Shashikant's tiny world

Yes, the title of this post is rather poor. There are bunch of ideas in my head for blog post, but many of them need to be polished. The fear of putting something half-baked is embarrassing. But, that also means, many ideas simply die out or lose relevance over time. So, here I am, ready to be ridiculed for these ideas.

Google Upgrades Picasa Storage:

Google unveiled new storage pricing for Picasa and Gmail. Now, you get 20GB for $5 per year. This is 87% lower than earlier pricing. A staggering 87%. At this pricing, if you are using all of 20 GB, the price is roughly 2 cents per GB per month. To see how ridiculously low this pricing is, compare it with Amazon's S3.

Amazon S3 charges 15 cents per month per GB. Amazon also charges for the data transfer. Putting 20GB in S3 costs you $2 at the rate of 10 cents per GB. And reading data out of S3 costs 17 cents per GB. (I am not putting the charges for GET requests on S3, because most of the time it is rounding off error.) In comparison, Google is charging $0 for data transfer. This makes Amazon S3 almost an order of magnitude more expensive if not more.

There is more. Picasa also stores various sized thumbnails to show when you browse the album. I suppose, that doesn't count towards your usage.

It doesn't end here. Google uses Google File System (GFS) to store all of its data. GFS follows rule of 3 - that is there are 3 copies of all the data. The reason being even if one copy is lost (quite routine with commodity hardware), the data still has redundancy. Essentially, per GB per month cost offered by Google is fraction of a cent.

The raw storage costs around 10 cents per GB. That is boxed disk sitting on table. Pretty much unusable. Think about all the costs involved to make it workable - power, machine to connect to, data center, cooling, security, sys admin and so on. The cost of usable storage is actually significantly higher than the raw cost.

Now, you know, why I called this a ridiculously low price.

So, what's the catch? To start with Picasa allows only photos and video (of size upto 1GB) to be uploaded. This storage is also available with Gmail, but Gmail also has a limit of 25MB on attachment size. Unless you are pro photographer or having tons of photos of your kids (my friends do have), it's difficult to hit the limits of available storage.

Then, the next question is if anyway very few need this kind of storage, why didn't Google give 5GB space on Picasa for free instead of current limit of 1GB? Anyway, Google isn't making any money from this $5 per year plan. You may as well earn people's goodwill.

I think, Google is testing waters of "freemium" (yeah, a much abused term, but only one fits the description) with this ultra low pricing. $5 per year won't even register a blip on many wallets. Millions of wallets.

Next idea is to build on this is a backup service. A 100% paid one. Mostly at the price point of $10-20 per year. Comfortably below the Mozys and Dropboxs of the world. Figment of my imagination? May be. But there is non-zero chance that Google will do this.

I really wish Google comes up with competition to S3 at Picasa-style prices.

-----------

Google Closure and the entire toolkit.

As much as I dislike the name, I sense it is utterly useful. Reducing javascript size will ease the stress on bandwidth and will run faster in the browser. Double win for the user. Google is creating tools for developers, spreading awareness about good practices.

Now, some of these tools are actually beneficial for Google as well. Google is strongly suggesting websites to use compression while distributing the text content (html/css/js.) With gzip, 75% or more size reduction is easily achievable on text. When a insanely popular service like Twitter doesn't bother about gzipping its content, you can imagine the (sorry) state of rest of the web.

And you know, Google is hurt every single time the Googlebot hits a site which sends uncompressed content to user. And when Google is crawling Billions of pages every day, you can imagine the bandwidth bill not to be trivial. Small compared to their $22 billion revenues, but not trivial. 75% reduction in that bill would be really sweet.

To lower it's bandwidth bill, Google will push out lots of tool to community. I wouldn't be surprised if they release Google Web Server - Lite edition to the community.


-----------------

Minor Crib about Google Adwords:

Google Adwords offers 4 types of keyword match - Broad, Phrase, Exact and Negative. It is explained here.

