On a Saturday morning in August of 2006, Sergey Brin and a team of Googlers flew to Los Angeles to meet Tom Anderson and his MySpace team. By the next afternoon, the two founders shook hands on a three year $900M contract. About twelve months later, I found myself as the product manager for the team in Marrakesh, Google’s executive board room, reporting the state of affairs to Eric, Larry and Sergey.
After signing the letter of intent, Google assembled a superb five-person team of machine learning experts and tasked them with improving ad targeting on MySpace and other social networks. At the time social networks like LinkedIn, Bebo, MySpace, Hi5, BlackPlanet and others generated about 40% of ad impressions for AdSense, but only 2 to 3% of revenue. The disparity was extreme.
When Sergey signed the contract with MySpace, Google hadn’t built the technology to properly target ads on social networks. But he believed it could be developed. And he was right.
Over the course of the next 18 months, the engineering team experimented and tinkered. They discovered that the language on social networks varies greatly from the language on the rest of the web. We threw away the dictionaries used for AdSense targeting and built new dictionaries using social-native dialects. The team modeled the dispersion of conversations through out the network in different ways - testing, testing, testing. Every week, we reviewed the experimental results.
Marrakesh is a long, thin room with grey walls, red couches. Two monitors hung on the wall at one end and Eric sat at the head of the table on the other side. When we stood up to present, we reported that in just a few quarters, the engineering team had increased revenue by greater than 10x. More importantly, they proved intent could be extracted from social media conversations at scale.
As social media usage continues to grow exponentially, users will pour out their opinions into conversations across the web in ever greater numbers. Within the morass of LOLs and emoticons, there is real value to be found.
Redpoint has invested in two companies which also pursue this idea: BlueFin, which Twitter just acquired, and Quantifind. Both BlueFin and Quantifind harvest data from the social web to glean meaningful insights and grow revenue for their customers.
We are only at the beginning of understanding the value of data contained within social media. I believe there is another 10x hidden in that tweet.
In baseball, on-base percentage is the best predictive metric for team success. In content production, the best predictor is user viewership behavior.
Because of Netflix’s scale, 25M+ subscribers, and because of its highly granular recommendation technology which measures viewership on a per household basis and includes all the metadata (actors, directors, plot) about each movie and TV show, Netflix has arguably the best return-on-investment calculator for content on the planet.
This content evaluation machine drove the decision to resurrect the cancelled “Arrested Development” series. Now a terrifically successful show, “Family Guy” was also cancelled. But it wasn’t the studio chiefs who decided to return the show to prime time. By the thousands, fans complained loudly enough to convince FOX of the inaccuracies of their data, rescuing the show and driving billions of revenue to FOX. How many of these kinds of mistakes have been made? It’s impossible to say.
Data-driven content production will come to dominate TVs and movies. We know the models work - we have seen it first hand on the web - and in order to survive, access to such data will be table stakes in the entertainment industry.
Like Netflix, the raft of startups is building channels on YouTube including Machinima, Maker Studios and many others leverage viewership data to tune and match their content to their viewers' desires. These channels represent the future of television - a future in which the demands of viewers are satisfied because the shows are tailored to them.
When interviewing product managers at Google, we ranked candidates on four metrics: technical ability, communication skills, intellect and Googliness. A Googley person embodies the values of the company - a willingness to help others, an upbeat attitude, a passion for the company, and the most important, humility.
In the past week, I asked two heads of engineering to identify the most important characteristic in new hires. Both responded, “humility”. For one startup ascertaining humility is so important, it is the first filter in the interview process.
Disruptive companies reinvent. They don’t copy and execute someone else’s playbook. To be disruptive, a startup’s team must cast aside preconceived notions and assumptions about doing things the “right way” and start inventing new ways.
The more time I spend in venture capital working with startups, the better I understand that there are no templates or stencils or best practices. Each startup team faces a unique market opportunity with distinct market dynamics, sales processes, competitive forces, assets and challenges.
In such circumstances, the best expeditionary force keeps open minds about the way forward. They learn from each other and the market. The first step to learning is accepting we don’t know everything.
Kenny Van Zant is a marketing wizard. Before his current role at Asana, Kenny managed products and marketing for Solarwinds, a publicly traded company that sells networking equipment to the mid-market. Solarwinds pioneered the low-friction, high-velocity sales model in their segment to great success.
SolarWinds offered free products to their customers to gain usage data that informs their sales and marketing efforts. As one might expect, inside sales reps would call upon the most likely customers to up-sell them to paid. And of course, SolarWinds employed the enterprise sales team to win the business of the largest customers. Plain vanilla freemium, right?
But over a conversation earlier this week, Kenny pointed out that freemium businesses switch the hunter and farmer sales team roles, an important twist that had escaped me.
