Alex Schulz (part 1): an introduction to growth hacking

Stanford course CS183B: How to start a startup . Started in 2012 under the leadership of Peter Thiel. In the fall of 2014, a new series of lectures by leading entrepreneurs and Y Combinator experts took place:



Alex Schultz: I watched lectures on this course, the content of previous speeches is amazing! And now you have to suffer me. Let's see how things go.
')
Unlike Paul, who, speaking at the Q & A session, said that if he was in college now, he would have been involved in physics, I had given my own will in this regard at the time. In Cambridge, I studied physics. It seems to me that physics is an amazing subject, it gives you skills that can really come in handy in other areas of life, but today we will not talk about this.

I paid for the training by making money on internet marketing. In the 1990s, I started with SEO. I created a website about paper airplanes and took a monopoly position in this small market [Alex is obviously making fun of Peter Thiel, who in previous lectures talked about monopolies].

But when you create a startup, you should also keep in mind the possibilities of your chosen market for subsequent growth (the long-term prospects of my chosen market were disappointing). But I really learned how to work with SEO. In those days, we used the Alta Vista system, and the whole SEO optimization was to put white text on a page on a page on a white background. Having indicated the word “airplane” 20 or 30 times in such a text, you could take top positions in the Alta Vista search results. So did SEO optimization in the 90s. It was a snap to learn.

When I went to college in the Faculty of Physics, I thought that thanks to the paper-airplane theme I would look cool. In fact, I turned out to be the most boring botanist in the group, so I made another website dedicated to cocktails - in the process of creating it, I learned how to program, and the site later became the most visited resource on cocktails in the UK. I was really into SEO when Google appeared on the market. With the advent of Google, businesses had to worry about the place in the search results, and that their sites were links to other resources. True, at that time, it was possible to get to the top of Google’s issue by having just one link to the site from the Yahoo directory and all the same white text on a white background on the resource itself.

When Google launched AdWords, I was just starting to work in marketing. I bought paid clicks from Google and resold them on eBay with a small margin of about 20% using their affiliate program. This process literally pulled me into the topic of startups, into working with what is now called growth hacking , growth marketing.

It seemed to me that this is the same Internet marketing, I just need to use all possible channels and, thus, to achieve any desired result - this is what I paid for in college, and this is how I turned from a physicist into a marketer - turned to the dark side Forces.

So what do you find most important for growth? You have listened to a lot of lectures, you have been told about this thousands of times, so what do you think is most important for the growth of the company?

Audience: Steep product.

Schulz : And what does such a product give?

Audience: Customers.

Schultz: What do you need from these customers?

Audience: To have more of them around the world.

Schultz: Exactly, you need to retain customers. This is the most important element of growth. Now, Facebook has a great growth team and I am very proud to work in it, but the point is, we have a fantastic product. The opportunity to work on Facebook growth is a huge privilege, because we are promoting what the world really wants to use, and this is absolutely amazing. If we manage to attract a client and work with him, he remains on Facebook.



I acted as a startup mentor many times. Most of all I liked working with AirBnb, but I worked with Coursera, and with other companies that did not achieve outstanding success. But the essence in each of these cases is the following: if you build the “Percentage of active users per month against the number of days since the user’s involvement” curve, it turns out that your retention curve is an asymptote that tends to a straight line parallel to the X axis, then you have there is a viable business and product, and your product fits the chosen market niche. But for most companies that “soared” on the wave of growth hacking, virality and similar things, the retention curve eventually asymptotically rushed to the X axis and eventually crossed with this axis.

Now, when I talk about it, they answer me: “Well, of course, when you joined Facebook, the number of its visitors was already a million people a day” or “At that time you had 50 million users, this is a huge number of people, that you already had large volumes of information to ensure further growth. ”

But we used the same methodology [for retention curve estimates] for the B2B segment, attracting users to subscribe to advertising messages in order to understand how we can ensure growth in this market as well. And in that sense, when I joined Facebook, our product was only three days old. For 90 days of launch, we were able to apply this technique and understand what the value of one advertiser will be for a year - our predictions for the first year were justified with a 97% accuracy. Therefore, monitoring your user retention curve is very important.

