Class 9 Notes Essay — Reid Hoffman, John Lilly, Chris Yeh, and Allen Blue’s CS183C Technology-enabled Blitzscaling course
Here is an essay version of my class notes from Class 9 of Stanford University’s CS183C — Technology-enabled Blitzscaling — taught by Reid Hoffman, John Lilly, Chris Yeh, and Allen Blue. Errors and omissions are my own. Credit for good stuff is Reid, John, Chris Yeh, and Allen’s entirely.
This class was a talk with Allen Blue and Reid Hoffman— two of the co-founders of LinkedIn — on lessons learned from scaling LinkedIn.
I wasn’t actually able to attend this class in person so a big thank you to Ryan McKinney for helping to record this class and share the audio with me.*
Video of the class, notes are below:
I. The change from OS1 to OS2 to OS3
- During OS1 (The Family) — You are trying to figure out if you can create something customers want, that is unique.
- At OS2 (The Tribe) — You are trying to figure out if you have product market fit, and when to put on the gas to occupy the full market
- OS3 (The Village) — This is when you decide to scale up. When the company reaches 150 people is usually when you go through the next substantial shift.
II. Why scale?
- 150 people is Dunbar’s number— a suggested cognitive limit to the number of people with whom one can maintain stable social relationships — this is estimated to be ~150 people.
- Once your company goes beyond 150 people, this is the time when people at your company don’t know everyone else in the company and coordination becomes challenging.
- At OS3 the company has a sense of real traction, company scale, and either has a revenue stream or can see where the revenue stream will come from.
- The coordination challenges causes the company to need to scale up and support the larger organization.
- If the organization is scaling 50–100% per year, you are essentially building an unstable organization. Typically companies will change their organization structure many times over during this period.
III. Key considerations of when to scale
- This is when market size becomes an important consideration. One question is if your company scales up, does it grow into a market that is worth taking over?
- At this point you need a plan for access to large pools of capital — This can either be revenue (in Google’s case) or fundraising (Airbnb, Dropbox, Uber, etc).
- Companies scale fast usually because competition becomes much more intense. However there are examples — for example in the Minted Story Mariam Naficy decided not to hyperscale because her competition were the old school stationery companies. No one was trying to “take her out.” LinkedIn was very similar in this regard.
- During the first stages (OS1 and OS2), competition is less important because the market opportunity is not obvious. Once you get to OS3 — other companies can see more of the opportunity you can see (it becomes less contrarian) and this turns on the clock to compete.
- The decision of when to scale is very much a judgement decision — depends on competitive circumstances, market opprotunity, the value of being first to market, access to capital, etc.
IV. How and why LinkedIn made the choice to scale, in 2008
- LinkedIn launched in May 5th, 2003. They decided to scale 5 years into the creation of the product.
- During LinkedIn’s initial launch they made a decision to keep things trim (very much the same decisions Miriam made) — and they were very selective about hiring and raising capital.
- Before 2008 it was difficult to articulate what they were doing. In 2003 there was a site that launched called Friendster, and the only way to get the press interested in LinkedIn was to say they were “Friendster for Business.”
- After Friendster and Myspace came around, the average user started to understand what networks were and how to use them. In LinkedIn’s case, users started to understand how to keep a profile updated, why profiles were valuable, and the ability to use networks to find a job, etc.
- LinkedIn started off as a consumer company and moved towards enterprise because they found out businesses wanted to pay LinkedIn for recruiter access — moved towards hiring account managers, sales, customer success, etc — started to see early signs of product market fit.
- In 2008, they made a conscious decision that now was the time to move fast into the market opportunity — and win both the consumer and enterprise market.
V. The LinkedIn plan of how to scale
- Much of the below comes from a presentation that Jeff Weiner gave in 2010, while they were in the middle of the village stage, to the whole company.
- When Jeff joined (more on this below) he wanted to bring to the table the foundation for scaling. In OS1 and OS2 you don’t need as clear of a plan because everyone can work together directly. In the OS3 stage you need a far sharper and succinct business plan to manage a large organization — past Dunbar's number.
- The mission of LinkedIn which Jeff laid out was to “create economic opportunity for every professional in the world.” This became the same mission they had followed for the last 6 years.
- A mission is important because it becomes the touchstone which guides decisions. When you get to 400, 1,000, 10,000+ people, the key is to create a common language to make sure everyone is on the same page.
- Once communication moves past 100’s of people — there are many conversations within the organization that you can’t be a part of — how can you resolve conversations you aren’t a part of? You need a way to articulate things.
- Bringing in executives brings in a whole new set of skills which the LinkedIn team didn’t have before.
VI. Strategy of scaling LinkedIn
- The strategy of LinkedIn (as laid out by Jeff) was to connect talent with opportunity at massive scale.