In traditional businesses, the enterprise team hunts. They acquire leads, cold call, and try to push the customers to close. Their counterparts, the inside sales team, nurture and cultivate these relationships to prevent churn and grow revenue through up-sell and by growing with their customers.
Freemium businesses reverse the roles. The inside sales team fields the freemium leads, make the cold calls and push the customers to close. The enterprise sales team looks for multiple teams paying separately within one customer, build a relationship, and grow revenue through up-sell, helping their customers grow.
It’s plain to see the marketing departments of freemium companies must be constructed differently, with a strong quantitative bent. But freemium businesses also require a fundamentally different sales team structure.
In one hour, Kenny completely transformed my understanding of freemium software companies. I hope he decides to teach a course at Stanford one day his framework for success. It would be a huge a success and might very well change the trajectory of many SaaS companies.
David Brooks has a great op-ed this morning on the Philosophy of Data. He argues that data offers one major advantage and one major drawback. Data enables humans to discover patterns otherwise unobservable by our senses/intuition or patterns that violate human intuition. But the religion of data engenders a fallacy: that everything can and should be measured; and with this data, the best answer will emerge.
Belief in the power of data has become a sort of religious debate which has manifested itself in product design (data driven vs intuitive design), in politics, in baseball, in climate change debates and many others. At some level, it’s an extension of the religion/ science debate: what I feel/believe vs what my instruments indicate. Which is better?
Great innovations can come from both disciplines. Apple’s product design is built from intuition and belief - not from programmatic user testing. On the other hand, Google innovation in search focused exclusively using large scale multivariate testing and metrics optimization.
I believe the best innovations blend the two, mixing art and science. Great product ideas are borne of intuition but honed and fine tuned using data. In fact, when a company or product has reached has plateaued, when innovation has reached a local maximum, it’s often because one of these two disciplines has dominated thought. And it’s time to swing the pendulum back the other way.
Google’s recent focus on design and the reinvention of search’s user interface in Google Now exemplifies such a transition. A traditionally data driven company, Google has reinvented the feel of their brand and products using intuition.
In this debate as with most, zealotry will often command attention, but compromise will drive success. Both ways of thinking promise breakthroughs and innovation. Mastering both means balancing the two.
Venture capitalists are increasing market prices in Series A and Series B rounds aggressively in effort to reap disproportionate returns. And the variances in the prices of different startups in these early rounds is enormous, indicating a relatively inelastic market. Investors are chasing fast-growth startups irrespective of the price.
I gathered data on 77 rounds in the last 18 months by “high flying” companies using my own admittedly biased definition. Below is a table of the median, average and standard deviation of those Series A, B and C rounds.
Round
Median
Average
Std Dev
Count
Series A
31
65
135
46
Series B
120
289
638
21
Series C
212
318
333
10
I’ve also plotted the distributions below. Click to see a bigger image:
The Factors Behind the Trends
Returns in venture capital follow a power law - the top 1% of companies return the lion’s share of capital. As a result, venture capital demand to invest in high fliers is inelastic. Look at those standard deviations. They are massive!
Venture capitalists have raised growth funds to pursue companies generating $10M or more in revenue - Redpoint and many other firms have sought to invest in startups 24 months or so before an IPO, providing growth capital as startups remain private longer. The return characteristics on these funds enable them to pay higher valuations than traditional funds.
These growth funds are playing in all parts of the market including the Series A and Series B. GitHub, Lynda, and many other startups are bootstrapped and their Series A is a $35M+ round at a multi-hundred million dollar valuation, skewing the results.
Caveats
With a data set so small and subjectively gathered, there is bound to be bias and error. But I think the flawed data does support the anecdotal trend of rising valuations and enormous variances among top startups at all stages. For these startups, it’s a seller’s market.
Even the greatest minds fear missing out. Nobel laureate Richard Feynman who assisted in the development of the atomic bomb, contributed substantial advances to quantum mechanics and particle physics, discovered the cause of the Challenger Shuttle disaster and popularized science as a witty and successful author, faced this fear when confronted with a menu.
The number of dishes to try = √2(Meals remaining at restaurant+1) - 1
Fear of missing out is a paralyzing force. It even drives geniuses to mathematics for consolation. Having calculated the number of dishes to try, Feynman could rest, his mind at ease knowing that in all likelihood, he was eating the best plate on the menu.
With the panoply of options before us as founders, investors, managers and employees, the fear of missing out on key meetings, conferences, marketing initiatives, employment candidates, investment opportunities is rampant. There is always one more meeting to attend, one more person to meet, one more option to consider.
Within that last meeting, we seek assurance and validation that the choice we have made is the right one. But the byproduct of the relentless pursuit of the “best” can be debilitation. FOMO diffuses attention, sapping the focus which is often so necessary to success.