As you can see, this red line represents the number of customers who have already used your product for a certain number of days. So, you have a lot of users who have worked with the product for at least one day, but if your product has a year, then you do not have a single customer who has been with you for 366 days, right?

Then you rate users who have worked with your product one day. How many percent of them are among the active users per month? Of course, in the first 30 days this value will be 100%, because all those who stayed with you for a day appeared in one month. But then comes the 31st day. Count all users on the 31st day after registration: what percentage of them are among the active users per month? Repeat the procedure on the 32nd, 33rd, 34th day. This will allow you, with only 10,000 customers, to really understand what the user retention curve will be for your product. And you can tell if this curve is like an asymptote.

The curve will become less and less smooth closer to the right edge (on the slide I show an approximate version and do not use real data), but you can still understand whether the curve is approaching a straight line or not. If not - do not use growth hacking tactics, do not try to increase virality, do not hire specialists for this.

Focus on product conformity to the market, because ultimately, as Sam said [ here and here ], opening this course: the four components of a startup are an idea, a product, a team, and implementation. If you have a weak product, it makes no sense to work on growth, because your company will not grow. The problem number one that I came across - and this is true both for new Facebook products and for startups - is that the product does not fit the market, and its creators do not understand this.



The next question they ask me again and again is: What percentage of user retention can be considered a success?

Of course! This figure can be as high as 5% (although I think Facebook’s figures are higher). But in this situation, your business cannot be called successful. This question annoys me, because, as it seems to me, you yourself can answer it.

I like one story, I can say that this is one of those stories that I will tell here “not in the topic” (there is a link to the text in my presentation), although its ending will fit very well into the framework of our conversation. Right here [on the slide above] I put a photo that was published in Life magazine in 1950, the result of one of the tests of the Trinity atomic bomb. So, there was a man named Jeffrey Taylor. It was a British physicist who, by the results of his work, won the Nobel Prize.

Just by studying this photo, he was able to determine what the power of the American bomb was. The Russians also published similar pictures based on metric considerations. Metric considerations [dimension analysis] is one of the best skills I have mastered while studying physics in the UK.

Dimensional analysis is that you evaluate the metrics that are involved in a particular task: you need to figure out the amount of energy. Guided by the equality 1 N = 1 kg · m / s ^ 2, you need to calculate the number of kilograms, meters, seconds, and you are trying to determine how to get these values, based on the data that you have. The mass can be expressed by estimating the volume of the sphere - based on these considerations, Taylor was able to calculate the power of the atomic bomb of the Americans and the ratio of the power of their bomb and the bomb of the Russians. In fact, he revealed one of the greatest secrets of the time.

It was a daunting task. But finding out what the percentage of retention of Facebook customers will be is much simpler. How many people use the Internet? 2.3-2.4 billion? Okay, let's take into account that Facebook is banned in China, how long will it be?

Audience: 2 billion

Schulz: So 2 billion people use the Internet. According to Facebook, the number of active users of the service is 1.3 billion. You can share these values, but it will not give you the right answer.

Of course not! But you’ll get a rough estimate of how Facebook’s user retention rate might look. If we can convince every Internet user to become our client, you will understand that the retention rate has grown. The same can be said about WhatsApp. They announced that the number of active users of their service for the month is 600 million people.

How many people own smartphones? You can calculate it - no one hides the numbers. They can tell you how great the WhatsApp user retention rate is. Amazon has a pop-up ad message that captures almost every Web user in the US. You know how many people access the Internet in the US, and you know how many of them are Amazon users, the company freely distributes these numbers.

Within various vertical structures, to create successful businesses, different extremes of customer retention are needed. If you work in e-commerce, calculate retention based on the number of active users per month, and your customer retention rate is 20-30%, then you are doing very well. But if you work in social media, and less than 80% of the first group of registrants remain, your social network will not become massively popular. It all depends very much on a particular industry. In this regard, you need tools that can answer the question with whom you can compare yourself, so that you can calculate a competitor's performance and see if you are approaching a level that can be called successful in your vertical.

Hold is the most important thing that keeps a company growing. It is formed if you have a great idea, a great product that implements this idea and meets the needs of the market. Therefore, when we assess the level of retention of users of a particular product, we look to see whether those who installed this product remain with the company in the long term, performing normalization based on a cohort analysis . This is a good methodology for evaluating your product: you are therefore constantly looking for the answer to the question of whether the first 100, 1000, 10,000 people will stay with you in the long run.

