Friday, July 21, 2006

Starting Findory: On the cheap

When I started Findory, I was extremely sensitive to burn, probably too sensitive.

Maybe having witnessed all the dot com flameouts caused me to be overly cautious, but I wanted to be very careful about money flowing out the door. Many companies do fail early simply by running out of cash. I thought the best way to avoid this was to be careful about spending money.

Findory was built on the cheap. It is self-funded. There is no office space; the company is virtual. There is almost no marketing or advertising.

There are no salaries or benefits at Findory. For most tech startups, salaries dominate the burn rate. Even at 1/3 salary, each additional employee is about $4-6k/month (fully loaded, includes benefits, expenses, office space, etc.) At full salary, it is about $10-15k/month. Salary and benefits would have burnt through initial funding in less than a year.

By the way, it is worth noting that not drawing a salary is roughly equivalent to taking a salary but investing an equivalent amount immediately back in the startup. The latter is a lot more complicated, but both serve to fund the company using the money normally paid in salaries and benefits.

Findory uses free open source software (Apache, MySql, Linux, Perl, Berkeley DB). This allows the company to keep its costs low in its infancy. A quick note here, using Fedora Linux was a mistake. As it turns out, the upgrade path to get patches on major releases (e.g. Core 3 -> Core 4) is quite painful.

Findory uses cheap, leased, commodity servers. Findory currently runs on six servers. The cost per server is under $100/month. They are simple, low end, AMD boxes with a reasonable amount of memory and a cheap IDE disk.

Rather than pay for a news feed, Findory runs its own crawl. That allows us to customize the crawl to our needs and quickly launch new products (e.g. video, podcasts), but does take a fair amount of time to maintain and extend.

In retrospect, I think I have been too cheap. The tradeoff here is between death by burn versus death by resource starvation. While Findory has lived a long time with very low burn, it has been starved of resources, slowing growth and making it difficult to pursue many paths for expanding the products.

Update: By the way, I always welcome comments, but I especially would love to hear feedback in this particular series. Think I should have done something differently? Please comment and let me know. Think I have analyzed the issues incorrectly? Chime in and tell me about it. I would enjoy hearing what you think.

Update: There are some good discussions going on in the comments to this post.

20 comments:

John K said...

Good stuff.

So Fedora was a mistake? What would you go with next time?

Personally, I'm a FreeBSD guy.

Greg Linden said...

Not sure, John. But, yep, I've heard FreeBSD is pretty nice. Don't have a lot of experience with it myself though.

Anonymous said...

I think that's how a lot of web-based companies are getting started these days, especially for those who have been around the block with VCs. My two cents is that your approach makes sense in early on when you are proving the concept and figuring out how to make money. Risk is high and if you could raise money, you'd give up too much equity and regret it later (assuming it was successful). At some point though, if you are getting traction, the risk of under-investing outweighs the downside of taking in money. This is the point to place your bet and go big. After all, there is no reward without risk and you'll give up a lot less of the company when you can show your company works. You want to raise money when you don't *need* it, but rather *want* it to fund growth.

typingtalker said...

Did you do any revenue projections?

Greg Linden said...

Yes, I did do revenue projections in the business plan.

In the end, the estimates worked out to roughly $0.01 per page view from the personalized advertising.

$0.01 a page view seems to be a conservative estimate. Google gets about $0.10 per search. Gizmodo, a tech gadget news site, appears to get about $0.05 per page view.

By the way, Findory's advertising engine is personalized in the sense that it targets ads based not only on the content of the page, but also using each reader's reading history. It is quite unusual.

Anonymous said...

I think you did most of it right, but there are two things I disagree with.
By being completely virtual, even with management, you lose the momentum generated by a few smart people bouncing ideas off each other all the time.

This leads to a trade-off; you sacrifice initial momentum for slow-burn. The problem with momentum is its easy to lose but hard to gain and hard to build up.

Here's how I would have done it: put the three key people (founders) in one office and made everyone else virtual. Spend 3-6 months building and testing. Once you get user traction with your initial concept, then quickly build momentum by building your business plan and getting revenue from ads or subscriptions. If you need more money, then consider VC money. (If you have some kind of revenue coming in, at least you will get a better valuation, and have better funding choices.)

Greg Linden said...

Thanks, Paul. That's a great point on office space.

Bouncing ideas rapidly off people sitting in the same room is important. Eliminating the distractions of working from home or from coffee shops is also useful. Finally, office space creates a separation between work and home and can help reduce the feeling that the founders are always on and always working.

I do think not getting office space was a mistake.

MakkhiChoose said...

You might have mentioned this in a previous post but I'll ask it anyways; what is the general category of machine learning algorithms you are using?
Amazon personalization seemed to be based on the correlation of users buying histories. (Of course, this is painting the picture using a very very broad brush stroke) These user histories are sparsely populated and that does pose a significant engineering problem. Do you have to deal with this problem in generating user reading history vectors or do you work around that by only assigning articles as belonging to some previously chosen, large number of categories?

And here are a few (too many?) questions I would love to see answered in future posts, if they aren't part of something you believe should be confidential.

What is your general opinion about the diversity of findory readers?
Are they as varied (say in some distance measure in some dimensional space) as Amazon readers?
Do you think a company like Amazon would have a significant advantage in this area by viewing this as a transfer learning problem where they could learn some core similarity kernel from user-product histories and apply this as a starting point for learning user-reading histories?

And lastly, thank you for all the wonderful, thought provoking posts.

Anonymous said...

Hi,

I had the same questions when lauching our gay dating site in France (www.cleargay.com).

