How to really learn from your experiments
At any stage of a start-up’s life, there are dozens, if not hundreds, of different directions to take. In the early phases, you need to pick the right product direction and find your product-market fit. Later, you may wonder which marketing channels will work at scale and which ones won’t.
Since you never want to approach each crossroad blindly, the key is to continually test out the options. Success is often defined by how well you hone these experiments and learn as an organization. What separates great start-ups from the rest is how many experiments they can run at the same time, and how effective each is experiment is.
Building out your test
Approaching your experiment is not that much different than creating an A/B test – although you’re testing for a much bigger matter than tweaking the font and image choices on a landing page.
Any experiment needs to be well planned, grounded in a specific hypothesis and potential outcomes. Follow these steps to create your test:
1. Define the problem. What are you trying to fix or improve?
2. Define potential solutions. What are all the different possibilities that could solve the problem?
3. Define potential outcomes. What do you expect will result from your solution? What’s the best case scenario? The worst?
4. Run the experiment.
5. Analyze the results.
6. Double down on what worked, drop whatever didn’t. If the results were inconclusive, run another experiment.
For example: Let’s say you want to figure out if paid search marketing can help you scale your business profitably.
Wrong approach: You might be tempted to dive right in…start small by investing $50/day on Google to buy a few keywords and look at the results over the next few months.
This strategy won’t work for a few reasons. One, $50/day is too small an amount to derive any meaningful conclusion at scale. Second, it’s far too general: there’s no defined timeline and no defined outcome.
Right approach: Follow the steps above to fully define all aspects upfront: Define a meaningful budget (i.e. $500/day); define a specific timeline (i.e. one month); define the outcome (i.e. need to have a CAC of $50, otherwise this won’t be a profitable acquisition channel); and define different solutions (i.e. bid on general search terms vs. long-tail keywords).
Too many start-ups fall into the pitfall of running experiments for the sake of running them, and then never actually learn anything since the parameters were too fuzzy. Drive your activities toward reaching tangible learning milestones…as many as possible, as fast as possible.