I’ve been thinking of blogging about my employment with Yahoo! for a while, but never got around to it. Reading today, about yet another impending Yahoo! reorg, I decided to write about what I felt like while I was there, what worked, what didn’t, and how Yahoo! can get out of the rut it finds itself in.

I was employed by Yahoo! from Oct. 2005 – Nov. 2006. It was my first job out of school, my first full-time job indeed. I had only done a couple of internships (in Singapore & Santa Fe, NM) before it.

It started out great. I got the offer an hour after I interviewed with Yahoo! (Burbank/Pasadena office), which was a welcome change since I was going back-and-forth with Google at that point. Google had flown me from Tucson, AZ (where I was finishing up grad school) 3 times to their Mountain View headquarters, and after having talked with 22 people (yes twenty-two, I kept a count!) on phone and in-person they still couldn’t make up their mind. But I digress…

I joined Yahoo! in the YPN team, which was going to be the adsense competitor. I was excited for multiple reasons

• Somewhere around that time I had started reading blogs, and was totally hooked onto the online “thing”. I was graduating with my thesis in AI/Multi-agent systems and was excited about applying what I had learned to a new Yahoo! product.
• I was going to live in LA, and get to live a fast-paced life
• I knew ad-serving had lots of computational challenges, from scalability to being able to process large data-sets, relevancy of ads, personalization etc. etc.

My team was great, and I made some very good friends. Then in my first week at Yahoo!, Panama happened! Panama was going to be an overhaul of the Yahoo! advertiser product and almost everyone at Yahoo! Burbank was going to be working on it. Except for my team, since we were publisher focused.

That was a bummer, since that meant our team was not going to get the attention/resources it needed. There were other bummers, like every team at Yahoo! “pushed back”. I didn’t know if it was territorial or not but people just didn’t want to take “risks”. There were lots of meeting, but usually nothing came out of them.

I also realized that the team I had joined was going to be managing the front-end interface which publishers were going to be logging into etc. I wasn’t going to be designing algorithms, writing pieces of ad-server code or facing any scalability challenges at all That all was being done by another team, which seemed to not like us at best, hate our team at worst.

There were other problems with Engineering @ Yahoo!, which I didn’t realize back then, since I didn’t know any better. There were virtually no spec/design meetings, no code reviews (not including paper code reviews once it had already been written!!), we were using CVS (seriously?) to manage our code. There were motivational issues too as I saw it. The VP (Engineering) to which my team reported to barely knew me. Now not that its a big problem, but in my 1 year there, he did not get the entire team in a room and talk to us, our problems, motivate us etc.

Now there were exceptions. A star employee Qi Lu. Even though he was a SVP, within a month of moving to Burbank, he called my entire team for a meeting, wanted to know each of us, was just an all round great manager. Most stars like these have left Yahoo! (Qi Lu is now with Microsoft – announcement). Most of the product managers I worked with were rockstars too, so it wasn’t the foots on the ground that was the problem, it was more the VP and above employees that have failed Yahoo! They had the talent, they had the resources, they just weren’t able to execute.

In short my advise to Yahoo! is:
+ Develop common engineering infrastructure so that every group does not re-invent the wheel in terms of server technologies, persistent storage/caching etc. Ala Google app engine.
+ Keep closing properties that suck, or merging them with pre-existing ones
+ Stop trying to beat Google at search. Come up with innovative advertising related products, that are not simply search. Maybe focus on more on product search, or travel etc.
+ Small teams, trim down and promote the good guys/gals!
+ Focus on the enterprise space also, its huge!
+ Hire senior execs who truly “get it” (case in point, Jeff Weiner who is now COO of LinkedIn) and care about the people who report to them, directly or indirectly.

This isn’t meant to be a rant, or anti-Yahoo! commentary. I think the company always had potential and I’d like it to do well and innovate (case in point YOS). Hoping for the very best, and wishing Carol Bartz Best of Luck!

As many of you know I recently started working at Tivix. Working to solve problems faced by social entrepreneurs and Non-Profits is a good feeling. I came up with an idea similar to igive (there are others out there too), but wanted to make it simpler and easier to use (without forcing people to login for example)

I came up with a rhyming name for this concept, Givix: Giving while shopping! Thanks to Amazon Associates, commission junction and linkshare it was made possible!

Also thanks to the simplicity and free scale of Google App Engine I am able to write the code, test and launch Givix (on another note I think the Django model interfaces are easier to use, or maybe I’ve simply become used to them. I do like App Engine’s online dashboard, ease with which I can edit model instances, see logs/performance etc.)

I enjoyed writing this simple app, I hope to enhance it more next year. Hopefully Givix will be able to raise some (however small) money for a few causes this year. Given the looming recession I don’t have high hopes, but if you’re reading this and are going to be shopping at Amazon.com, buy.com, sketchers, macys.com, cafepress.com or others, then do consider clicking through Givix before making the purchase. Every cent counts!