So, if you are bidding for "tennis shoes"
- Broad Match will show your ad for queries "tennis", "buy tennis shoes", "tennis sneakers".
- Phrase Match will show your ad for queries "red tennis shoes" but not for "shoes for tennis."
- Exact Match will show when the search query is exactly "tennis shoes" and nothing else.
- Negative Match will not show your ad if the specified term appears in search query. For example, the negative word "cheap" will not show your ad for the query "cheap tennis shoes."

I think, there is a need for something more stringent that Broad Match, but less restrictive that Phrase Match. Let's call it "All Keyword Match." The behaviour for this would be to match presence of all specified keywords, in any order, in the query.

So, if my keyword is "tennis shoes", the ad should be shown for "shoes for tennis" but not for "running shoes." That is as long as the query contains both "tennis" and "shoes", the ad should be shown.

Yes, yes, Google's ad matching is quite smart, but I need this control.

--------------

Google Enhancing the ads.

Google now allows enhanced product ads where, you could specify a picture of the product, it's price, payment information, etc. I expect these ads to perform a lot better than existing plain simple ads of "Buy Wii at Low Price." And these are going to be Pay for Action ads and not the typical CPC ads.

Essentially, the retailers are going to build a fantastic shopping search engine for Google. When you search for an electronic item, the rich ads will dominate the vanilla organic search results and attract disproportionate share of the clicks.

So, who gets affected by this? The shopping search engines, obviously. Many of the shopping search players themselves spend aggressively on Google (see here ) But it seems, Google wants to take a hard look at the arbitrage game. For a long time Google has been a spectator in the shopping search space. Google Products (earlier Froogle) is almost as old as the web search on Google. But they never took massive interest in making it a real product. Putting a better shopping search product would have cannibalized their ad revenues in a big way. But, now that the glory days of 30% growth seem a thing of past, they probably want to see how they can extract more juice from existing universe of searches.

Frankly, for someone who runs a vertical search (bandhan.com), this should not be a good news. But, since we operate in a rather unproven market, entry of Google will help us. Yes, you have heard this cliche umpteen times from entrepreneurs who built their companies painstakingly over the years, just to see Google entering into the business and eyeing a leadership position. Trust me, for us it's not just a statement to be issued to the media, but a reality.

-----

Sorry, right now, I'm too lazy to even proof-read this. Thanks for reading.

Labels: google

Last night, I posted a following message to my twitter stream.

"If LinkedIn is a threat to Monster/Naukri, StackOverflow is fast becoming a bigger threat to LinkedIn."

It received following responses.

"Not sure I understand the analogy. How are u equating the two?" - Arjun Ram

"Huh?! How is stackoverflow becoming a threat to linked-in?! Only techies on stackoverflow, right? Linked in is much-much more!" - Navin

"Really? How so? I'm becoming more and more of an online recluse now, so don't have much of my own exp to lean on." - Shyam

I realized, my original message was somewhat esoteric.

Let's start with basic. Job sites like Monster, Naukri connect the active job seekers with recruiters. The candidates upload their resume when they are looking for a change.

Then the Social Networking boom started where you created a personal profile with information about yourself and bunch on inane things. Social network connected you with your (mostly real) friends for easy communication. The professional counterpart of this idea, LinkedIn, helped you create network of people you are/have been associated with professionally. So, you put up your brief (or long) professional activities publicly. Though, LinkedIn wasn't meant to be a job site like Monster, given the amount of high-fidelity information and social proof, it quickly became an open repository of passive job seekers. LinkedIn realized the recruiting angle to professional networking and is aggressive in pushing recruitment solutions. If you see "Products" line on LinkedIn's home page, you will see links for "LinkedIn Jobs", "Recruiting Solutions", "Jobs Directory." Currently, LinkedIn has 50 million users and adding a million every 12 days or so.

LinkedIn didn't start as a career site, but as side effect of professional networking it has become a threat to Monster. Just like, you know, Facebook didn't start as a photo sharing site, but it quickly became a 5 times bigger than Flickr.

Cut to StackOverflow(referred as SO henceforth). This is a Q&A site for programmers. It is a great resource for every geek who codes for daily bread & butter and occasional beer. Growth of SO has been astounding. It is barely an year old, but it has decimated the long standing leader, experts-exchange. See the traffic graph from Google Trends.



