Sunday, July 08, 2007

The Twin Future of Search

Search is the most widely used application on the internet and while it has come a long way from the yahoo index or alta vista search there is still vast room for improvement in the search field. One area that at least needs to be distinguished is subscriptive search. I know, you're saying subscriptive isn't really a word, but it sounds so much better to me than "persistent search". You see, I think there are really two primary search use cases.

There is the one we are all familiar with and use Google for daily. That is ad hoc search. Ad hoc search is used when we need to find something. It may be something we used to know but have since forgotten, it may be what the weather is going to be like somewhere we're going to, it may be figuring out what the GDP of Austria was in 2006. These are tasks that we don't repeat day in and day out. We perform the search when it is warranted, often looking through only the first page of search results, plucking out the desired information, and are on our way. We don't spend a lot of time with the data, we find it, consume it, and move on.

The second form of search is subscriptive. There's that word again. Ok, we can call it topical subscription but that is so long and unwieldy. Subscriptive search is used when we need to know what's going on around a topic that we're highly interested in, and please keep us up to date. I don't want to go and re-search, I want you to tell me when there's something important and why I should care. A number of weather services have popped up to provide this in the specific information vertical of weather forecasting. I always want to know what the weather is going to be like where I live. Don't make me go search for it every day, just deliver it to my desktop or email me and definitely be proactive about letting me know when some dire weather pattern is about to descend on me. The same is true for a number of common areas of interest: sports scores, movie times, top news stories, and most of the things that you can customize your My Yahoo! page today.

The thing that is missing is in the Long Tail. If I have a passionate interest in antique lawn mowers I can find a couple of blogs or I can continuously search, or I can use something like Google Alerts. None of these get me what I would really like though, which is a page that I can go to that pulls any new posts from blogs or news (with rankings that are personalized to my specific interests), shows me about-to-expire auctions on ebay for those mowers, and whether there are any upcoming shows around the country. It would also allow me to tag the content I wanted to save so I could review it later. I could also expose my page out so that others could find it and use the knowledge I gained in building my page and just read that page, or use it to build out their own page with mine as a template.

This would essentially allow anyone to publish their own zine on the areas that interest them most including multiple content sources, and would be perhaps the 2nd stage (after blogs) in really creating an n-to-n publishing model that will fill a key role in next generation content distribution and discovery. Hey, what a great idea for a startup!

Wednesday, July 04, 2007

Supernova startup shakedown

These are my notes from a startup pitch session paneled by Michael Arrington of Tech Crunch and Paul Kedrosky, Josh Kopelman, Julie Hanna Farris. All of the companies at least had early stage funding and were given about 5 minutes to present each, sort of startup speed dating :)

trying to match online video to advertisers. over 1M ads currently Contextually analyzes video, audio, and metadata. Monitors users interaction with the ads and adapts to their prefs.
Pretty cool. continuously analyzes stream and finds most relevant ad in their ad database.

Adaptive Blue
Semantic web company. People + shortcuts = getting to info faster, also personalized web. Browser add-on -> blue organizer = Personalized Smart Browser. "i'd like to" menu shows contextual intent related actions. Can highlight text, tell the organizer that it is a book, then shows shortcuts to buy the book or see a book review. Has a trademarked piece called "Smart Links" that shows these contextually related links.

Aggregate Knowledge
Six months ago launched at demo conference. Powering discovery for 50M users per month. Discovery happens in offline world all the time. how do you discover online. finding serendipitous piece of content. creating better navigation metaphor for Implicit affinity matching on a massive scale. worlds largest implicit social network.
Video search. Matching users with content they would be interested in. Compared directly to Google. Searches across the web, unbiased compared to Google (why is google biased?) Has generated fan landing pages for all major shows (thousands and thousands of shows).
- crawling and indexing - prop technology to build a better index. gemstar is a customer uses to automatically generate a tv guide
- relevancy - something better than google. blah blah blah.
Tried to get on their site but it said was undergoing some improvements. Pretty lame.
funded by DFJ

