How to build an effective experimentation strategy
Published March 23, 2023 by Jacob Pretorius – Tech Lead & Optimizely MVP
In this article we will show what the benefits are of an effective experimentation strategy and tips on how to create and continuously improve a strategy for the best results.
To get the most from any experimentation platform, organisations need a good experimentation strategy. A good strategy helps by reducing the risks of experimentation and can ease the decision-making process. It can also lead to an increase in efficiency across multiple teams as everyone is clear on what they are working towards.
For organisations that don’t have an existing strategy, it is easy to start with a simple approach.
Ten steps to create an experimentation strategy
If a business is new to experimentation, here are some steps they can take to create an experimentation strategy:
1. Define objectives
Start by defining the objectives and identifying what the business wants to achieve through experimentation. This will help identify the areas of the business that need improvement and prioritize which experiments to run.
2. Set up testing software
Establish a process for conducting experiments and find a suitable platform to manage the testing process, including the tools and resources needed to run experiments. More on this later.
3. Choose metrics
Identify the metrics that matter to your business, aligning with business goals and objectives, and develop a clear understanding of what the team are measuring, why it matters, and how to interpret the data. Make sure these metrics are available in the experimentation platform of choice.
4. Develop a hypothesis
Develop hypotheses that can help test the initial assumptions and guide the experiments. A hypothesis should outline the what, how, and why of the testing. It should also include what the expected outcome or learnings are from the experiment.
5. Plan and prioritise experiments
Develop a plan for running experiments and prioritize them based on the potential impact on the business.
6. Run the experiments
Launch and run the experiments, making sure to document the process, data, and findings clearly.
7. Analyse results
Analyse the data collected and use it to identify what worked and what didn't work and make changes accordingly.
8. Iterate and optimise
Based on the findings, optimize the experimentation strategy and iterate on the testing process. More on this later.
9. Communicate results
Share the results of the experiments with key stakeholders, including leadership, to ensure alignment and promote a culture of experimentation.
10. Establish ongoing experimentation
Make experimentation an ongoing process, continuously testing and improving your business. Remember to learn from both successful and unsuccessful experiments, as both offer valuable insights.
With the basic strategy is in place and the business has run a few experiments successfully, the next step is to evolve the strategy.
Eight tips to improve an existing experimentation strategy
Here are eight tips to iterate and improve on an experimentation strategy:
1. Ensure clear and measurable objectives
Take the opportunity to examine initial objectives and what the business wants to achieve through experimentation. Make sure the objectives all have defined outcomes that are measurable and identifiable against the control or baseline.
2. Select the right metrics
Choose metrics that align best with the objectives and provide valuable insights into the users or experiments. See if the experiments can support additional metrics for a deeper understanding of user behaviour.
3. Consider alternative hypothesis
The first instinctive hypothesis isn’t always the best at finding the true root desire or problem users encounter. Consider a few alternatives to the first hypothesis and see where it leads.
4. Control for external factors
Try to eliminate or account for external factors that could influence the results, such as seasonality, changes in the market, or user behaviour. Look for recurring patterns in existing internal and external data and combine that with previous experimentation results for a deeper understanding.
5. Test more than one variable at a time
It’s a good idea to start with singular experiments that change one variable at a time. However, if the chosen experimentation platform supports it, running multi-variate tests can achieve results that may otherwise not be possible.
6. Share results widely
Record the findings and analyse the data to draw actionable insights in a more general sense. Share these results and analysis with anyone who will listen, even if they don’t work in the “experimentation” department. Generally, all business roles can benefit from a deeper understanding of their users or customers and experimentation results can often be the perfect way to start these conversations.
7 Learn from failure
There is no such thing as a failed experiment, only incorrect starting assumptions. Don’t be scared to push the limits and reach of what is testable.
8. Build a culture of experimentation
Encourage experimentation as a continuous process and integrate it into your organization's culture regardless of business role or seniority. Often time the best experiments come from outside of the experimentation “team”.
Choose the right platform for your needs
There are many experimentation platforms available, and while most provide a similar core set of features, it is important to evaluate the options and make sure the software fits well with the business goals, needs, and its users or clients.
If you are interested in reading more about the features and benefits Optimizely has feel free to read our articles on How Optimizely Web Experimentation enables fast validation of design and UX changes and How to go beyond web A/B testing with Optimizely Feature Experimentation.
Support for experimentation strategies
Experimentation can be an intimidating hurdle for any organisation, but with the right strategy in place any business can achieve positive results.
We’ve shared ten steps for creating an initial experimentation strategy and a further eight tips on how to refine the new strategy over time for the best results.
If you want to learn more or are interested in support with web experimentation don’t hesitate to reach out to us.