- This includes both the growth of people joining LinkedIn and product market fit with recruiters who were paying for the product. The value to the recruiters was a large pool of passive candidates.
- They decided to find customers primarily with the creation and scaling of a sales force.
- Secondarily to their strategy also included “greater engagement across LinkedIn” and “developing secondary monetization strategies.”
- The competitive advantages are the “moats” which are hard for others to copy.
- In LinkedIn’s case, their moats were: a focus on the individual vs. the businesses, continuing to grow to critical mass, network density, concentration on data (profile completeness), and using data to drive their recruiter business.
- In the middle of their strategic bullseye was their hiring solutions product (aka recruiters). After this was developing products for marketing professionals, sales professionals, etc.
VII. The operating priorities of LinkedIn were
- Build a world class team
- Focus on product
- Expand globally
Question from audience: What were the most surprising things when Jeff came in and did this?
Reid Hoffman: The things which stood out were:
- The clarity and sharpness in which he expressed the plan
- He made this plan rock solid and kept improving and reworking it.
- This plan was the same plan we used internationally and externally. He repeated the key points of this plan over and over again (See Mozilla section)
- Jeff had a strong focus on culture and the mechanisms for driving culture. He used these to expand our company while maintaining culture.
Articulating the plan
Reid Hoffman: The difference with articulating the plan in OS3 vs. OS2 or OS1 was this. Here's how I was articulating LinkedIn before:
- We were building a network as a platform
- On this platform everyone's identity would be real
- We could use this platform for a number of different applications — helping people find and match with each other, helping people form connections, etc.
Jeff’s insight was these ideas work well with geeks but we were trying to build a company with 1,000’s of people. For this you need to articulate the plan with much more concrete and succinctness. We were still doing the same core thing but needed to come up with a way to rationalize all of this and get anyone to understand what we were doing quickly.
There are ways to fail as a 450 person company which just don’t apply during a small organization. The current tech giants (Facebook, Apple, Google, etc) aren’t there just because they had the right app, the right market opportunity, and just hung on.
There is a bunch of art and science on how you build an organization. How do you articulate a plan in which most people know how to understand and can coordinate amongst themselves? Also how do you build a culture which defines parallel action amongst a large number of people?
Question from the audience — Which of these things would be inappropriate in OS2?
Reid Hoffman: In the plan, things like “world class team” mean something different in OS3 vs OS2 or OS1. At LinkedIn we hired very smart people before — but now we had to operationalize things and put processes into place. For example, with hiring, we needed to create on-boarding processes, videos to help teach culture, interviewing practices, etc. We had to bring on executives who are managers of managers and so on.
The “world class team” means people who can adopt and create these kinds of practices. We aren’t saying the people we had weren’t a “world class team” — we were saying we needed to focus on people who understood and could implement scale mechanisms.
We noticed that the key pivotal roles — which were once occupied by generalists — either had to become more specialized or they would have to move into different roles. Generalists tend to be flexible, love to experiment, could figure it out on their own, weren’t afraid to take risks and attack new problems, etc.
During OS3 when we needed to take this part of the organization, grow it by 300%, develop dashboards and metrics to manage this part of the organization, all while improving operational efficiency — this kind of work tends to be done better by people who have high expertise in this particular field (specialists) rather than by generalists.
VIII. Scaling in the engineering function
Allen Blue: During the OS1 and OS2 our technical platform had been optimized for agility, being able to experiment quickly, abandoning failures, etc.
During OS3, we needed to shift to being able to increase our capacity and load by 10x-100x and build the systems to support what was working. During this period the attitude towards building software is very different than before.
Question from the audience: Going back to the generalist / specialists divide, how can you determine which type of person someone is? How do you know if a generalist could become a specialist?
During the OS1 family stage, everyone we hired were generalists. Good generalists were people who could: tackle a new problem, know how to sort it out, and were comfortable with different things. For example generalist engineers can work on the server side then move over to building an iOS app.
With specialists during hiring you want to ask:
- What specific projects have they worked on in the past?
- What did they learn from those projects?
- How would they reproduce those same results in a different org.?
For example in hiring a sales generalists you would ask what would happen if I threw you in this circumstance? For hiring a sales specialists you would ask: how do you define a territory? How do you manage your sales pipeline? How do you measure your pipeline? You want to look for people who have done these things before and can do them again.
Question from the audience: Do you classify people as generalists and specialists? Do you interview differently for each type of person?
Reid Hoffman: Yes we do.
Question from the audience: Have you made bad hiring decisions?
Reid Hoffman: Yes we have made lots of bad decisions. Generally companies have one of two choices when they decide to scale:
- Hiring anyone quickly, monitoring their performance, then firing fast.
- Be more careful on hiring and more careful on firing.
Most organizations which I have seen scale — do the second strategy instead of the first. They do this because they need to build a community within the team, and its hard to do this if you are firing a lot of people.