SO also has a built-in incentive system, where it is beneficial for you to register before asking or answering the question. And your public profile contains all your questions, answers and reputation. (Check mine) If LinkedIn's recommendation is a good social proof, albeit with a small problem, your SO profile is even bigger endorsement coming from "random guys on internet."

Given the popularity of SO, they have recently launched Careers section. They already had a job listing, but they careers is more comprehensive solution. An active job seeker would upload the resume. And to filter out non-serious job seekers, it is a paid service for candidates. The recruiters will get an integrated view of a candidate's public profile and private resume. I suspect, if you go through a person's questions and answers, you will get a lot better idea about his/her competence than an hour long interview.

Yes, compared to 50 million users of LinkedIn, SO is still tiny. But, it's growth has been astounding. It does a million page views a day and I have read that almost one-third of world's programmers have visited this site at least once. It currently claims 120,000 active users. SO will soon compete with various product/technology specific mailing lists which are great for one-time solution, but extremely hard to use by subsequent programmers.

Combine it's usefulness, quality and growth, I wouldn't be surprised if grows by an order of magnitude or two in next couple of years. At that time, it will be a formidable force against LinkedIn & Monster as far as hiring techies is concerned.

Labels: linkedin, stackoverflow

Real time search is the newest kid on the block and everyone is quite excited about it. The excitement almost rivals that of the non-existent iSomethingOrOther from Apple. The obvious way to make money is now a decade old - show relevant ads for the search queries.

So far so good. Let's understand how real-time search is different.

1. Trends is a wave. When people are tweeting about something they are creating a wave. Now that essentially means the trends come to you rather than you searching for something. It's almost like news, where most of the time you want to know what's happening right now and may be use search occasionally to dig into the less prominent stories or simply find relevant information from thousands of news stories.

2. Waves die. The promise of real time search is to provide relevant information which is just out of the oven. A very relevant tweet which is 6 months old has almost no value. Contrast this with the Google world where good content gets rewarded handsomely over a long period of time (until something significantly better shows up.)

3. Bar is not high for quality of the content. After all, how much information can you cram in 140 characters? I don't see one tweet to be far superior than the next 10.

How do these idiosyncrasies of search affect the ads.

1. Since trends are coming un-announced, it is hard to catch the trend. In Google world, advertiser roughly know the keywords. They can create ads for those keywords, bid and let google handle the massive ad auction while you sleep peacefully on a soft pillow. You can take a look at your campaign once in a while (a day or a week), tweak a little and again go the soft pillow.

But the real-time search is different game. The google world of adwords will look almost static in comparison. The trends are getting created quickly, you don't have enough time to think of all the keywords & the bid amount, and you need to respond fast.

For example, you are a chocolate company and few people found worms in it. They are very offended by it and it is becoming a trend. You need to go out, put the ads pacifying all the consumers.

How will this be done? Well, I don't know. May be you need a service which gives early warning about possible trend.

2. Since trends die, you need to constantly keep on tuning your ads. You don't have the luxury of soft pillow in this world. You are on your toes all the time.

3. The current discovery model is not geared to handle the micro-trends or long-tail trends. More than 12 million tweets are created everyday. There must be so many trends getting created every hour. How do you sense those tiny trends and target ads for them?

4. Real time search is only about scanning the content. Like say checking headline and going away. You don't care about the quality of headline, just the message. So, ads are not really competing with content for users attention like regular search. This is sweet! If ads are just right, content will lag the ads significantly in terms of user attention.

I don't know if real-time search is for real. For all you know it simply is a mass illusion. But if it shows promise in early stage, we are in for an exciting ride.

Labels: ads, real time search, twitter

The national Unique Identity Card is the smart card is based on the specifications of Smart Card Operating System for Transport Applications (SCOSTA.) [Source:BS] Now, SCOSTA is of interest to me as I have personally seen it developing in the lab back in 2002. SCOSTA has been completely developed at IIT Kanpur. It was the M.Tech thesis Ravinder, who was senior to me. Ravinder belonged to the rare species who manage 10/10 without sweating too hard. The professors involved were Prof Rajat Moona, Prof Deepak Gupta and Prof Manindra Agrawal ("Primes is in P" fame) all from the Computer Science department.

