Monthly Archives: October 2010

why netflix’s recommendations are more interesting than amazon’s

I have always found Netflix’s recommendations to be more interesting than Amazon’s.  When you buy a Bruce Willis movie on Amazon, they often recommend other Bruce Willis movies.  Netflix, on the other hand, is much more likely to suggest movies I wouldn’t have thought of or even heard of.  Why is this?  

Generally recommendation systems can be tuned to simply please the user (emphasize overall popular items) or to wow users (emphasize less popular items).  

I think the difference is due not to difference in expertise but difference in goals.  Amazon sells products and Netflix rents them (at least in the case of physical DVDs).  Amazon wants an algorithm that simply optimizes sales, and showing a popular item is more likely to lead to a sale.

Netflix has a different challenge.  They have an inventory of DVDs.  If everyone is renting the most popular movies, they have to stock tons of those movies, and the “long tail” movies sit idle in their warehouse.  They want an algorithm that gets people to go down the tail and watch movies that would otherwise go unwatched and therefore keeping more of their inventory in circulation.  Hence the algorithm is tuned to “going out on a limb,” making the recommendations more interesting.


Why don’t you guys [Hunch] go after a single problem and solve it instead of trying to be a generalized product?

A question we got about Hunch from an anonymous user on Formspring:

Question: It seems to me Hunch is a solution in search of a problem. It’s like to Why don’t you guys go after a single problem and solve it instead of trying to be a generalized product? It seems like you’re making such an obvious mistake.

My response: Great question.  

1) First, I don’t think Hunch is a solution in search of a problem.  We’ve talked to probably a hundred large websites about partnering and have found that personalization/recommendations is at the top of the priority list for almost all of them.  It has surprised us, for example, how many large sites are either a) building significant internal operations to add personalization/recommendations to their sites or b) have recently hired senior executives to be “head of personalization.”  For better or worse, our biggest problem hasn’t been what lots of startups face – indifference – but instead people seem to see personalization/recommendations as so important that they want to do it themselves.  So our take is that what we are focused on is going to be a central theme in technology over the next few years and we are very well positioned having been working on the problem for almost 3 years.

Meanwhile, in the search space, technology like what we’ve built at Hunch is becoming increasingly important.  Eric Schmidt (CEO of Google) recently said:

Ultimately, search is not just the web but literally all of your information – your email, the things you care about, with your permission – this is personal search, for you and only for you. The next step of search is doing this automatically. When I walk down the street, I want my smartphone to be doing searches constantly – ‘did you know?’, ‘did you know?’, ‘did you know?’, ‘did you know?’.This notion of autonomous search – to tell me things I didn’t know but am probably interested in, is the next great stage – in my view – of search.

So in addition to all the opportunity for Hunch to partner with commerce, media, local etc sites, we see what we are doing as highly relevant to the most important market on the web – search.

2) Your question about focus is a good one.  I don’t know the details of the history of ThePoint/Groupon, but I my guess what looks obvious in hindsight actually required a lot of experimentation. Experimentation does not mean lack of focus – it can be part of the process of testing various markets to find the best product-market fit. We see ourselves at the stage where we are pursuing about 3 promising angles and then once we figure out which one is most promising we will focus much more on that angle.  One of the great challenges of having a broadly useful technology is building enough functionality in enough areas to find the best area without spreading yourself to thin – something we wrestle with every day.




A few comments on Paul Graham re superangels

Overall, the best written and most accurate description of recent
changes to early stage financing environment I’ve seen. A must read
for founders and investors.

A few comments:

– “But it was mysterious to me that the super-angels would quibble
about valuations. Did they not understand that the big returns come
from a few big successes, and that it therefore mattered far more
which startups you picked than how much you paid for them?”

Totally agree: as a seed investor I generally follow the Ron Conway
school of thought that outcomes are binary and so angels generally
should not be price sensitive. That said, in some cases (especially
non-hot markets with no VCs co-investing) a key risk for seed funded
companies is financing risk, specifically getting VCs to follow on to
the next round (at a higher price hopefully – down rounds are
psychologically devastating). I’ve seen VCs summarily pass on deals
when angel round valuation was too high. Keeping the seed valuation
reasonable lowers financing risk.

– “Because super-angels make more investments per partner, they have
less partner per investment. They can’t pay as much attention to you
as a VC on your board could.”

True, but I also would argue that practicing entrepreneurs can help
early-stage startups in more scalable ways since they are current on
lots of startup issues/people/vendors/partners etc that VCs aren’t.

– “Who will win, the super-angels or the VCs? I think the answer to
that is, some of each. They’ll each become more like one another. The
super-angels will start to invest larger amounts, and the VCs will
gradually figure out ways to make more, smaller investments faster.”

