Although we achieved a lot with Tutorspree, we failed to create a scalable business. I've been working through why. In doing so, I’m trying to avoid the sort of hugely broad pronouncements I often see creep into post mortems that I’ve read: “don’t hire people!”; “hire people faster!”; “focus on marketing at all costs”; “ignore marketing, focus on product” etc.

I’ve focused here on the strategic causes of our failure. While I learned a huge amount about operations, managing, and team building; mistakes made in those areas were not the ultimate cause of failure just as the many things we got right within the company did not ultimately lead to success. I also recognize that this doesn’t cover every detail, even on the strategy side.

SEO: Too good to be true

Tutorspree didn’t scale because we were single channel dependent and that channel shifted on us radically and suddenly. SEO was baked into our model from the start, and it became increasingly important to the business as we grew and evolved. In our early days, and during Y Combinator, we didn’t have money to spend on acquisition. SEO was free so we focused on it and got good at it.

That worked brilliantly for us. We acquired users for practically nothing by using the content and site structure generated as a byproduct of our tutor acquisition. However, that success was also a trap. It convinced us that there had to be another channel that would perform for us at the level of SEO.

In our first year, that conviction drove our experiments with a series of other channels: PPC, partnerships, deals, guerilla type tactics, targeted mailings, craigslist posting tools, etc. Each experiment produced results inferior to those from SEO. The acquisition costs through those channels were significantly higher than what was allowable based on our revenue per customers. We also found that potential customers coming through PPC were converted at a lower rate than those originating through SEO. Even as we sharpened our targeting, experimented with messaging, and sought advice and consulting from more experienced parties, we found that paid channels just weren’t good enough to merit real focus.

That dynamic put us in a strange position. On the one hand we had a channel bringing in profitable customers. On the other hand, we did not have the budget within our model and product to push hard enough on other channels.

The AirBnb Head Fake

At the end of our first year, the divergence between our success with SEO and our failure with other channels dovetailed with a whole set of lessons we drew from analyzing user behavior on Tutorspree. We realized that there were fundamental problems with the product of Tutorspree which both prevented us from converting visitors to customers at optimal rates and from having enough capital to spend on acquiring more visitors/potential customers.

We had modeled ourselves on AirBnB, believing we were a clear parallel of their model for the tutoring market. What we were seeing in terms of user behavior, however, was fundamentally different. Parents simply didn’t trust profiles and a messaging system enough to transact at the rate we needed. Our dropoff was too high, and the number of lessons being completed was too low. We realized that we were wrong in how we thought about the entire market, and radically altered our model to suit in March of 2012. Looking back, it is also apparent that we were able to ignore our error for as long as we did precisely because SEO worked as well as it did. The dynamics of our marketing provided air cover for any other issues we had.

We called the new model Agency as we pulled in aspects of a traditional agency’s hands on approach and combined it with our matching system and our customer acquisition channel – SEO. Within a month of the change, we doubled revenue. Six months after the shift, our revenue had increased another 3x and we’d increased margins from 15% to 40%. The new model gave us our first profitable month, and put us within striking distance of consistent profitability. It looked like we had cracked the product problem. Our conversion rates were way up and per user revenue was climbing rapidly. Those factors gave us the budget we needed to more productively experiment with other channels.

Virtually all of our customers came from SEO.

New and Better Model; Same Old Channel

By December of 2012, we had virtually infinite runway and were at the edge of profitability. We still wanted to swing for the fences, and, given the radically shifted economics presented by our new model, we made the decision to retest all the marketing channels we had tried with our initial model and then some. We knew that only having a single scaling channel – SEO – would not let us become huge, so we began pushing for another scalable channel.

Given the strength of where we were and the challenges we saw, we raised another round with the explicit purpose of finding the right marketing channels. While we considered raising an A, we played conservatively, deciding that we wanted to find the repeatable channels, then raise an A to push them hard rather than raise too much money too early.