We released it on very low budget, almost no budget in fact (we did it on extra hours).

But we rapidly saw the limits of this approach : managing a high traffic site is very demanding, especially on human time...

We decided to raise little with business angels, and, one year later we're happy with the decision.

We raised something like 200k$ and it was enough to get a strong IT infrastructure, finance sufficient human ressources...

We reached break even merely 8 months later, so this allowed the company to get really on rails...

Initially, the bad point was that it was very dilutive but now, we're preparing a new fund raising (for european and possibly US development) at a very good value because we demonstrated our potential.

In fact, seed funding seems better to me than doing everything on your own : you get easier bank and partners trusting...

Note also that we choose BA that we're mostly internet pros, so we got really good advices free, both in marketing, IT and corporate fields...

And, by the way, we're running CentOs linux distro and and it's rolling fine our our 15 servers...

Anonymous said...

Do you think that having active communication on something like IRC or IM is a good enough alternative to a physical office when you start out?

How many people did you employ in that initial phase? If it's just one or two people, doesn't a spare bedroom work just as well as an office?

Greg Linden said...

Thanks, Alain. That is good advice, and it is interesting to hear about your experience. CentOS looks convenient for us since it is based on RedHat. Perhaps we will switch eventually, thanks.

Greg Linden said...

Hi, Chasing. The recommender algorithms used at Findory loosely could be described as a combination of social (aka collaborative) filtering and content-based algorithms. It does look for correlations between people's reading histories.

The basic idea is to have everyone in the Findory community all recommending articles to each other, but have the computer do all the work. Unlike social networking sites, all the sharing happens anonymously and implicitly, all behind the scenes.

But, it is still people sharing what they found with each other. The knowledge of what is good and what is not comes from humans, not computers.

By the way, another way to look at these kind of social filtering systems is to think of them as generating millions of most popular lists, the most popular items for people like you. Best selling lists often are uninteresting to me because my tastes are not generic. But, a best selling list for people who read some of the same articles I do, that is likely to be interesting.

As for your other questions, Findory is tiny compared to Amazon, so I would expect the audience to be much less diverse. I agree that a good way to frame the recommendation problem is to look at it as trying to predict people's reading history (and future reading).

Thanks, Chasing.

Greg Linden said...

Hi, Mikal. I do not think IM works as well as being in the same physical space, but people differ on that one. There are companies like MySQL or IMDb where people are widely distributed around the world. Seems to work well for them!

Sure, a spare bedroom or a basement in a house might work well. However, there may be some advantage to having a formal office space with a reception. I have to admit I found it embarrassing to not have nice office space for meetings with potential partners or investors, especially if they were traveling from abroad. It seems there are plenty of relatively inexpensive options out there for getting a single room office with a shared reception on a floor shared with other businesses.

Anonymous said...

Jumping in on the office discussion: I underestimated that prior to having one (I'm now in an office community), but I think now that you should really separate your private and working life concerning time and space (you may have colleagues as friend though :)). It helps me to concentrate more when working and it let's me enjoy spare time more relaxed when a workstation is not in the room around the corner. Second to that it really helps if the office is somehow representional.

zippy said...

I'm with you on being frugal. When you spend money on things that make life easy, you often fail to think about the core problems and how to solve them. One thing I learned from Alexa Internet was that doing things on the cheap forced us to come up with some of our best technical innovations.

Being frugal forces you to think about what's really important. If you have money, you can buy gold plated connectors for everything. If you're thinking about money, you challenge assumptions about the necessity of (Oracle | SCSI | servers with redundant PSUs | ...)

I second the notion of the
importance of being in the same room. I don't know whether it's necessary to have office space, but it is very helpful to have unscheduled face-to-face time. Most of the time, nothing concrete comes from this time. But once every month or two, a real idea or business-changing innovation comes out.

--Pat

Anonymous said...

Hey Greg,

How about a reading list for those on the "outside" to get a good glimpse on these modern AI techniques used Findory and Amazon? Let's say I start with Norvig's AI book -- where to, from there?

BTW, right now I am reading "Starting Something" -- pretty awesome entrepreneural nugget!

Greg Linden said...

Hi, Anonymous. I posted a list of papers a while back that you might enjoy. Hope that helps.

amckinnis said...

It's a fine line between being cheap and ineffective. From my experience, I burned through money on the wrong things - and was not "forceful" enough with others within my company on where the money should go. By the time I realized my mistakes, it was too late.

But, as was witnessed by the big bubble burst -- having lots of money doesn't necessarily help either.

It's an age-old question between too much or too little. Keep going Greg - and remember to have fun with it.

amckinnis said...

I'll chime in on the office space situation....

it's great to have everyone together, for ideas, brainstorming, quick projects, etc. but that can also get in the way -- the same discussions that start off being productive can also quickly become bull-sessions.

I my opinion and depending on the cost of office rental - you are better off putting that money in better communications, finding a convenient meeting room (possibly for free - leverage friends) -- office space can be expensive - and then there are all those desks, and the network connections, and the expectations that go with an office.

But then again, I could be all wrong!!

Anonymous said...

I think you were correct to have done the office space the way you did it. One really does need the feeling like one is "going to work" every day. And not just working out of someone's basement or a friend's house.

And in terms of doing everything else on the cheap: I don't know, long term, if it was the right strategy or not, as I'm not a business-type myself. But I highly respect your approach. It is so much more sane, at all levels, than the burn and churn approach of 98% of the dotcoms of yesteryear, or of the YouTubes of now.