Professional Documents
Culture Documents
to customer-centered growth
Contents
03 INTRO
04 SECTION 1: BEFORE YOU START
The two approaches to product growth
Why you need an experimentation process
Set your strategic goals
The four key documents you need
insights to discover And that’s why so many people are looking for the one tactic, secret, or hack that will skyrocket their growth
curve. But we believe that focusing on specific tactics is the wrong approach.
the tactics that work In this eBook, we’ll make the case that, instead, you should focus on implementing a rigorous process of
best for you. experimentation. And that the key to driving sustainable growth is using customer insights to discover the
unique combination of tactics that will work for your product.
The material we’ll discuss is best suited for PMs working on products that have already found product/market
fit and have a good handle on retention. You need to have built a “must have” experience that customers love
before you transition into the growth phase.
Ideally, you already have a substantial amount of traffic, you’re ready to scale, and you want to make sure you do
it right—by putting the user at the center of the process because that’s what pays off in the long run.
And while every company will adapt this process to fit their own needs, our goal is to give you a
step-by-step guide for implementing a highly experimental, customer-centered growth process into
your organization. In this eBook, you’ll learn how to:
It’s a no-brainer. Your growth efforts will be much more effective if you establish a continuous, consistent
cadence of experimentation to identify the leaks in your funnel and come up with test ideas based on
multiple sources of data.
Peep Laja uses a great metaphor: “Would you rather have a doctor operate on you based on an opinion,
or careful examination and tests?” You’re like a growth doctor, and your product is your patient.
What works for other Distribution channels are There are no silver bullets.
products won’t necessarily changing faster than ever.
No one-off tactic is going
work for yours.
Channels that work for you to solve all your growth
Your business model, today will likely be saturated problems. The tactics that
product, and audience are and far less effective 12 months do produce explosive growth
fundamentally different from now. To stay ahead of are usually the result of
from anyone else’s. the curve, you have to always a combination of lessons
be experimenting to discover learned from previous
emerging channels. experiments.
Your growth efforts will be more productive and successful once you
move away from testing random tactics and start approaching it as the
scientific process it is:
Set strategic
goals “ There is no one right or perfect growth
process. The important part is just to have
Systematize Gather and one, stick to it, and improve it over time.”
analyze data
- Brian Balfour, VP of Growth, HubSpot
Analyze Generate
results hypothesis
This process isn’t perfect, but it will give you a solid foundation to build on.
Take it, copy it, and adapt it to your own team. But whatever you do, use it.
How will you measure success? Next, you have to decide how you’ll measure success. What quantifiable
results will you use to measure whether or not you have achieved your
Start by taking a step back and setting your high-level strategic goals. goals? For example, if you decided to focus on improving your activation
Ask yourself this question: rate, the quantifiable results you measure might look like this:
What’s the one area of the customer lifecycle that, if you 35% activation rate
made substantial improvements, would have the biggest
impact on your growth curve right now? 40% of users come back two times per week after week 1
The Objective and Key Result (OKR) framework is a goal-setting method used by many growth teams in Silicon Valley.
It was first invented by Intel and popularized at high-growth tech companies like Google, Zynga, and LinkedIn.
If you’d like to learn more about how to set OKRs, we highly recommend Christina Wodtke’s guide.
There are four key documents you’ll use throughout this process to
keep it driving forward:
1 Backlog
The purpose of the backlog is to provide a public place for your
team to dump and record all the growth ideas they come up with
as they go through the experimentation process so they can
come back to them later.
2 Pipeline
This is a list of all the experiments your team has previously
run, are currently running, and are going to run next. It allows
anyone on your team, including your new hires, to see every
single experiment you’ve run and the lessons you’ve learned
along the way.
3 Experiment docs
Every experiment gets its own document that outlines the objective,
hypothesis, experiment design, results, analysis, and lessons
learned. Experiment docs force your team members to critically
think through their ideas to implement them successfully. To
download an example experiment doc template, click here.
4 Playbooks
These are step-by-step guides that anyone can follow to repeat
what you’ve learned without having to talk to you.
Now, let’s dive into the details of each step of the experimentation
process.
But not all data is equally useful. You don’t just need any data; you need actionable data that gives you insight
into how to optimize your specific area of focus. For example, if you’re focused on improving activation,
break down your product’s onboarding and activation flow into its individual steps. And then collect data for
each step.
Narrowing your scope down to the individual steps allows you to gather the right data that will give you insight
into which levels you can pull to impact your area of focus—and prevent you from getting overwhelmed by
having too much irrelevant data.
Start by gathering quantitative data that shows you where the problem is—where the biggest dropoffs are
and where your funnel is leaking.
Once you know where the issues are happening, you can gather qualitative insight to understand why.
What’s causing people to have doubts, uncertainties, and hesitations?
Once you know the source of the problem, you can come up with hypotheses to test about what you think
would remove the friction.
As you gather and analyze the data you collect, you should be dumping all the issues you identify and experiment ideas you come up with into your backlog.
The next step in the process is to translate those findings into a prioritized pipeline of hypotheses to test.