We finished that fundraise in January, began a much needed redesign of the site to fit with our significantly more high touch model, hired a full time growth lead and began to push rapidly into content marketing, partnerships. Then, in March of 2013, Google cut the ground out from under us and reduced our traffic by 80% overnight. Though we could not be 100% certain, the timing strongly indicated that we had been caught in the latest Panda algorithm update.

With our SEO gone, we took a hard look at our other channels. While content may have played out in the long run, and in fact showed signs of the beginning of a true audience, the runway it needed was far too long without the cushion provided by SEO. PPC - mainly through Adwords (though also through FB) – was moving in the direction of being ROI positive, but the primary issue turned out to be one of volume rather than cost. Because of our desire to focus heavily on the markets in which we had the highest tutor density (and therefore the greatest chance of filling requests), we had to carefully target our ads in terms of geography and subject. Given that dynamic, we simply couldn’t find a way to generate enough leads, no matter the price. In the end, that calculus applied to nearly every paid channel we could identify.

Common Thread

Our reliance on SEO influenced nearly every decision we made with Tutorspree. At the beginning, it influenced our decisions to allow tutors to sign up anywhere, for almost any subject. On the one hand, that brought in leads we could never have specifically targeted. On the other hand, it spread our resources out across too many verticals/locations. That problem was compounded by our move into Agency. While we were converting at a higher rate and price than ever, we were also forced to spend too much time and money on completely unlikely leads. When you build your brand on incredible service, it becomes very hard to simply ignore people.

When our SEO collapsed, we routed virtually all of our technical resources to fixing it. In that effort, we had significant amounts of success. We regained most of the traffic that we lost with the algorithm switch in June. We regained a significant portion of our rankings. However, the traffic that we were getting at that point was not as high quality as that which we had been getting beforehand.[1]

Because of how successful SEO was, it was the lens through which we viewed all other marketing efforts, and masked the issues we were having in other channels along with important realities of how the tutoring market differed from how we wanted to make it work. We were, in effect, blinded by our own success in organic search. Even though we saw the blindness, we couldn’t work around it.

Lessons Learned

Tutorspree taught me a lot of lessons. I learned about product, users, customers, hiring, fundraising, managing, and firing. I made some bad hires because of my own blind spots and desire to believe in how people operate. There were periods of time where I avoided conflicts within our team too much – decisions that were always the wrong ones for the business and that I regretted later. Those mistakes were not ultimately what caused our failure.

Nor is the largest lesson for me that SEO shouldn’t be part of a startup’s marketing kit. It should be there, but it has to be just one of many tools. SEO cannot be the only channel a company has, nor can any other single channel serve that purpose. There is a chance that a single channel can grow a company very quickly to a very large size, but the risks involved in that single channel are large and grow in tandem with the company.[2] 

That’s especially true when the channel is owned by a specific, profit seeking, entity. Almost inevitably, that company will move to compete with you or make what you are doing significantly more expensive, something Yelp gets at well in their 10K risks section: “We rely on traffic to our website from search engines like Google, Bing and Yahoo!, some of which offer products and services that compete directly with our solutions. If our website fails to rank prominently in unpaid search results, traffic to our website could decline and our business would be adversely affected.”

For me, this is a lesson about concentration risk and control. In this case, it played out on the surface in our only truly successful marketing channel. That success wound its way through everything we did, pulling all that we did onto a single pillar that we could not control.

By necessity we had to concentrate risk on certain decisions (something likely true of most small startups). I did not have the time or resources to do everything I wanted or needed to do. I never will. But I need to be cognizant of the ways in which that concentration is influencing everything I do. I need to make sure that it doesn’t dig me into holes I can’t work out of on my own.

Ultimately, this post mortem is about the single largest cause I can identify of why we failed to scale Tutorspree. In examining our SEO dependence, I was surprised at how deeply it influenced so many different pieces of the company and aspects of our strategy. It powered a huge piece of our success, and ultimately triggered our failure. There’s a symmetry there that I can’t help but appreciate, even though I wish to hell it had been otherwise.