Critical Metrics
Music discovery recommendation and search field. lots of competitors. why would they come into this field at this late stage. no matter how much you use the services, they won't keep you up to date with NEW music. Why? because it's pretty much impossible. There's too much music that comes out. Each day there are about 1000 songs. Is a recommendation engine that ploughs through all the music.

phone talking company. communications through your social network. Phone is not currently attached. Phone is not part of the profile because of privacy issues. They handle privacy issues. Sends email with voice message to user. Can put a widget up on your page that allows readers to call you. You then permission people as to whether those calls go through.

pando networks
peer assisted media delivery. Cuts cost of deliver a 1G media file from $200,000 to $5000. Currently serving 8.5 million clients delivering 70TB per day.

Just came out 2 months ago. Old school polling with web2.0 social. Can do a lot more now, share poll with friends, comment on poll, answer with video, pictures, etc. Capture aggregate data with individual commentary

people search engine. Looks pretty cool. Put in blogger, returned Michael Arrington, clicked on related term "Tech Blogger" and Tim O'Reilly came up first. Will need to check it out more. Uses user tagging to build relationships between people and concepts. Good presentation.

Online product research is still too hard. Hard to help father buy computer. normalizes rating systems across all the prod research sites. Created product sentiment database from user reviews, bloggers, expert opinions, market buzz, and manufacturers sites. WizeRank - consumer report for the future. Aim is to create a true product satisfaction score. Starting syndication relationships now. Should talk to this guy.

The shortest distance between great food and your tummy. online meal order and delivery service. ebay for takeout.

Cosumer centric services, + power of sync + focus on design. MS research background. "Adaptive and scalable UI" 16 tile customizable views of what's important to you on your mobile device. Updates by polling with any new info, basically a 2 dimensional widget space on your phone. can nav and then zoom in on tiles. viral spread of widgets by monitoring widget heat.

Untether online services (such as so you can access and use from your phone.

13 companies presented, 1 was fake. got to vote via Soda which one was fake. Definitely ZapMeals. I must have missed a couple there. oh well.

Panel Feedback
Kedrosky - interesting that there were no wiki mentions, no ajax this year. Lot's of competitors to google but loathe to mention google name (awaken the monster). nature of demographics on web have changed so much in the last 5 years. Much has moved to entertainment sites and bloggers. Which companies appeal to innate laziness. Likes Big unsolved problems attract big buyers. Monetizing video traffic is such a big problem, esp. with creating live overlays. Consumer related search technologies must painfully avoid Google or will just become another tab in google (without consent or payout).

Arrington - new phone apps are exciting, perhaps driven by iphone functionality and screen real estate. touch screens are fascinating. not easy to impress with just a web application. Likes just acquired by AOL. he thinks is better. Better video search is hot.

Julie Hanna - Bias against companies wanting to be destinations sites. Jangl brings two common forms of communication and blends them together.

Doesn't feel like a bubble. Not enough froth out there. Need more people out the other side getting rich to create a bubble. Lot of companies in tweener stage. Easy to get early stage money, not so easy to get to the next round. Seed stage is used to validate or improve hypothesis. iPhone may be the catalyst for the next rich ecosystem for startups out there.

Tuesday, July 03, 2007

Sales Process

In the early days of a startup the word "process" does not present itself, except in the occasional meeting where someone on the team feels like they are drowning and need more structure. It is all in the doing. Sales in particular feels fairly process free to start with (from the CTO's point of view, so take it for what it's worth) and is almost strictly opportunistic. Don't get me wrong, there is a strategy, but what process there is, is fairly loose.

The methodology goes something like this: call friends and acquaintances to ask for "feedback" on the product. These are either truly prospects or fishing expeditions. Most people want to help and will at least give you feedback, or will point you to someone who might actually care about your product, or in the 1% case be a great fit for your current product and move forward with you. You work closely with the 1%, incorporating the features they're asking for that make sense to the core direction (or change the core direction), ask for money, and close a few early adopter deals.

Later as the product and strategy mature, process must be introduced to make selling repeatable and scalable.

