@andrewchen

Originally answered on Quora:

Day-to-day, I would break down what a growth team does into two major buckets: 1) Planning/modeling and 2) Growth tests.

But first, you need a great product
Let me note that if people aren’t using your product, then you’re wasting your time spending too much time optimizing growth. You need a base of users who are happy and then your job is to scale it.

With that caveat in mind, let’s start with the planning activities:

Planning and model building
The planning/modeling side of things is really about understanding, “Why does growth happen?” Every product is different.

• You might find that people find you via SEO and then turn into users that are retained via emails
• You might find that people come to your site via web and then cross-pollinate to mobile, and that’s the key to your growth.
• You might find you need to get them to follow a certain # of people.
• You might realize they need to clip a certain # of links to get started.

These are all things that are product-specific, so I can’t give specific advice in this answer, but this is the foundation for understanding why your product grows. You can come up with a model by looking at your flows for how users come into the site, by talking to users, and by understanding similar products. You can look at successful users and unsuccessful ones.

Once you have a good model, you can create more specific criteria in evaluating the outcome of a good or bad growth project. Your mental model doesn’t have to be perfect at first- the goal is just to get started. As you execute your project successfully, if your growth goes up, then your confidence will grow. (Or you’ll have to revisit things if you keep improving that one metric significantly but overall signups doesn’t go up)

At a more tactical level, eventually this model gets more fine-grained and you can start thinking about individual things that you can change to increase overall growth. Ideally you can model a lot of this in a spreadsheet so you can do scenario-planning around what works and what doesn’t.

The goal is to create some kind of feedback loop that results in sustained growth. Maybe you buy ads, make money, and then reinvest even more in ads. Maybe you get people to create content, driving SEO, which brings in more people that create content. Or maybe you have something invitation based. The important part is to model this process and its component parts.

Project execution
Once you have a model for how to drive your growth, the next part is to actually come up with a bunch of project ideas that can make those numbers go up and to the right. Ideally you can do lots of A/B tests for pretty short ideas that prove out the concept. If it works out, then keep investing.

For something like this, you’ll need a bit of A/B testing infrastructure, a lot of creativity, and some dedicated engineers to get the tests out there.

Because the majority of A/B tests don’t do what you want (maybe the number is <30%) as a result, you’ll want to have many, many A/B tests going at the same time so that you get a couple winners every week. Sometimes people do 1-2 A/B tests per week and then complain that it doesn’t work for them – they probably need to 5-10X their A/B test output in order to get a win or two per week.

To execute each growth project, you may also need to develop some instrumentation around tracking where users come from, and what they do. This can be a bunch of SQL databases and reporting at first, but might move to something fancier later on.

Eventually, the results of these tactical projects feed back into the uber model – you have to constantly reevaluate your priorities and understand which places in the product are the most leveraged in driving growth. So there’s a feedback loop of jumping from the strategic to the tactical, and back.

Summary
To summarize the above:

1. Have a solid product where your users are happy
2. Coming up with a model for how your site grows
3. Trying out ideas and deploying them as A/B tests
4. If the site grows, then try out more ideas. If it doesn’t, rethink the model in step 1 because it might be broken

Hope that helps.

Like this post?
Get new updates via newsletter..