__

[1] This is a whole other issue I explored in the Tutorspree blog at the time. It seems that Google is increasingly favoring itself in local transactional search.

[2] RapGenius recently ran into this issue, but were able to overcome their immediate problems through some impressively fast and thorough work.

Why convertible notes exist

Convertible notes beat equity as the financing of choice for early stage startups largely because they were faster. While later stage companies can afford the time necessary to negotiate long sets of terms, early stage companies don't have that luxury. Not only are prices for such early stage entities nearly impossible (and therefore time consuming) to determine, time is itself the most limited resource a startup has - it's critical to raise quickly and get back to work. Though equity financings can be standardized up to a point, there's always a huge sticking point for negotiations: price. Rather than try standardizing, convertibles notes simply got rid of price as a term.

While the notes eliminated price, they needed structures to govern what the investment would be worth when a price finally did come into play. Price was replaced with two other mechanisms - a discount rate and a cap.[1] While each of these was a negotiable point, they were each and cumulatively less meaningful than price, because they deferred the core pricing question to the next round of financing. Ideally, that meant shorter negotiations on less significant terms. With a priced financing meant to take place within the year (based on the note's maturity), investors were willing to punt on pricing while founders were willing to roll their risk a bit further down the road and get back to work.

The cap trap

But a funny thing happened on the way to this ideal state - caps became implied prices. It isn't all that surprising that this happened. The best founders know that setting goals based on real numbers is critical to sustained growth. They're also incredibly competitive. By removing real prices from fundraising, the convertible note nullified a part of the fundraising game that was important to the egos of founders (and investors). Without a number on which to hang their success, startups lost a public way to "prove" how good they were. At that point, the cap became an obvious choice for an approximation of price: It was a hard number, it could/should reflect something about the note holder's expectations of the price at which the company would next raise money, and it is typically not that well understood.[2]

This introduced several unintended problems. The most extreme of these issues was the advent of the "uncapped" note. To some founders, this was the holy grail. In a very naive sense, this implied an infinite valuation - companies were so hot that investors were saying they'd accept any price at some point in the future just to get in. Practically speaking, this scenario - paired with not having a discount - creates a misalignment of incentives between the investor and the startup. Rather than trying to help the startup move as fast and far as possible prior to the next round of financing, investors are actually incentivized to get as low a price on the first equity as possible so that they are able to convert in at a reasonable price. Conversely, the startup wants to move as fast as possible - as all startups should - and give up as little as the company as possible. While that tension always exists on some level, the dynamic is particularly extreme in these cases.

The second problem came from startups who raised too little money with too high a cap. Convertible notes are debt. They accrue interest and have maturity dates. As a result, it tends to be a bad idea to raise too much money through them. While investors rarely call the notes, the accrued interest can become a significant dilutive factor. But, since founders decided that high caps = good caps, startups would do crazy things like raise $1mm on a $20mm cap with one year maturity.[3] The hurdles set for a company in that scenario were almost unimaginable - and contradictory! Inside of a year, that company would have to grow to a point where raising money at above a $20mm valuation was not just possible, but likely. Otherwise, again because of the common misperception that cap = price, they'd have to raise a "down round". However, if by some miracle they raised above the cap, the investors would get the same % they initially agreed to, but would also get higher liquidation preference.[4]

Ironically, while convertible notes were initially designed to allow good companies to raise money quickly and get back to work, the best companies are precisely the ones that most frequently fell into these cap traps.

Being smarter when raising money

Recently, Y Combinator released a new financing instrument, called the safe, designed to eliminate parts of these problems. By removing interest and maturity dates, the safe makes it easier for companies to raise larger amounts than on notes while avoiding putting themselves in impossible situations. However, the ultimate responsibility for restraining the temptation to raise at ever higher caps rests with founders. Understanding how and why convertible notes and safe exist is the first step. That's easy, it's just reading. The second step is much harder, but equally worth it - founders need to overcome their egos. That might be impossible, but it's well worth the effort.