These data sources are a combination of ideas from a book called The Innovator’s Solution by Clayton Christensen and the ResearchXL framework, originally developed by Peep Laja.
Now that you have a backlog full of potential tests to run, the next Once you’ve crafted the hypothesis, the next step is to prioritize each
step is to turn your findings into prioritized hypotheses. experiment based on three core factors, and, of course, to document
them in your experiment doc.
Every test idea needs to have its own experiment doc. And in this
step, you’ll use it to document five things: the objective, hypothesis, • Probability
probability of success, resources needed, and potential impact if the What’s the probability that this experiment will be successful? Use
test is successful. The purpose of this exercise is to make you think a rough estimation of low = 20%, medium = 50%, and high = 80%.
through why you want to run this experiment and whether or not it’s You might give your test a high probability of success if it’s based
actually a good idea. on something you learned from a previous experiment. But if
you’re testing a new channel that you have no experience in, you
First, describe what you’re testing, and why you’re testing it, in one might rank it low.
or two sentences. This is your objective. Describe it so that someone
else on your team will understand what you did if they read it a year • Resources
from now.
Roughly estimate how much time you’ll need from your design,
Then, you need to write down a hypothesis for every experiment. engineering, and marketing resources to implement the experiment
Surprisingly, most people don’t do this. Here’s an example of what (i.e., 1 hour, 1 day, 1 week, 1 month, etc.).
your hypothesis should look like:
• Impact
If this test is successful, what’s the total impact you expect to see?
If successful, [variable X] will increase What do you predict the results will be? How much will it improve
by [impact] because [assumptions]. the targeted metric?
Business
1 2 2
that would force you to rerun the test.
? Okay, now that you’ve outlined your prioritized experiment, then next
step is to run it. There are a few things you need to do to make sure your
tests are statistically significant:
LOW HIGH
Complexity / effort Test with the appropriate sample size
- Values vs. complexity quadrant. Jim Semick, the founder of ProductPlan. Test for an entire business cycle
Once you’ve run your controlled experiment, the next step is to analyze the results and extract the lessons learned.
As you determine the success or failure of an experiment, look at three factors and document them in the experiment doc: “Why?”
Impact. What were the results of the experiment? is by far the most important
question you can ask. The
Accuracy. How close were the results to your hypothesis? answers you come up with
will lead you to new ideas
Why? Why do you think you saw the results you did?
and experiments to run next.
“Why?” is by far the most important question you can ask because the answers you come up with will lead you to new
ideas and experiments to run next. Use this question to extract as much insight as possible and, again, document all the
lessons you learned in the experiment doc.
Once you capture what you learned, make a list of action items that consists of tests you want to run next. And then transfer
those ideas into your backlog. Now that you’ve gained a deeper understanding of your customer, product, and channel, go
back to the experiments in your pipeline and adjust the predictions you made based on what you know now.
Feeding what you learn back into your backlog and predictions is the most critical part of this experimentation process—
it’s what allows your experiments to become increasingly accurate and successful over time. If you skip this step, you’ll fail
to build on your experience and you’ll constantly have to start over from square one. You’ll essentially be throwing away
the value you’ve created.
This is the final step in the process. Once you discover a successful experiment, try to implement it as a new product
feature or initiative. For example, if you only ran your test with 15% of your audience, then launch it to all of your users as a
new part of the product experience.
But sometimes your team will discover things that are more manual and can’t be productized—for example when a growth
marketer on your team discovers an effective method from promoting a piece of content to a specific segment of your
audience. If that’s the case, turn it into a playbook, which is a step-by-step guide that anyone can use to repeat what you
did and get the same result. Make your playbooks publicly available to everyone on the team to read, comment, and
collaborate on.
This allows you to capture, retain, and share the knowledge and value you create with successful experiments.
your backlog and Repeat steps 1-5 as many times as you feasibly can in your 30-90 day period. Take the lessons you learn from
analyzing experiment results and continuously feed them back into your backlog, and use them to refine the
using them to adjust predictions in your prioritized experiments.
your predictions is This feedback loop plays a critical role in making your process more accurate with each successive cycle.
increasingly accurate the effectiveness of your efforts from a macro level and think about how you can optimize them. You need to
look at three things in particular:
time. What’s your team’s ratio of successful experiments to failures? This number should be getting better over
time.
2 Accuracy
Are your hypotheses getting more accurate as you gain more experience in a channel? The longer you
work in a channel, the more accurate your experiments should get.
3 Throughput
How many experiments you do you run in a given period of time? How can you increase the velocity?
You can improve those three things by constantly optimizing your process, your team, and the tools you use.
And as a result, your experimentation efforts will become more accurate and more successful over time.
The key to this process is that your team puts the customer at the center
of it—constantly gaining a deeper understanding of your customers,
how they use your product, and which channels are effective for reaching
them. As Sean Ellis eloquently said:
If you don’t deliver a “must have” experience, it won’t matter what process
or growth tactics you use. It’s hard to justify investing in top-of-the-funnel
acquisition if every new user eventually leaves. Bottom line: the better
your customer experience is, the more customers you’ll retain, the more
people they’ll share your product with, and the easier it will be to grow.