Feynman quelled his fears with probability. Most of us won’t approach problems with the same rigor. But all of us are seeking the same peace of mind.
I think it comes down to accepting that, as is written on Facebook’s walls, done is better than perfect. It’s more important to keep moving forward with a good decision than to slowly optimize for the best decision every time.
I don’t need a menu, thanks. I’ll have the spaghetti and meatballs.
The most important principle of start up fund raising is:
Raise enough money to achieve a set of milestones that will attract a subsequent round of investment from new investors.
Last week, a founder of a seed stage company came to pitch. When I told him the opportunity wasn’t a fit for us, he asked me what milestones he would need to achieve to raise a Series A - as he was raising a seed round! He was calibrating how much he needed to raise for in his seed to make sure he could raise a Series A.
After we spoke, he drafted a financial plan to calculate the size of the seed investment he would need to achieve those milestones and added a reasonable cash buffer. In effect, he reverse-engineered his Series A.
This is the same process investors follow when investing in a company. During the term sheet negotiations, VCs and founders discuss the size of the investment and the corresponding milestones the company will be able to accomplish with more or less cash. Like founders, investors are concerned about follow-on fund raising risk and reverse-engineer the next round to mitigate that risk.
The long term success of a venture-backed startup hinges on the ability of the founding team to raise follow on capital to finance research, fuel product development, discover product market fit and ultimately scale sales and marketing. To minimize financing risk, raise a seed that will enable you to raise an A. Raise an A to raise a Series B and so on.
Brokers serve two key roles within ecosystems. First, they introduce buyers and sellers. Second, they lend their expertise to help buyers and sellers make the right decisions in the market.
The internet neutralizes the first value proposition of brokers by leveling the information asymmetry between buyers and sellers at a far greater scale with much better data than any broker ever could. And for products and services with relatively small price points, commodity goods or where the cost of failure is low, market places often nullify the value (and cost) of an advisor.
Market places are arguably the best examples of this kind of disruption. Whether in ad tech (Google AdX), real estate (LoopNet, RedFin), accommodation (HomeAway, AirBnB), goods (eBay, etsy), classifieds (Craigslist), taxis (Uber, Hailo), travel (the GDS systems), financial services (AxialMarket, Angelist) market places are black holes within ecosystems.
Market places exert such gravity that once at scale, they are almost impossible to disrupt. Craigslist is the canonical example but all of the companies listed have constructed almost unassailable moats.
In real estate and financial services, where the value of the assets is measured in the millions or more and each asset is relatively unique, advice is valuable and customers will pay for it. So market places serving such verticals provide tools for brokers to succeed. Often, these tools provide both lead generation (customers) and management systems (CRM or deal tracking).
But as data volumes continue to grow exponentially, I believe that the business of brokers of almost every kind will be disrupted. And technology will become the ultimate market maker.
Once each month I met Peter at Café Habana in Nolita for huevos rancheros drenched in tomato sauce and a glass of fresh orange juice. Mopping up yolks with tortillas, Peter and I chatted about his business: the techniques of scalable customer acquisition, the priorities of the product and engineering team, the structure of sales quotas and the ebbing and flowing dynamics of the market place he and his team were building.
After two years, Peter called to say he was raising capital for his company; he invited us to take part in the process. I lept at the opportunity to work with Peter. I had witnessed how Peter managed adversity and uncertainty in his business and trusted his instincts and intuition. We had forged a working relationship over those breakfasts.
During the fund raising process, Peter encouraged me to spend time with his team. In parallel, the Redpoint deal team researched the market opportunity heavily. Ultimately, our investment memo totaled 40 pages, laden with interviews of market place participants and industry experts.
Then we began discussing terms. Peter negotiated terms aggressively and we shook hands on a partnership between AxialMarket and Redpoint in April 2012. In addition to having a special significance because of our friendship, AxialMarket was my first investment at Redpoint where I would be point and a member of the board.
A year into our partnership, Peter and I still see each other at least once each month and have the same kinds of conversations we always shared. Sometimes those conversations occur in the board room instead of Cafe Habana.
Reflecting on those two years, I admire the way Peter managed his fundraising process: building a relationship, evaluating our working styles and negotiating a fair set of terms. To be plain, those twenty-four breakfasts were quite possibly the longest investor interview process in history.
Peter understood from the very beginning that founders and investors work together for quite a while. The data confirms it: the median venture-backed company requires seven to ten years to reach an acquisition or IPO.
Most importantly, Peter showed me the value of building a long term relationship before seeking investment, the importance of a shared passion for the business, and the power of an alliance built on healthy, strong and honest relationships between investors and founders.
And to think, it all started at a little Cuban place in Nolita over some rancheros…