__

[1] Notes also have elements of debt: interest rates and maturity.

[2] I've had conversations with both founders and investors in which it has been apparent that this is one of several terms that are commonly misconstrued. Those misunderstandings lead to different actors using caps as placeholders for different things, which complicates the picture.

[3] I'm being extreme, but have heard of crazier.

[4] David Hornik explores this more in his post Just Say No To Capped Notes. He rightly points out that the capped note is an at times awkward compromise, but I disagree that the resolution should be all uncapped notes or all equity financings. 

A friend of mine was a professional poker player for years. One of the more interesting things he taught me was that he could predict his earnings when playing online poker. I had assumed, based on my own inexpert poker playing, that, while there were knowable odds that would improve your chances of winning, you always had to deal with other people as an unpredictable variable. That's why you ended up with big pots that could break players - they judged the odds wrong and lost.

As I thought about it, I realized that those scenarios were not mutually exclusive. You could play thousands of games for small stakes and edge up by predictable, if limited, amounts. Part of the positive expected return came from the basic odds of poker, and some came from playing against people who simply didn't understand the odds as well. At the other extreme, you could play single huge games where the variance of outcomes could empty your bank account or massively increase it on a single hand.

The parallel to startups wasn't immediately obvious to me. Risk based scenarios can all be modeled given certain assumptions. If an investor could model those variables accurately, they would know exactly how much return they would expect out of investing x dollars over y companies. If they had enough of the right variables predicted, they could just bet the winners - imagine an investor so good they could divine the information needed to only invest in Google, Facebook, Twitter, LinkedIn. Given how valuable solving that risk equation is, investors are obviously going to do everything they can to solve for those variables.

The dynamic that starts playing out, then, is that you have different investors playing different games, even though they think they're playing the same one. If you start with the assumption that startups are extremely likely to fail [1], then investing with no further information is a bit like roulette. Place your bet, and maybe you get massively lucky - but that's all it is. So the investors seek information to change the game into something closer to poker, where their odds rise dramatically and quantifiably. Meeting the founders, validating the market, testing the product, seeing traction - each of these edge the game closer to one with a positive expected outcome. The best investors find ways to change the game before anyone else does, reaching decisions faster and more accurately (professional players among amateurs). The worst investors don't even realize what's wrong with playing roulette.[2]

Founders looking to raise money can use this dynamic to their advantage. The simplest read on that advantage is that founders should do that by being tricky with the information they release and to whom they release it. By doing that, you could keep all the investors off balance and hyper competitive with one another. That should theoretically drive your value up. This seems clever, and clever seems good. But that's actually a bad idea for a few reasons. The first is that investors talk to one another and so your information will get around anyway, but it will happen out of your control and will probably reflect badly on you.[3] The second, and more important one, is that if you are tricky with releasing information, the best investors, the poker players, won't feel comfortable with you. You'll restrict your potential pool of investors to the roulette players, which tend to be the least good investors.

It's probably better to treat your interested investors fairly similarly in terms of what you tell them. The goal is to build a company so good that the information you release forces investors to rush to a decision they feel good about before anyone else does. You want to help the good investors get past the roulette stage so that they'll make an offer they believe will help you make them a lot of money.

If you do that right, you'll have good, smart, investors who are aligned with you and your interests. If you only have the roulette players, you probably did it wrong.

___

[1] I've seen this number pegged as high as 99%, but who trusts statistics?

[2] Of course, since they're playing a game of chance, they might actually get lucky and win big. Some will admit they got lucky, most will call it skill and inspire other similarly foolish people to do the same.

[3] It also starts to get very difficult to keep straight what information you gave to who, and your energy should be spent on building a company, not managing a web of information dispersal.