it no longer served its purpose and was important enough for us to clean up. We learned a lot along the way, balancing what we defined as a minimum feature set and implementing an iteratively cleaner separation of concerns to minimize technical debt in the long term.
Lifecycle of Modules
While the main tC application was being built, we created modules that we called tools that run inside the application with their own development lifecycle of PoC, Prototype, MVP. Their code is managed in separate repositories allowing for a manageable way to replace each tool’s Prototype with its MVP.
Features of one of the tools, WordMAP, also has its own lifecycle as well. Two years ago, it was a PoC that helped cast a vision for automated alignment with minimal requirements. It worked offline and in a browser. We used it to test assumptions and cast a vision. It sat idle for nearly two years before it was rewritten in a more planned manner as a Prototype. While it is a stand-alone module, it is able to be integrated into tC and it’s tools. The external nature of the module allows it to have its own lifecycle and not have to be directly tied to the fate of any other project and be used in other modules and applications. It is undergoing testing and we plan to have features of WordMAP with their own PoC and Prototypes.
Hopefully in reading this far you understand the value of pacing your software development into phases such as Proof of Concept, Prototype and Minimum Viable Product and have an idea on how to set the proper expectations of each phase. PoCs are great for building excitement by trying something new without being held back by existing limits. They are hacked together by design and typically not meant for users but to test assumptions. Prototypes are great for improving on the PoC by learning from the mistakes made and seeing how the software fits into the greater workflow it will serve. Minimum Viable Products are great for building usable software without unnecessary features that will slow the time to market.
This lifecycle is scalable and when coupled with modular design decouples the pieces of a monolithic application into bite-size pieces. This fosters best practices and makes it easier to contract out work on portions of the application. This also encourages external collaboration on software from partner organizations and the open source community.
“Remember, planning is a tool that only works in the presence of a long and stable operating history. And yet, do any of us feel that the world around us is getting more and more stable every day? Changing such a mind-set is hard but critical to startup success.”
—The Lean Startup
There is no silver bullet in software development. However, iterative learning is still paramount even if enough field research has been done to study the use cases. Since the world is continuing to change at a growingly rapid pace, research may be out of sync by the time the software application is built. Think of this approach as a part of field research. Test assumptions made in the PoC and Prototype during the research phase, that way there is less of a disconnect between the research and the development efforts. Continue testing during the MVP phase to ensure the user needs are adequately met.
Have you had a difficult time setting expectations for your software? Share these ideas and concepts with your team to see if it is a good fit for your current or future projects. How might applying the concepts shared help you turn frustration into satisfaction? Let me know in the comments how you have applied this in the past or any stories of success and failures.