De Facto Guide to “Startup Metrics for Pirates”
Every day companies are required to make mission-critical decisions. Trouble is, they’re often made without all the facts — particularly in the land of startups, where speed and execution are the only things keeping you alive.
Generally speaking, some of the best decisions are data-driven. In that vein, I’ve pulled together a guide for a popular framework I’ve found effective in making informed product-oriented decisions. The framework is Dave McClure’s “Startup Metrics for Pirates”. AARRR!
Let’s get started!
First off, what are Pirate Metrics?
Pirate Metrics is a framework that condenses the customer life cycle into a single, measurable funnel.
Pirate Metrics allows you to identify ‘leaky’ areas at different stages of your customer life cycle and make informed decisions to improve it. The 5 stages AARRR:
Acquisition: prospective users visit your site from different channels.
Activation: users perform the very first step or event to start using your product. The ‘aha’ moment or first taste of your core benefit.
Retention: a measurement of users coming back to your product/service.
Referral: users enjoy your product/service enough to refer others.
Revenue: users complete some sort of conversion event that translates into $$$.
Here is a visual representation of what the funnel looks like:
How do I… Pirate Metric?
To make a framework like pirate metrics work, you need to do 5 things:
Define, Measure, Analyze, Improve and Control.
Your Measurement Plan:
Before you can measure “success”, you have to understand your company’s objectives. A measurement plan is a great way to do this. Be sure to write it down and engage your team when drafting it. This is super important. People support what they help build, so engage them early in the process.
Steps to create a measurement plan:
- Define business objectives. What is important to the business? What are your main problems, challenges and opportunities?
- Ask questions to form hypotheses for each life cycle stage. How are users currently interacting with us? Our competitors? What channels are they using? Who are our ideal customers? How can we remind users of the value they receive from our product? What could we do to make an impact on our business objectives?
- Ideate experiments. What key events in our customer life cycle can we use to test our hypotheses from above? Select a few measurable events to test your hypotheses (Event: triggered when users complete a specific action like visit a page, download a .pdf, or sign-up for your product).
- Establish a baseline. After you’ve formed hypotheses on your customer life cycle, establish a baseline from your current dataset (ex. over the past 90 days ~20% of our newly activated users came from Facebook campaigns).
- Keep it simple. In a recent article 500 Startups recommended using only 3 events. I tend to agree.
Here are various business model funnels using just 3 events:
After you’ve defined your measurement plan and ideated experiments, you need to prioritize them. A good question to ask here is what experiments are likely to have the biggest impact? Pitch ideas to your team and peer review them to prioritize.
Also, you’ll need to plan ahead for things like:
- Setting up tracking codes.
- Design work.
- Naming conventions.
- Multiple sub-domains.
- Responsive web design.
- Creating multiple views of your data to accommodate filters (i.e. one unfiltered, Master View and a Test View is considered best practice for Google Analytics).
Contrary to popular belief, mindlessly rummaging through data does not actually help you accomplish anything… So don’t stare at your metrics every minute of every day. You’re wasting time.
Use the hypotheses in your measurement plan to guide your analysis. This way you’re goal-oriented, time efficient and more likely to find valuable insights.
If you learn something about your experiments, share it with your team! It may come in handy in future. You could even apply it elsewhere.
Your Leaky Funnel:
Ask ‘why’ during your analysis to identify areas of improvement. If you spot a high bounce rate or see users drop off at a stage in your on-boarding process, ask why! To determine if your experiments are successful, ask yourself if they confirm or disprove your initial hypotheses, and to what degree.
Were your initial hypotheses proven correct or were they proven wrong? If they’re proven wrong, use your data to find the cause, refine your hypotheses and create more experiments. Continue this process with your team until you start to see positive effects.
For a deeper dive on experimentation, growth guru Sean Ellis has a great presentation on High Tempo Testing with a Growth Model:
Improvements Over Time:
Many of your improvement experiments won’t work and some might be ‘out-of-the-park’ successes. The key is to identify when something is working and double-down on it. Apply resources to things that work and learn from the ones that don’t.
- Define business objectives, problems challenges and opportunities.
- Measure growth experiments at different life cycle stages (i.e. AARRR).
- Analyze and share results.
- Improve product/service based upon findings from analysis.
- Control improvements by doubling down on what’s working.
Thanks for reading! I’d love to know what YOU think. Tweet your comments to me @dannyjpwilliams or comment below. :)
Check out thoughts from Venture for Canada and its Fellows here:https://medium.com/@VentureforCanada