There are organizations which do the first pattern though.
One of the goals for this class is if you hire someone for the family, think if they would make sense for the tribe — it’s important to start to think about these kinds of questions early. Part of the challenge is if you have someone you really want to keep but doesn’t work for later levels — the only chance of keeping them is to start these conversations early.
If you can see issues coming up in the future you can take preemptive steps today — coaching, mentoring, learning from others, talking directly, etc. People can understand if you need to hire a person above them if they understand the decisions around it — better to have these conversations early.
This is also part of the reason why scaling is a lot of real work.
Allen Blue: Founders are a little different in the fact that they associate their success with the success of the company — not a given role. Founders tend to be willing to make changes (with their own role) and continue to do this over time for the benefit of the org.
Reid Hoffman: The short answer on hiring over stages:
- Now that you more data about the person from the stage before — would you hire this person in this role today?
- If you don't answer this question 100% yes — you need to figure it out or make a change.
Question from audience: For LinkedIn, 60M users in 2008 seems like a lot, why did LinkedIn not have to scale until this point?
Reid Hoffman: Before they knew how to massively grow the revenue line we were only investing what we made back into the company — primarily to grow the user base. In 2008 they got to a point where they knew how to scale in terms of capital.
I wrestled a lot with the term “blitzscale” which we got from the word “blitzkrieg.” I don’t like the word but it has a lot of good parallels.
Prior to blitzkrieg, all of the war was done through supply chains. You would extend your front to only what the supply chain would handle — it had a max speed.
The innovation in blitzkrieg was, it said screw the supply chain, whatever you could carry you carried to go fast without the slow supply chain behind you. Once you got the battle you either won big or lost big. If you lost, you collapsed because you had no supply chain (no backup, no ammo, no food). It was very much a gamble.
Similarly in a startup when you scale — you are really going to crank up the burn rate — hire a lot more people, really make a go at it. If you are wrong, it’s very painful, most likely the death of the company. Before you scale you really have to know where investment will come from (either revenue or VC) before scaling.
Question from the audience: Do you have a thesis about if bringing on an outside leader is critical for blitzscaling (Jeff Weiner, Sheryl Sandberg, Eric Schmidt) or can the old team adapt?
Reid Hoffman: Roughly speaking it should be a combination of internal and external.
If it’s all internal you tend to drink your own koolaid, unless you have a lot of experience in scaling. This is tough because very few people have the experience of going early to late stage.
If it’s all external you tend to lose all of the people who care deeply about the problems you are working on, the people who are emotionally committed, work 100’s of hours a week.
The art is to balance between these two. Some of it comes down to the founders recognizing what their key strengths and weaknesses are. External people (investors/board) can help with this dialog and critique with the founder.
When you are founding a company the question to your board/investors isn’t “am I doing a good job?” it’s — "What could I be doing better? What do I need to be doing that I am not doing?" You need to have an accurate judgement about this and about yourself.
One of the things which was impressive about Mark Zuckerberg (Reid was an angel in Facebook) was watching Mark grow. He was trading out execs as he was scaling along the different stages to figure out what skills complimented him best. At the end he found Sheryl who excelled in many areas Mark didn’t have expertise in.
IX. Execution plan
Reid Hoffman: One thing to add is in OS1 and OS2, we were focusing on lots of experiments. One of the key execution components in OS3 was to pick a few things to focus on.
Allen Blue: One of these components was going international. Going international was something that didn’t play a role in OS1 or OS2.
For LinkedIn the critical components were:
Each of these plans were presented on by a new executive, people we specifically brought on for OS3.
The background for the product plan was LinkedIn had been working on many of these items before, but never constructed it into a plan which 500 people would work on together.
The product plan included:
- Member growth
- Professional identity (profile)
- Search (for recruiter product)
- Knowledge sharing (Q&A, groups, etc)
- Hiring solution (the paid product of LinkedIn
- Marketing solution (the secondary paid product of LinkedIn)
- Monetization (payments, invoicing, etc)
Go to market plan
- Primary sales broken up by field sales teams
Another new component of the go to market plan and the OS3 stage in general is using dashboard and analytics to manage the company and forecast projections. You don’t really need this in the OS1 or OS2 stage.
- The engineering plan was to scale our product lines 10x.
- This happened primarily through focusing on developer productivity which included building our own internal tools.
- Before this, we would just build things for product market fit — now we had to build systems in place for engineers and operations — uptime, distributed computing, disaster recovery, security, etc.
- We had to change our whole stack and infrastructure, continue building our product, and scale our sales team all at once — it was very difficult.
Question from audience: How many engineers were you at, at this point?
Allen Blue: We were at about 200 engineers. One other reason why process is important is to to ensure we didn’t lose coherency across the product, tech, and sales offerings.