So how do you get started on growth? Suppose your product is ideally suited to market needs. You have created an ecommerce site, 60% of users return to you monthly and make purchases, which is certainly amazing. What do you do next after you decide at some point that you need to scale? At this point, as I understand it, the teams responsible for ensuring growth come to the rescue.

But I have a different point of view: if you work in a startup, do not create a growth team . They are not needed in startups. Such a team should be the whole company. The CEO must be the leader of this team. You need someone who will give you a guideline, tell you where to go, and, based on my experience, this person can only be the one who leads the company.


If you are creating a messenger application, the number of sent messages is likely to be your most important indicator. If people use your application once a day, it may be great, but it means that for them this is not the main mechanism for sending messages - so Jan [Kum, WhatsApp] began to publish these numbers.

In AirBnb, this indicator refers to the total number of days for which apartments and rooms are booked through their service for a certain period — based on this data, they create an infographic that is published in TechCrunch. They always compare their numbers with similar indicators for the largest hotel chains around the world. Each of these companies has its own reference point. Not in every vertical this guideline can be the number of active users per month. For eBay, this was the guideline for total sales: how many products do people actually buy through eBay. Although external analysts judge eBay solely by their revenue.

Benedict Evans conducted an excellent analysis of the activities of Amazon - it is very interesting to look at their marketplace and compare the performance of its work with their traditional business. eBay is the marketplace, right? And although eBay is judged by their revenues, the total sales volume through their site exceeds their revenues 10 times. When I worked in eBay, they paid particular attention to this indicator. Every company, thinking about growth, needs its own guideline. And, working on scaling, you, as a leader, are very important to establish this landmark.



The reason this is so important is that as soon as more than one person starts working in your company, you stop controlling all the others. I assure you, now, heading a department of 100 people, I cannot control them. I can only influence them.

I can get someone to do one thing, but the remaining 99 will do what they want. And in such a situation, it is not obvious to everyone what the main goals of the company are. It’s very easy for eBay employees to say, “You know what? Let's focus on revenue ”or“ Let's focus on the number of people who buy something on our site ”or“ Let's think about increasing the number of potential sellers. ”

But Pierre [Omidyar], Meg [Whitman] and John [Donaghou], being different leaders by nature, always said: “No, we care about the total amount of sales that are conducted through our website, the percentage of transactions made through our website “From all transactions in the e-commerce segment, this is what is really important for our company.”

This means that in any situation it was clear to all employees that this was not about revenue, but about total sales, or not about how to get more people to register on the site, but how to make them active users in the long term.

A great example of a similar policy from my experience in eBay

In 2004, we changed the method of payment for attracting new users to our partners. Affiliate programs to attract users now look a bit old-fashioned, but the idea of ​​attracting is left - you pay those who catch traffic to your website, but now it’s more about getting access to big market players who do it themselves.

We paid for confirmed registered users, so all our partners were interested in attracting those who subsequently register on eBay. We changed our payment model and began to pay for activated confirmed registered users. The user had to confirm his account, and then offer a price to buy a product, or buy a product, or put something up for sale, so that we would pay for its attraction to our partner.

During the day after these changes were made, we lost about 20% of confirmed registered users who were attracted by our partners. But the ACRU (Activated Confirmed Registered Users, activated confirmed registered users) fell by only 5%. The relationship between CRU (Confirmed Registered Users, Confirmed Registered Users) and ACRU has grown dramatically, and the growth in attracting ACRU has accelerated significantly .

The reason for this is as follows: if you want to increase the value of the CRU index, and you know that someone needs a trampoline, for example, you give that person a link to the eBay registration page so that the user first registers, then confirms his identity and then already bought a trampoline.

But if you want to increase the value of the ACRU indicator, you give the potential client a link to the eBay page with search results for the trampolines on the marketplace - the user likes something, he is interested, registered and buys the goods. If you develop only the CRU metric, users will not experience the slightest interest: when they get to the eBay page, there will be no “magic”.

Therefore, the next thing you need to think about is how to create this “magic” that will force users to “attach” to your service.

[ The second part of the translation of Alex's lecture ]