I hope not. It will require discipline on the part of successful
super angels to keep their funds from growing. If Founder Collective
is successful, I expect we will disciplined and continue focusing on
seed investments (and not raise bigger and bigger funds as seems to be
the historical pattern).

– “The seriousness of signalling risk depends on how far along you
are. If by the next time you need to raise money, you have graphs
showing rising revenue or traffic month after month, you don’t have to
worry about any signals your existing investors are sending. Your
results will speak for themselves. Whereas if the next time you need
to raise money you won’t yet have concrete results, you may need to
think more about the message your investors might send if they don’t
invest more. I’m not sure yet how much you have to worry, because this
whole phenomenon of VCs doing angel investments is so new. But my
instincts tell me you don’t have to worry much. Signalling risk smells
like one of those things founders worry about that’s not a real

1) In my experience a significant majority of companies are in that
grey area without super strong concrete results.
2) I have seen negative VC signaling play out a number of times. I
guess I just disagree here but as Paul says this is all so new that
time will tell. Particularly interesting will be to see how rounds
with multiple VCs in seed deals play out. Could be good or bad for
founders – I’m really not sure.

– “The best thing for founders, if they can get it, is a convertible
note with no valuation cap at all. In that case the money invested in
the angel round just converts into stock at the valuation of the next
round, no matter how large. Angels and super-angels tend not to like
uncapped notes. They have no idea how much of the company they’re

As I see it, the main problem with uncapped notes is that seed
investors have perverse incentives – they want to see the company
succeed but are economically rewarded for a lower valuation, hence
economically (I stress “economically” to distinguish it from moral
imperatives) have no incentive to help the founders raise a high
valuation VC round.

Thoughts on incumbents from a startup’s perspective

– By “incumbents” I mean the big companies that are loosely competitive to your startup.

– The first thing to do is try to understand the incumbent’s strategy.  For example, see my analysis of Google’s strategy. 

– Being on an incumbent’s strategic roadmap is a double-edged sword.  On the one hand, they might copy what you build or acquire a competitor.  On the other hand, if you build a valuable asset you could sell your company the acquirer at a “strategic premium.”

– Incumbents that don’t yet have a successful business model (e.g. Twitter) might think they have a strategy, but expect it to change as they figure out their business model.  An incumbent without a successful business model is like a drunk person firing an Uzi around the room.  

– Understand the incumbent’s acquisition philosophy. More mature companies like Cisco barely try to do R&D and are happy to acquire startups at high prices.  Incumbents that are immature like Facebook only do “talent acquisitions” which are generally bad outcomes for VC-backed startups (but good for bootstrapped or lightly funded startups). Google is semi-mature, and does a combination of talent and strategic acquisitions.

– Understand the incumbent’s partnership philosophy.  Yahoo and Microsoft are currently very open to partnerships with startups.  Google and Facebook like to either acquire or build internally. If you don’t intend to sell your company, don’t talk seriously about partnerships to incumbents that don’t seriously consider them.

– Every incumbent has M&A people who spend a lot of their time collecting market intelligence. Just because they call you and hint at acquisition doesn’t mean they want to buy you – they are likely just fishing for info. If they really want to buy you, they will aggressively pursue you and make an offer.  As VCs like to say, startups are bought, not sold.

– Try to focus on features/technologies that the incumbents aren’t good at.  Facebook is good at social and social-related (hard-core) technology.  Thus far they’ve kept their features at the “utility level” an haven’t built non-utility features (e.g. games, virtual goods, game mechanics).  Google thus far has been weak at social and Apple has been weak at web services.

– Try to focus on business arrangements that the incumbents aren’t good at.  Facebook and Google only do outbound deals with large companies.  With small companies (e.g. local venues, small publishers) they try to generate business via inbound/self service. Building business relationships that the incumbents don’t have can be a very valuable asset.

– Be careful building on platforms where the incumbent has demonstrated an inconsistent attitude toward developers. Apple rejects apps somewhat arbitrarily and takes a healthy share of revenues, but is generally consistent with app developers.  You can pretty safely predict what they will will allow to flourish. Twitter has been wildly inconsistent and shouldn’t be trusted as a platform.  Facebook has been mostly consistent although recently changed the rules on companies like Zynga with their new payment platform (that said, they generally seem to understand the importance of partners thriving and seem to encourage it).

– Take advantage of incumbents’ entrenched marketing positioning.  The masses think of Twitter as a place to share trivial things like what you had for lunch (even if most power users don’t use it this way) and Facebook as a place to talk to friends.  They are probably stuck with this positioning.  Normals generally think of each website as having one primary use case so if you can carve out a new use case you can distinguish yourself. 

– Consider the judo strategy.  When pushed, don’t push back.  When Facebook adds features like check-ins, groups, or likes, consider interoperating with those features and building layers on top of them.