This is the fourth post of a series that will try to show how Lean Startup can be useful at various phases of an organisation’s life.  This post is the continuation of Lean Startup in the Discovery Phase – Part 1 and will show you how to conduct a Lean Startup experiment in the discovery phase of your project

Your first Lean Startup experiment

In Lean Startup all ideas that impact your business model are to be considered as assumptions to be validated. Hence, experimentation is the hearth of Lean Startup. A Lean Startup experiment can be summarised by the Build, Measure and Learn loop.

Build

Build Measure Learn - Baker MarektingIn order to build a Lean Startup experiment, you first need to transform your assumption into a hypothesis. A hypothesis has the form of a simple, objective statement that can be validated (or invalidated). This may seem simple but it is often one of the most difficult parts of the experiment.  It entails that you already have a significant understanding of your experiment participants’ lingo, environment and mindset.  Hopefully, you obtained some of this knowledge during your ideation phase when you informally discussed your project.

Your hypothesis must also produce results that will be actionable whether you validate or invalidate it.  If, for example, you state your hypothesis as: Customer segment A will always prefer my product to the other solutions presented, all you need is one potential customer to choose another solution to invalidate your hypothesis. What action will you take if you invalidate your hypothesis? None; as your business doesn’t require 100% of a customer segment to adopt your product to be viable.

Simply dropping the word ‘’always’’ in your hypothesis statement will allow you to take action if you validate your hypothesis (you’ll go forward with the segment) or invalidate your hypothesis (you will either pivot on your segment choice or adapt your product further to fill the needs of this segment).

It is also interesting to note that this hypothesis statement contains an unverified assumption. You assume that the ‘’other solutions presented’’ will be the most important substitutes to your product. The experiment, if well designed, will allow you to verify this as well. You may find out that your target segment uses other substitutes that you didn’t account for.

Once you have stated your hypothesis clearly, you need to design a Lean Startup experiment that will enable you to validate or invalidate it. As you now know, you will have a whole lot of these Lean Startup experiments to conduct during your start-up process. Therefore you will be thinking of designing not just any experiment, but the experiment that will require the least amount of resources to achieve your goal.  If your product requires expensive machinery to produce, you certainly don’t want to build the entire factory to validate your assumption. You’ll need to be creative and find ways to get your machine’s concept across to the people in your experiment in the least costly way (in $, time and other resources) possible.

The way you will have found to get your product concept message across will be your Minimum Viable Product (MVP). Hence, in this particular example, your MVP could take the form of a drawing on a napkin, on a computer (2 or 3D), in a video, a physical model with or without moving pieces or using a similar product that you supplement with explanations of what your product will do differently. Whatever you feel is necessary and sufficient (no more, no less) to get the message across.

You also need to keep in mind that your Lean Startup experiment should be replicable. This means that in a year from now, if someone else needed to audit your results, they could replicate your experiment and get similar results. In other words, don’t choose a period, a setting or participant profiles that will skew your results.

Measure

Build Measure Learn - Baker MarektingOnce you have identified your MVP, you will need to decide what metric (data) you will be collecting and well as the standard it will need to measure against. In the example cited above, you would be collecting how many people in your experiment chose your product over other solutions that were offered to them. The standard, by definition, would have to be above 50% of your sample given your hypothesis states a preference.  Any results below 50% of participants in the experiment choosing an alternative solution would not show a preference. If your hypothesis statement did not imply a comparison but simply an interest in using your product, the standard (% of participants in this case) could be set at whatever you feel would be sufficient to demonstrate clear interest.

In order to reduce the objectivity of standard setting, you should try to get the input from someone who is very familiar with the stakeholders in your experiment in respect to your product/service.

When measuring, you aren’t looking for statistically significant results. This would be ideal but too costly in time and other resources. Instead, you will simply be looking for a clear direction in the data. If your experiment data doesn’t show a clear direction, you will need to include more participants. Occasionally, your experiment will not show a clear direction. You gather whatever learnings you can from it and move on to designing another experiment to validate your assumption.

A word of caution here, despite wanting to go as fast as possible in your Lean Startup experiments you still want to conduct them in a way that will get you to validate or invalidate your hypothesis.  If your data collection methods are not done properly and your results are seriously skewed, you will have wasted your time and energy.  Hence, you still need to have some basic knowledge of primary data collection techniques. Here are a few good sources of basic information on data collection that you should take a look at if your knowledge on the topic is nil or very low.

Learn

Build Measure Learn - Baker MarektingDuring your first Lean Startup experiments, which should be conducted face to face with your participants (this would be the Get out of the building part), you will have the greatest amount of learnings.

While you conduct your experiment, you should always leave room for the unplanned. If a participant reacts in an unanticipated way, run with it and dig into it.

Try to avoid justifying your product/service. Embrace all the comments you will have, especially the negative ones. Dig as much as you can into those. They will give you the best insights. Let the participants talk and listen to the words used, observe body language and notice expressions they use. If at all possible video or tape-record your interactions.

Keep an open mind. It won’t help your start-up if the participants understand your point of view or for you to tell them how you understand their problem. It will however help you tremendously to understand their point of view and what they perceive their problems are or your solution to be.

Your first Lean Startup experiments will help you learn what your stakeholders think about your business model assumptions, as well as all the assumptions you didn’t know you didn’t know. On top of this, you are also learning how to conduct an experiment.

Expect to make mistakes, a lot of them, in the entire build, measure and learn process. That’s ok. Learn from them as quickly as possible and always keep your ultimate goal in mind. If you start an experiment and realise you didn’t state your hypothesis correctly, design it right or choose to collect the right metrics, make immediate changes. Don’t persist with 10-20 other participants with a flawed experiment. Your ultimate goal is to validate a hypothesis, not run an experiment. Similarly, if you had planned to involve 50 participants and you have a very clear direction after 15[1], stop.

The experiment log

Build Measure Learn - Baker MarektingIf at any point in time in your start-up process you will be going to venture capital firms for funds, you will want to record your experiments in a log. Your Lean Startup experiment log will document your progress as a start-up. It will show how you dealt with the unexpected, all your learnings and the progress made towards a product/market fit. The data collected in your experiment log will also serve as input in your innovation accounting. Innovation accounting, developed by David Binetti, is one of the more complex but exciting component of Lean Startup. You can learn more about innovation accounting in these previous posts on how innovation accounting is changing the rules, how it’s done and on the concept of innovation options.

Practice makes perfect

Practice Lean Startup - Baker MarketingAs you may have figured out by now, applying Lean Startup requires effort, lots of it, and discipline. Well, that is the reality of starting a business.  Lean Startup experiments have one purpose; to bring you closer to having a profitable business or perennial organisation.

You may not be very good at the whole build, measure, learn thing at first. That’s ok. Just keep at it and practice, practice, practice. It gets easier, much easier, I promise.  After a few Lean Startup experiments you will start getting the hang of it. Not only will you already have a better understanding of the mindset of your stakeholders, you will develop tools and processes that will make experimenting much faster. You will also develop skills that will enable you to identify critical assumptions quickly, as well as spot, absorb and act on useful information more efficiently.

After a few dozen Lean Startup experiments, the build, measure, learn process will become second nature to you and others, applying it, in your organisation.

Using Lean Startup is similar to buying an insurance policy. It does require more efforts initially than if you simply build your product and service and take it to market. The benefits are that you won’t be building a product or service that will never lead to a profitable business.

Our next post will present some of the tools to help you run your experiments faster, cheaper and more efficiently.

[1] As long as your sample is not completely biased. Your participants sample has to somewhat reflect the overall population of the stakeholders of your experiment.