New API’s that would be awesome:

+ Friend list API to get list of friends (users) of currently logged in user. This will enable sharing applications.
+ Jabber messaging api (for currently logged in user and/or any of his/her friend(s)). Real-time messaging anyone?
+ API to schedule execution of arbitrary code at regular time intervals, which can span > 2 second request time limit currently imposed (hopefully)

More feature requests:
+ Be able to set how long the user session lasts

UPDATE:

+ How about more than 3 apps per developer?
+ Also would be neat to be able to take *1* app and have it run in several Google Apps domains?

This post could easily have been named “Targeted Blogging” or “Feature requests for Feedburner”

So I started thinking about a RSS/Atom feed that is intelligent. Meaning each of my blog’s feed subscribers can potentially be offered unique content. So let me present a few scenarios:

- Wouldn’t it be great to be able to write a blog entry about music, and have only people who had “music” as one of their interests, be the only ones who see it in their Google Reader/My Yahoo!/NetNewsWire ?
- What if I was organizing an event in San Francisco, and wanted to let only people who are in California know about it ?

Problems:
- Doing Geo Targeting for ads within feeds is easy, since they load within the browser as a separate request, hence the ad server can act with intelligence about time of day/location etc. Sadly the same is not true for my blog entries. I write them once.
- Scenario#2 above is a bit flawed, because what if my friend from NY is going to be in California during that time period, then he/she will not be able to attend, since my blog entry won’t show up in his/her feed!!

Solutions:
- When the user comes and clicks on my “Add to Yahoo/Google” button I ask them for some basic info, such as age/location/profession etc. They may chose to ignore it of course, but if they don’t they are giving me important information.
- A better solution is for feed burning platforms such as Feedburner to maintain a persistent profile of users who are subscribing to feeds. So my profile would read: “Sumit Chachra, Male, 26, San Francisco, CA ……..”. This information will *never* be shared with feed publishers, but they can use it to target content (much the same way Facebook ad targeting works)
- The problem with scenario#2 can be resolved using a few techniques. Assuming Google Calendar has a feature called “Location” where in I can feed in my coordinates for specific dates, and this information can be shared with Feedburner, then my feedburner profile is really “dynamic”! We’re headed down the path of intelligent agents here really, and its all starting to sound like a scifi (scary) movie at this point

All this implies that each subscriber to my feed has a unique feed url, which will spit unique content to each such subscriber, based on certain data points that the feed url platform knows about them, and using which the feed publisher can do targeted content delivery!

I wish this existed!

So I know personalized search is hot and all. But how long do I have to be using the search (logged in as a user) for that engine to start personalizing results ?

Recently (last week or so) I noticed that search for TIVO on Google stopped returning their stock graph etc. at the top. It returned the TIVO website instead as the top link, with an additional “Stock quote for TIVO” below it.

Now I’ve clicked that a few times, forced myself to search for “TIVO stock” instead to get what I want; but still Google hasn’t learnt that when I am searching for TIVO I am really looking for the stock ticker. What they have learnt of course is that “most” people when searching for TIVO (rightfully so) don’t want to know about its stock. How is that personalization? Thats just the law of averages being forced on me.

Please fix this, and the fact that the SERP shouldn’t keep changing on a weekly basis. Thanks!

It was a long time coming! Yahoo! chose to use Flash (Flex?) to build the online version of messenger. Will I use it, probably on my personal laptop at home (one less piece of desktop software to install… phew!). Does it even come close to the Gtalk that comes integrated with Gmail (launched more than a year back)? Not for me, but I can imagine people who would want all the bells and whistles of their regular desktop messenger (fonts, text size, bold/italic etc.) to like the new Yahoo! Web Messenger. Some of my comments on this new products and ways to improve it:

History: So finally I can see my chat history. But why not give a search box right at the top instead of making me click on a link? Pretty basic interaction designer stuff I assume? Instead a web search box is provided at the top.

Cntrl-W / Cntrl-G / Cntrl-Tab: None of these nice shortcuts works. Cntrl-W should close my current chat tab, Cntrl-G should let me buzz my friends (my friends can buzz me, but my window doesn’t vibrate!) and Cntrl-Tab should let me move between chat tabs, just like I can on my Firefox (or at least Tab like the way it does in the Gmail/Gtalk)

Copying multiple lines: Doesn’t work. Only one line at a time. Thats pretty lame.

Banner ad on the right: Was it very essential to put that ugly banner ad on the right, specially when a new product like this launches? If I were the Product Manager on the web messenger team, I’d instead match (and rotate every 10 minutes or less) text ads related to my past/current chats. I’d bet this would increase CTR significantly. Irrespective its worth an experiment for sure. Simple bucket test to conduct.

