Showing posts with label google. Show all posts
Showing posts with label google. Show all posts

Thursday, January 17, 2013

Move. Faster!

Today we were lucky enough to get some time with Mike Cassidy at Google.  He was far from the expected big tech management speaker.  Mike’s a veteran leader of four successful startups (Stylus Innovation, Direct Hit, Xfire, and Ruba) and came to Google by way of acquisition.  Now he’s part of the clandestine Google X Labs which works on big picture projects like the driverless car. 

Mike is the rare person that carries himself and his interactions with urgency.  Everything from his presentation to conversation style screamed “Lets get going!”.  Quite appropriately the theme of his (well known) talk was on the role of speed as a central business strategy.  My takeaways:

  • Raise before an inflection point in valuation – it lowers closing cost/time
  • Move fast and get all decision makes in the room before pitching
  • Get contracts and LOIs from enterprise customers
  • If you haven’t built it at least have a plan down to concrete actionable details before pitching
  • Hire experienced people – no amount of all-nighters can make up for the speed 5-10 years of experience programming will bring.
  • Make an offer same day.  Draft the offer in the morning, complete a reference check at noon (if the interviews are going well), and hand the candidate an offer letter on the way out.  Ask for an answer the next morning – you don’t want to let them shop it around.
  • Set up all the infrastructure an employee needs day one so they can hit the ground running.
  • Probability of a deal closing falls 10% every day in doesn’t close.  Put pressure of the other party to close or cut your losses early.
  • Pivot hard and fast.  3/4 of Mike’s companies were very different from the original idea.
  • Find great people with great ideas. 3/4 of Mike’s companies were initiated by his co founders (housemate/fellow alum/industry expert).  He also mentioned scanning the MIT 50K (now 100K) site for interesting projects: http://www.mit100k.org/accelerate/2013_semi_finalists/
  • Start small – his first market was only $20M (telephony). You learn the ropes of a business and can build great personal value in that space.
  • CAC < LTV.  The CAC (Cost of Acquiring a Customer) must be less than the LTV (LifeTime value).  This seems obvious, but it’s important to plan for and have a good feel for how this metric is evolving in your business.

image

From the launch timeline Mike laid out (above) I feel pretty good about BuildAFlock.  On a small/prefunded scale we’ve follows this pretty closely to get from ideation to launch in about 3 months.  However, I’m very curious to see if we can maintain this pace at a Series A scale. If you’d like to take a look at his slides I found an old version here (content looks very similar):

http://www.slideshare.net/dmc500hats/best-strategy-is-speed-startup2startup-may-2008

Tuesday, July 24, 2012

Is your web server load high?

A couple weeks ago we were getting worried about CIMLS.com's server.  We were running a larger AWS instance and hitting 80%+ utilization on a continual basis. For a site of our size slinging simple HTML this is ridiculous.  So, what was the problem?

Crawlers....

The bots that work for various search engines were furiously indexing CIMLS.com resulting in almost 40% of our server load.  About half of that was Google.  While a worthy cause, we cut the load in half by lowering the crawl rate to .5 requests/second.  This hasn't significantly impacted the number of indexed pages (let alone organic traffic/rank).

Adjusting the Googlebot's crawl rate can be accomplished by visiting google.com/webmastertools >> configuration >> settings >> crawl rate.  We tried a number of options, but found .5->1 requests/second to be a sweet spot that got our pages indexed without tanking server performance for real users.

Another big surprise for us was that the bulk of the rest of the load wasn't from other crawlers like Bing or Yahoo!  Instead, two foreign crawlers Yandex (Russian) and Baidu (Chinese) were just hammering the system.  As we are a US oriented site, the traffic from these sources was negligible (and probably irrelevant).

We found that modifying the robots.txt was not effective (these bots just ignored it).  We had to use a RewriteCond to block the bot activity:


RewriteCond %{HTTP_USER_AGENT} Baiduspider [NC,OR]
RewriteCond %{HTTP_USER_AGENT} YandexBot
RewriteRule ^.*$ - [F]

Well, finally that did the trick - and removed nearly 20% of our server load overnight.  Phew!