The Purpose of Experiments in Startups

Robbert van Geldrop
April 5, 2016

The connotation of the word ‘experiment’ is about some scientists in lab coats holding up and looking at test tubes, while they carefully measure outcomes and make sure their findings are statistically sound.


It’s what the Doc is doing in his garage all day:


However, neither is a good framing for what type of experiments you should do when you’re starting up.

Experiments in your startups serves the purpose of finding out two things:

  • Validating your business model
  • Finding out if your team is able to execute

Sourcing your experiments

The first step for running your startup experiments is to source them. *Lean Startup *expert Tristan Kromer has a great framework for this:


The diagram shows two axises. First of all there’s the distinction between generative research and evaluative experiments. Quoting Tristan Kromer:

  • Generative* — Research techniques which don’t necessarily start with a hypothesis, but result in many new ideas. e.g. Customer Discovery Interviews.*
  • Evaluative* — Testing a specific hypothesis to get a clear yes or no result. e.g. Landing Page a.k.a Smoke tests*

Secondly, there’s market vs product. Since you are required to find product/market fit for your business, you are probably going to have to research and experiment at least either side of this to succeed.

Startup teams who conduct on Experiment Design often jump right into crafting some evaluative experiment. However, without a good hypothesis to test, it will not lead to meaningful insights. These hypotheses typically emerge from preceding generative research, which give you signalson what to test.

Inhibitors and how to overcome them

Some teams start without a technical co-founder and hope to add or hire one later. This is a fundamental element in your strategy and can strongly influence your course of action.

One of the biggest misconceptions by non-technical founders is that they expect to find a technical co-founder based on the idea alone. If you have an incomplete team you need to collect compelling evidence that you have a sound business model and some proof that it’s worth joining your startup. Also, you need to demonstrate your ability to execute on all the other things engineers typically do not prefer to do.

If your team lacks the ability to create a Minimum Viable Product all by itself, you need to include this in your strategy and change the way your going to test your business model. The good news is that there are many tools, platforms and services these days which allow you to work around that inhibitor. Many startups have preceded you in doing this.

The textbook example is Groupon:

On their blog, they published an article which offered a group discount to > the pizza joint around the corner. The deal was such a success that they > pivoted to the model of discounts and daily deals. Read the full story here.

Build your Minimum Viable Product without code

Many startups are now paving the path on how to run your startups, experimenting iteratively and avoiding big chunks of development, which slow you down.

Here are some blog posts on how and why to build your Minimum Viable Product without coding: