This presenta,on will be delivered in two parts. First I will discuss my thoughts on
“ideas” and share three of my favorite ideas with you. I will then share 7
recommenda,ons with you about how to take your idea from start to ﬁnish.
Raise you hand if you have children. Keep it up if you have ever read Runny Babbit.
I will give you a minute to read this excerpt.
Runny Babbit is a book full of poems with leGers moved around. Because our brains
recognize paGerns, it is hard for adults to read the book. This is the only book where
my 6 year old daughter can read faster than I. She sees the words as they are printed.
I see paGerns from past experience and have to ﬁght my ins,ncts. She loves it.
This paGern recogni,on usually serves us well. But past paGerns and data aren’t
always right, as Richard Feyman says here.
Every morning I come up with a new business idea. This prac,ce has made me a
more crea,ve product professional. They key star,ng point for me is to look for the
gap between conven,onal wisdom and facts as I observe them.
When you look in the mirror and share your new idea with yourself, you see Albert
Einstein. But everyone else sees her.
Does this sound familiar? “If it’s such a good idea, why hasn’t it been done?” “If it
works, [insert big company name] will just do it and kill you.” The good news is that
people aren’t likely to steal your ideas – most people will think that you are crazy.
Share your ideas. Doing so will make you feel like you need to go do them, because
of the small risk that someone will take your idea now that it’s “out there” and beat
you to it. Sharing your idea will expose you to diverse feedback on it. Your idea will
get pressure tested. You should iterate based on that feedback, staying true to the
core idea but improving it using the power of logic.
There is one key advantage of secrecy, and that’s marke,ng. Le[ng candidates in on
the secret can build up excitement. Giving the press an exclusive can get you good
marke,ng. But for most of you, le[ng openness win in the baGle with secrecy is the
Please, keep this between us and sign the NDA on your way out ;‐‐)
Ads: This is an adver,sement I received on CNN.com this Wednesday when I was
pu[ng this presenta,on together. I was so conﬁdent that any display ad would be
terrible, that I simply went to the ﬁrst site that came to mind and took a snapshot of
the ﬁrst ad that was delivered to me. Terrible. Behavioral targe,ng models can be
leveraged to improve targe,ng. But the models are ofen very high‐level and take
,me to build. Imagine if instead of random ads or even behavioral ads in big buckets,
that any developer with an algorithm or heuris,c could improve the targe,ng
I’m not looking for a job. But if I was, I wouldn’t use any of the exis,ng job sites (with
the excep,on of LinkedIn). I tried to look for a job on Monster at a startup in Silicon
Valley. I’m probably not the “target” market for Monster, but why not? And do you
really believe that this is the best we can do?
Browsing what is going on in the news is something that most consumers do most
days. Yahoo! solves the relevancy problem in news by leveraging feeds, technology,
and editors. Google tried to solve the problem with scraping and algorithms– and the
product is not even close in quality. Digg solved the problem by making the
community editors – it’s a very strong broadcast solu,on, but I’m convinced that
opening the core algorithm development in addi,on to the editorial func,on to third
par,es is where things will get really interes,ng.
Facebook realized that they could not do everything. They realized that they couldn’t
scale business development – that they would do some deals that they shouldn’t and
not do deals that they should. So they built a self‐service applica,on development
plamorm and changed the game (and the trajectory of their company) by opening up.
In most startups, there is some key IP that is usually delivered through a set of
algorithms. Yet, most of the eﬀort of a startup is focused on everything but the
algorithm. Take a 20 person company and you probably have 1 or 2 people working
on the core algorithm. Innocen,ve spun out of a pharmaceu,cal ﬁrm that wanted to
tap into (now 155,000) scien,sts around the world. Chemists can apply knowledge
about they have learned in their craf to help other scien,sts. Mathema,cians can
help companies solve energy problems. There is a huge amount of untapped
knowledge in great minds around the planet. Let’s tap that knowledge.
So take that Innocen,ve model, but make it self‐service. And make it objec,ve.
Focus only on algorithms. Build a A/B tes,ng plamorm to ﬁnd the winners. Allow the
atomiza,on of your site so that algorithms can be delivered in a very granular way
(e.g., open up ad targe,ng and let anyone build their own behavioral targe,ng
model). Or use this approach to build a web‐based email product that is spam‐proof.
How about a news site that like Google News, but which delivers an order of
magnitude improvement in relevance.
Visit a da,ng site, like Match.com and try to ﬁnd a good match. You need to add all
sorts of informa,on into the system and then you wade through “results” to ﬁnd a
match. If Match had more informa,on about me that was passively collected, they
could make it an order of magnitude easier to par,cipate in their system. The results
would also likely be much beGer.
I am an Apple fana,c. I worked as an Apple Student Representa,ve in college and
have fought the good ﬁght in several companies against the Windows‐IT guy complex.
But despite the fact that my purchases from Apple are all direct – either online or at
their store, I get a generic experience at Apple.com. Despite the fact that I have
registered at Apple.com and use iTunes and my iPod every single day – Apple oﬀers
me a “broadcast” experience. Apple knows that I worked at Yahoo! and that I’m not a
student, yet they marke a special student oﬀer to me that I pursue. They know that I
own Leopard as I just bought a new Macbook Pro. I don’t have an iPhone yet. I
would love to ﬁnd out more about cool third party sofware for the Mac.
Most sites ask for too much informa,on. They collect too much and hold it for too
long. And then they oﬀer me a broadcast experience in return? We can do beGer
Amazon does a very good job of merchandising to me. In fact, Amazon is the only
service I have found on the web that comes close to delivering on the promise of
giving me value in return for me providing informa,on about myself and my interests.
Why do sites require registra,on? 1. to authen,cate that you are a human, not a bot
(catcha, email conﬁrma,on, etc), and; 2. to collect data about me to garget content,
oﬀers, and marke,ng. Most sites on the web does a terrible job at both.
Personalizer will bring the personaliza,on func,onality of Amazon to the en,re web
with NO REGISTRATION requirements. We will do this by using your data which has
been passively and ac,vely collected – which you have full transparency into – and
leveraging the fact that compute cycles and storage costs are approaching zero.
We could also have quality of service levels triggered by authen,ca,on. For example,
we could have a base level registra,on which would use captcha and email
conﬁrma,on and a second level that ads SMS conﬁrma,on. We could add an
electronic two factor authen,ca,on token, which would remove phishing aGacks.
Taken together, you would get a massively personal and safe web experience with NO
A few years ago I bought a Linksys 54g wireless expander. Se[ng it up from my
Macintosh was the single most frustra,ng consumer experience I have ever had. First
I went to Google to ﬁnd an answer
I was taken to CNET, where my ques,on was not answered on the message boards.
At least I couldn’t ﬁnd the answer.
Finally I went the the Linksys web site. Search was broken. It was impossible to ﬁnd
the answer. Afer hacking around, I ﬁnally got it working. It was a several day ordeal.
Why do products work like this? When they do, why can’t I ﬁnd assistance?
Most products aren’t user friendly. Unfortunately, that’s probably not going to
change. And you cannot change every product in the world. But you can ﬁx the
support informa,on problem…
Build a giant product taxonomy that includes not only all products from today, but
throughout ,me. Scrape all support content available on the web, and use cheap
labor and a community to map it to your product taxonomy. Hire a large staﬀ in India
and oﬀer local snail mail addresses to people so that they can send all of their
manuals in to be scanned and placed for all to see on the web.
Ideas are worthless without execu,on. Great ideas are valuable. Coming up with
ideas is fun. Using the power of your idea to recruit a team and raise money is
inspiring. But if you can’t execute on your idea, it’s worth zero.
I will now share 7 sugges,ons with you about how to make one of these, or any
other, idea a reality.
Any large market can be re‐invented. There is always room for improvement.
However, innova,on may fail if you are inven,ng a new market. While it’s possible to
invent a massive new market at the same ,me as building the market leader in that
market, it’s a low probability bet. I would strongly encourage you to look at big
markets and ﬁnd customer segments that just aren’t being well served. Delight users
by showing them how the world should work. Leave new market crea,on for basic
Startups create something out of nothing. While you clearly need to aGack a big
market with a great idea and good people, the winner in your market will likely be the
one with the best leader. A startup CEO needs to convince people to make
“irra,onal” bets with their own lives. He needs to convince customers to be early
adopters. He usually needs to raise capital and manage risk averse investors. A great
group of people with a weak CEO will lose. A strong CEO has a strong grasp of
strategy, can sell anything to anyone, and can inspire people to do what they never
realized they could achieve.
Most great technology companies are lead by great founding CEOs – Steve Jobs at
Apple, Larry Ellison at Oracle, Bill Gates at Microsof, Jerry Yang at Yahoo!, Larry &
Sergei at Google, Mark Zuckerberg at Facebook. The list goes on and on. Work hard
to have the right CEO on the founding team.
Take risks with your science, not with HR. In big companies, managers are
encouraged to “ﬁll reqs” before they go away. That type of behavioral will absolutely
kill a startup (and I would argue kills big companies, too). A bad hire is more painful
the smaller you are. You don’t have ,me to deal with moving people out. While it
may feel like a tradeoﬀ between speed and quality, put all of your eggs in the quality
bucket. And never hire ahead of your needs – you business is moving quickly. If you
hire the right person into an unimportant role, they will be de‐mo,vated and bring
Once you launch a product to the world, it’s hard to un‐ring the bell. Conven,onal
wisdom says that you should launch to the world as quickly as possible, because that
is the only way you can learn. But what if you learn that your hypothesis is wrong?
What do you do?
Get your product to a state where you believe that it’s at the minimum level required
to see if your hypothesis if right. Launch it to a small group (100?) of friends and
family and let them know that you need their help to test your product. Whatever
they do will likely be lost. Giving your product team the ﬂexibility to toss everything
and start over, or even giving engineering the ﬂexibility to build the next version
without having to be backward compa,ble with the last version if a massive
advantage. Do this over and over un,l you feel like you’re ready. Then launch the
product to the world and understand that you need to support your customers as you
evolve your product.
Big companies wait to long to launch, launch to a huge audience, and then give up
before the product team can learn from the experience and iterate. This is
par,cularly true with community products which take ,me to evolve. Startups launch
to the world too quickly and then lose some of their natural advantage of ﬂexibility by
inheri,ng backward compa,bility.
Sending people to a site which isn’t viral and which is losing money on every visit is a
bad business model. PR is great for BD, sales, and recrui,ng but it will not move the
needle short term in terms of traﬃc.
Have a clear model of what ac,ons you want and what value those drive. Model
them out and measure your marke,ng eﬀorts against them. Marke,ng is ofen far
too much art and far too liGle science. When you are ready, ﬁnd a marke,ng partner
who will work with you on a CPA basis so that you get a high probability ROI.
I have watched entrepreneurs spend YEARS looking for the perfect co‐founder before
star,ng a company. They reference them, they test them, they spend weeks or
months with them socially as well as working on the idea together. Then they raise
money afer a few mee,ngs and take the oﬀer with the best economics or with a
name brand ﬁrm.
If you take money from investors, they will likely have an equity take that is similar to
that of a founder. Apply the same ﬁlter on your investors as you do on your founders.
If you have trac,on, you have a good deal of leverage. Even if you don’t have
trac,on, you are beGer oﬀ boot strapping than going into a long‐term business
partnership with the wrong person or ﬁrm.
Good investors can help you create a great outcome. Bad ones can make your life a
Everyone has heard some second hand data that 1 in x startups succeed while the
rest fail. Yet, certain venture capitalists either pick or help make companies win at a
very high rate. Why is that? I don’t believe that startups are as risky as people fear.
Many employees can make close to the same salary as they do at a big company, but
the upside on the equity is massive. The real cost of a startup is not career risk, but
rather the amount of ,me and psychic energy they consume. Startups are a big
commitment. But there is nothing in the world like taking an idea from start to ﬁnish.
The two best jobs I ever had were Epinions and Bix. I hope everyone here has the
opportunity to be as happy in their work as I have been in mine.