You can check the presentation by Prof Moona to learn more about SCOSTA.

If you need the proof that world-class technology can be developed in India, you don't need to any further. I wish the management gurus of IIM learn something from this. IIMs recently chose Prometric ETS (the company which conducts GRE) to conduct computerized CAT. Apparently, it simply gives you one of the 30 question papers with random ordering of questions, and not completely online and adaptive test like GRE. From whatever little I have learnt about technology in last 10 years, I can assure you developing operating system for smart card is incredibly harder compared to conducting computerized exam however sophisticated.

Labels: IIT Kanpur, SCOSTA

Nilekani has been appointed the Chairman of Unique Identification Authority of India and he will be equivalent of a cabinet minister. I must admit I was quite disappointed with UPA govt's last term and was hoping they will lose election. But a thumping victory has catapulted the mood of the government. And now I have nothing but admiration for the new govt. This definitely marks a new beginning in governance where an eminent, competent person has been entrusted with a major responsibility. According to some estimates this project will have budget of Rs 10,000 Cr spread over 3 years.

But, guess what, not everyone is happy with govt's this decision. Immediately, the red flags of conflict of interest have been waived. Yes, Infosys will be competing to get this project (or part of it as I see.) Nilekani has already resigned from the Infosys office. He has also made it amply clear that he will excuse himself from the decision process which involves Infosys. It seems that's not enough. They are asking what will he do with his some $700 million dollar worth holding in Infosys.

Few things.

First. Infosys is one of very few companies in India which made "Corporate Governance" way of life long before the term gained currency. Nilekani has been an integral part of the process with which they built Infosys. The conduct of Infosys management has been impeccable. (Short exercise. Pull out the list of top paid management executives on India. Check how many of them are promoters of the companies. Check if their compensation was in line with the performance of the company.) If you are asking for highest (and unreasonable) ethical standards from the man who has amply demonstrated it, you need a break. And let's not forget we are the country which elects hundreds of murderers, embezzlers, criminals, rapist, robbers as our representatives.

Second. Let's find out how much Infosys stands to earn if they win the project. The budget of 10,000 Cr will be spent over 6 years. A significant portion of this money will be spent on hardware which includes hundreds of millions of smart cards. (When you buy a new vehicle, the RTO issues a smart card with details of vehicles. They charge Rs 350 for that. If you take Rs 50 per card or $1 for mass production we are talking about hundreds of millions of dollars, which is significant portion of the budget.) I would be surprised if the actual spend on software is more than Rs 3,000 Cr in all - that is Rs 1000 Crore per year. Obviously this is a huge project and you will have multiple software vendors. Let's say we are talking about 3 vendors. That roughly translates to Rs 300 Crore per year per company. If Infosys is one of the vendor, it will get 300 cr revenues per year. Don't expect the profit margins on this project to be superhigh as there is no labour cost arbitrage.

To put numbers in context, Infosys reported revenues of Rs 21,693 Crores. An additional 300 Cr will hardly change their financials in a tangible way.

For the sake analysis, let's say Infosy's stock prices moves up by 10-15% due to revenues of this project. Now, what is it in this world that Nilekani can do with $800 million but not with $700 million? Nilekani has spent his lifetime to build reputation and name for himself. He is not going to take any action which will jeopardize that pristine reputation. It's foolish if someone thinks he will trade his reputation for few extra millions. Last I heard he didn't own a private jet or a yacht or an island. He seems like a frugal man to me.

It's been 15 years since the start of Voter's ID card project and even today the cards are issued with wrong name and address. This is a himalayan waste of public money. We need competent person to look after such gigantic projects. By making fuss out of nothing, we are scaring lots of good people from corporate world whose expertise govt may wish to utilize. Now, if Deepak Parekh is asked to head a new authority on affordable housing and there are going to be similar noises due to his connection with HDFC, we will keep him away. It would be a shame if such great people, who are willing to commit their time and energy for the larger good of society, are kept away due to some non-issues.

All the best, Nadan!

Labels: Infosys, Nilekani