Making it repeatable means that you've found a message that resonates with a customer domain, you know how to reach the people in the domain, you can identify 1) the sponsor in an account, 2) the gatekeepers in an account, 3) the buyer in an account. Further, you know how the budget process and cycles work within the customer domain, and the product packaging and pricing make it easy to sell into the budget process. Most of this is intuitive for great sales people and it is amazing to watch them orchestrate this process, translating it into a consistent revenue stream for the company. Great sales people are also few and far between. If you find one, hold onto them.

Making sales scalable means that the cost of sales is not too high. This comes down to about three things in my opinion (and remember I'm the CTO ;)
  1. Sales cycle is not too long. If you're only hunting elephants it will take a long time to get revenue in the door, the customer feedback will be sparse and may take the company in a direction that prevents the first sales process tenet of repeatability.
  2. Doesn't require too much setup and configuration time that the customer doesn't value (i.e. doesn't pay you for)
  3. Doesn't require custom development work
So, when you're getting started, run fast and loose, but when you see the signs of the startup project transforming into a product start instituting the process. And be sure to recognize whether you have these process skills or not. If you're not the person, then go out and find someone you can work with, that is passionate about the product, and has the skill set and proven track record to go knock it down.

Sunday, July 01, 2007

Wallstreet and Web2.0

Since Collective Intellect is in the business of finding novel information from consumer generated media sources I often get asked about other opportunities that exist in mining information from the web for trading. Activity is really just starting to heat up in this area with multiple companies getting started and several acquisitions going on. It also seems to me with the popularity of more powerful web interfaces via AJAX and Flash that there are even some opportunities to create research platforms that are easier to use, more robust, and with more information than the current king of the road, Bloomberg. I'll reserve bitching about Bloomberg for a future post however.

I've jotted down an unordered list of what I've come across so far and will spend some time in future posts describing each one in more detail.

  • Rumor Detection - Find the people that are speculating on some aspect of the company that would likely have material impact to the stock, such as: big deals, management change, mergers, drug trial results, etc.
  • Sentiment Measurement - Detect sentiment towards a company, it's management and directors, and products and run statistical models to compare this against stock price and volatility
  • Event Activity Correlation - Track a company event, such as an earnings announcement, that has material impact to the company and measure the amount of chatter before and after the announcement in order to gain insight into what the crowd thinks will happen to the company stock relative to the event.
  • Citizen Analysts - Find the people in Consumer Generated Media (CGM) who speak regularly about a company and offer unique insights into that company or its products.
  • Trend Death - For consumer packaged goods that end up on auction sites, mine those sites for popular product pricing. When prices reach an asymptote and start to trend down, it is possibly time to short the stock.
  • Fill Rate Prediction - Anyone who takes online reservation for resources whether it be airlines, hotels, cars, that have variable rate pricing models, scrape pricing information. Lower prices = lower fill rates and thus lower top line revenues for the company.
  • Passive Site Visits to Stock Price - Use Comscore or information to track the number of visitors to a consumer packaged goods site. This shows invigorated consumer interest which may translate into stock price movement. I believe there was a study showing AAPL site visits as a predictor of stock price movement.
  • Active Site Visits to Revenue - For companies that sell products online, track the number of unique visitors and clicks within the site. If you can figure out the customer conversion to average order size you can estimate revenue contribution from the online property.

New bike

After accidentally running over my old Wheeler bike in the driveway, I thought it was high time to buy a new bike with lighter frame, good components, etc. And after trying out a bike made by Seven in a recent triathlon determined that I definitely needed to upgrade!

I tried a few bikes from Trek, Specialized, and Giant and decided on this one. It's a Giant OCR Limited Ed. from 2006. I guess they ran out of OCR2's and decided to do a limited run with the OCR3, upgrading the components to all Ultegra and new shiny paint job.

Anyway, picked it up on Friday and rode it today for the first time. Good speed and handles well in the corners. It's definitely a more comfortable ride than my Wheeler with the carbon frame softening the road shock. I'm going to have fun on this bike this summer :)

pull your banner ads until google does a better job