X. The CEO question
Reid Hoffman: Once you identify the scale challenges, one of the tough questions to ask yourself is: “Do we have the right CEO? Am I the right CEO?”
This isn’t a question about just now but it’s a question about what happens when you are 1,000 people, 2,000, 3,000, etc. The classic path is to wait till things are broken then try to fix it — this is much harder with organization problems. It’s better to anticipate where the organization is moving towards and then make the adjustments while you are going there.
The CEO directs how the whole organization works at scale. The two options are to either go into the steep learning curve or bring in a CEO/COO who has done this before. This is the same question with all of the other executive functions as well.
We fixed the ones which were absolutely critical — then used some of the early founders/employees who were generalists to fill in some of the other roles.
For myself personally (Reid), I was good at product and strategy but not so much in growing an organization. We originally hired Dan Nye as CEO when LinkedIn was 65 people. Dan was a good enterprise excellence person but after talking further, LinkedIn was at its core a consumer centric company and needed a CEO who had an affinity to this problem. Reid moved back into the CEO role then brought in Jeff Weiner to help. Very quickly Reid realized Jeff would be a good CEO and transitioned him to that role.
Allen Blue: Jeff knew we were going to need to grow fast — he worked to make sure our culture was relayed in the correct way. The challenge is when you hire in a distributed way — to make sure everyone hired the same in terms of quality, fit, match, interviewing style, etc. Jeff helped articulate a vision, mission, plan, etc but he also articulated a set of values and put these values into the hiring process. With these kinds of things, you can’t just talk about it either, they have to be embedded in how you manage on a regular basis.
People recruiting was also the #1 thing the engineers were worried about. When you are a small company, part of the pitch to engineers is to be able to define the tribe, the culture, processes, etc. When you are scaling up, this pitch doesn’t exist anymore. How do you now talk about the mission to get high quality people excited about LinkedIn?
Reid Hoffman: Another part of scaling is to make sure we had enough capital — enough money in the bank while they were about to throw on the gas and blitzscale.
The thought was to raise enough capital to smooth out our scaling challenge. This is much easier to do before scaling thann to do while you are scaling. Reid went out and raised the LinkedIn Series-D which was the last round before they went public.
Many people might be familiar with the LinkedIn Series B deck. Once you get to scale, the pitch changes to become much more succinct.
In the LinkedIn Series D deck was a:
- summary of revenue growth
- summary of user growth
- revenue model (for LinkedIn it was mean revenue per member)
- key execs and backgrounds
In later stage companies, the pitches are more succinct and much more model driven than in early stage companies.
For LinkedIn why we were so interested were:
- No cost of customer acquisition (users not recruiters)
- High margin on their recruiter product
- Highly scalable model (able to replicate to many users and recruiters)
- Network effects
- Huge market (recruiting, marketing, sales, business media)
Question from audience: How many companies fail at this stage? Is failing at this stage fatal?
Reid Hoffman: It depends on the competitive landscape. Many companies fail at this stage — sometimes it's product market fit, sometimes it's technology (Friendster failed in this way). I don’t know what the exact percentage is, we in technology tend to have a success bias and only look at the successful ones. At least in the social space there were Friendster, Orkut, Myspace, etc.
Question from audience: How did you approach new locations especially with the viral loop of email address, was this the same everywhere?
Reid Hoffman: We didn’t change the viral loop in other markets. Even in the early days we had a 50% international user base and we never tried going after people. It was only in the past couple years we have started doing specific changes for certain countries.
Question from audience: In terms of fundraising — how do you not fall into the temptation of just raising early and outspending the competition?
Reid Hoffman: By default, companies tend to raise more capital than less — more capital brings you more optionality.
The problem with raising too much capital are: less exit options, less discipline, setting too high expectations for new investors, it can complicate things.
On the other hand, a bigger war-chest is better than a smaller war-chest. The obvious tradeoff is dilution. In today’s market, it’s easier to get capital so founders are saying "let's go for it." Even if your valuation is above where it should be, you can use the capital to grow into the valuation.
This strategy works, only if when winter comes and financing dries up — you can get to a break-even state — because when you look badly capitalized, it's hard to raise money.
Allen Blue: Sometimes raising money at this stage is different than others — for example raising for acquisitions — common to start looking at acquiring companies during OS3 — LinkedIn didn’t, but it is common.
Question from audience: When you are at family or tribe stage, were you only hiring people who could make it to later stage? If not, did you have a discussion with them in the beginning?
Reid Hoffman: The short answer is, with very few people you have 100% confidence they can make it through multiple stages. Looking back on LinkedIn — some people broke earlier than expected and some lasted much longer than expected.
If you have the conversation — “there is no entitlement — even I’m not going to be the CEO forever" — we are defined by our mission and how well as a team we are reaching that mission. My job as a manager is to give you feedback, start the conversation, help you understand what you are good at and not good at, and helping you to find the best role possible.