How To Build A Kick-Ass Growth Team

sensei.jpg

“The bleeding edge today is Growth Hacking and Data Science, and it’s where analytics, statistics, computer science and marketing meet. Growth Hacking quintessentially looks for platforms that provide one to many relationships (read: one to millions), and develops smart ways to harness them quickly.

Tech companies are literally hiring rocket scientists to exploit these channels and acquire customers quickly before they get crowded and the advantage fades as they get costly or restricted. If you want hypergrowth, conventional marketing won’t get you there, because traditional marketing methods involves absolute, linear growth and not relative, exponential growth.”
Matt Barrie - Freelancer.com CEO

In this blog, I will share how I built our Freelancer.com growth team from ground up.

So I don’t have online marketers in my team. That’s not because I disregard their existence and importance but mainly because what we do in the company is very mathematical, scientific and calculated. Marketers mostly being mathematical neanderthals are the exact opposite.

So I have a team of twelve - comprised of product managers and data scientists.

1) 4 University Medallists/Valedictorians (I used to have 6)
    All the twelve did graduate with High Disctinction / First Class Honours
2) 4 PhDs
3) Mechatronics/Robotics Engineer
4) Computer Scientist - Machine Learning Expert
5) Computer Engineer
6) Mathematician/Statistician
7) Quantum Physicist

So what are some of the traits you want?

1) They are engineers and should know how to program

All the growth hackers in my company had to go through the same intensive interview process that the normal engineers would have gone through on top of separate analytics/products interview.

Why? Well, they are different kind of breeds - hybrid of marketers and engineers as Andrew Chen would have kindly put it. I do think that they should be strategists too and hence our team plays a very critical role in the company core product strategy and decision making.

Concrete Example: AirBNB managed to automatically post on Craigslist without any API. If you are not a programmer, you woulld not be able to do that. Needless to say, this is how AirBNB achieved its initial growth.

Additionally, there won't be any context switching. You don’t have to go back and forth to engineers to ship minor tweaks/fixes. Our team ship about 15-20 A/B tests a week and if we are lucky, we will get 2-3 minor improvements. So if we are to involve core engineers for those, imagine how much longer this would take.

2) Avid Tech-News Reader

To put it simply, if you don’t read hackernews or techmeme or techcrunch or e27(if you are in Asia) or other sites alike, you shouldn’t be in the team. This is our university.

3) Hire The Best

I have one philosophy: You can't go wrong with university medalists. At the very least, they have proven that they are very smart and willing to work hard - you can’t be graduating top of your class if you are not committed.

4) Passionate - Hungry - Driven

A smart manager always hires people smarter than him/her, the best you can afford. Look for passionate, hungry and driven team members. I'm more comfortable hiring 22 year old graduate who is smart, deeply technical, passionate, hungry who is easy to get on than veteran "outsider". The average age of our growth team is about 23 years old.

5) Knowledge in Statistics & Data Mining

You probably don’t need the whole team to be experts in statistics and machine learning but all of them should have high level understanding of those fields like for example:

- A/B testing and its statistical significance
- Binary state classifier and the applications of it
- Regression Analysis
- etc

So the idea is to have thinkers in your team and have these data scientists to validate the ideas by series of number crunching.

You have to track basically every improvements that you push. You have to be able to quantify whatever you do have impacts on the core metrics you are trying to improve. That’s why in Freelancer.com we have thousands of graphs in the dashboard created by the team and it is available to everyone in the company because we love transparency.

Oh yeah, you will also be amazed at what insights can come from Machine Learning.

In the case of the identifying paying customers for example:
ML - Binary State Classifier e.g. Random Forest Classifier can essentially predict whether a user is going to be a paying customer or not. Even better, it can also tell you - by weighting - which attributes to tackle so that more users are paying customers.

So what kind of interviews/questions you ask in hiring?

1) Technical Engineering Questions

I will probably start them off with engineering related interviews - the normal interviews software engineers would have gone through. Algorithmic related questions are probably the emphasis here cause you want to know how they think.

2) Basic - Advanced Data Analytics Questions

Depending on hiring product managers or data scientists, you probably need to ask them basic to advanced data analytic questions such as:

- What does 95% Confidence Interval mean?
- How would you run your A/B tests if there are over 10 simultaneous tests running?
- What are Type1 & Type2 error and why is it important in data mining?
- If you can have any data from Google, what would you yearn the most for our company and why?
- What is Expected Value?
- What is curse of dimensionality in machine learning and how would you go to solve it?
- Say you have the whole log file of our website, what insights can we have from it?

3) Random General Knowledge Questions

I also like to ask general random questions as it shows how diverse and versatile they are
- What is Subprime Mortgage crisis and why did it fail in hindsight?
- What is the Big Mac Index?
- Why do you think milk powder can is cylindrical and not a box?
- How much wood would a woodchuck chuck if a woodchuck could chuck wood? or, explain in detail why a query run at the second time takes faster than the first and for bonus, how can you circumvent this to a certain extent to benchmark performance?
- What do you think of online marketplace that allows users to outsource small jobs and tasks to others in their local area e.g. taskrabbit.com, airtasker.com? What problems do you see from those businesses?

4) Case Studies

Finally, run them through some case studies preferably problems that your current team has encountered and subsequently solved.. This may include funnel optimization problem for example.

I hope you enjoyed my first of many posts to come :).

/