Login popup: Why does a new popup window need to be opened to make my login? Its a very jarred user experience, different from every other Yahoo! property I’ve used before.

Gmail has been down for me for the last couple of hours, out here in India. Anyone else facing this downtime too? I’m getting the following message:

Temporary Error (502)
We’re sorry, but your Gmail account is currently experiencing errors. You won’t be able to log in while these errors last, but don’t worry, your account data and messages are safe. Our engineers are working to resolve this issue.

Please try logging in to your account again in a few minutes.

I wouldn’t really call more than 2 hours “temporary”. Someone missed their SLA targets!

UPDATE 1: Its weird, I get the following error now (choosing the HTML view does open up my inbox… its broken though since I have no vertical scrollbars?!?):

Loading…

This seems to be taking longer than usual.

If you are using a slow Internet connection, you can wait a bit longer for this page to finish loading, or just use basic HTML view for now.

If you are using your normal Internet connection and you usually get past this loading step without any problems, please refresh this page in your browser. If you continue to have trouble loading your account, please visit the help center for troubleshooting information.

UPDATE 2: Its finally up after a 4-5 hour outage!

UPDATE 3: Its been 5 days since I wrote this post. But as can be seen from the comments below (and the traffic I am getting to this post), people are still facing this issue with their gmail accounts (I am not). This seems to be a user specific access issue, rather than a mass outage. It might be that gmail is performing some system/software upgrades. Being patient is the only way out! I did not lose any of my emails/data, so don’t see a need to be concerned as of now.

Previously I wrote about Google Calendar and Gmail improvements, here and here respectively. In this post I gives more product ideas for Google Reader and iGoogle (Seriously why doesn’t Google have a URL like http://igoogle.com or http://igoogle.google.com for it ?? Thats the first thing someone would try the moment they hear Google has launched… err re-branded something to “iGoogle”):

Google Reader

search and auto-filter: Search has been the most asked feature for all GReader users out there. I’d like to see a general blogosphere search (similar to Google Blog Search) built into GReader and a search box specific to the feeds I subscribe too. The auto-filter feature would enable me to save some wildcard searches that would remove matching entries. This would be great since for example I hate reading blog entries about everyone’s del.icio.us bookmarks etc.

- Easy “mark read”: I treat my feed entries in GReader like emails in Gmail. But sadly GReader lacks the archive capability. I’d like to see a easy (one-click) way of marking entries as “don’t read”. These are entries I don’t want to open (since after reading the one line snippet I know that its not worth my time), but don’t want to make the effort of opening them and hence marking them as read. Cause of the sheer slowness of this process (and the fact that I don’t want to mark things as having been read, when its not the case) what ends up happening is that I have a bunch of entries that are unread and clutter my view

- Better auto-sort: I know this feature is sort of in beta, so I’d like to make a few suggestions on how to improve it. So whats wrong with it? The relevance of this feature is for the top 10-20 posts only. Beyond that it starts clustering posts from the same feed in some random cluster size of 4-8 posts. Visually seeing such a pattern, anyone can realize its far from intelligent. I’m sure the GReader team is thinking of using the page rank/starring/subscription/read count of feeds/feed entries etc. to come up with some kind of ranking (apart from posting frequency which I don’t think is the best measure; at least not when its used in isolation). If they can implement the “mark read” feature mentioned above, then they would have an additional measure to use and the read count spurious behavior (I manually mark all del.icio.us entries as “read” to force them to go away) would go away too

- Add to Google: My experience with adding feeds to GReader is less than ideal. For example clicking on the “Add to Google” button on the Official Google Blog leads users to this page. Now the problem with this page is clicking on “Add to Google Reader” actually opens up GReader again (again cause I usually already have it opened). I’d rather make this adding procedure quick and snappy with the ability to add labels etc. and a link to GReader. I usually add new feeds cause my readings on GReader lead me to one, hence the situation where I already have an instance of GReader open.

iGoogle

- Variable width/height for each category: Giving the ability to widen/shrink the width/height of widgets would be great. For example I have 24″ monitors at work and my stock module (which Google calls “stuff”!) has the stock ticker and price/moment etc. too far apart.

- Mobile Version issues: I like the fact that iGoogle has a mobile version. I have 3 issues with it though. One being the fact that I can only add a limited set of modules to my mobile version. Secondly any module I want to add to my mobile version of iGoogle needs to be on my web version too…. for example I don’t want a GReader module on my web version, but on my mobile one!? Lastly I’d prefer not having pagination on the mobile version… or at least not after 3-4 modules only. When I want a quick glance at all my modules, I don’t really want to make an additional click to get to my other modules (mobile interfaces/data speeds are not conducive to excessive user interaction).

My first post on Google Calendar improvements can be found here. In this post I focus on Gmail:

Spam Sort: Gmail spam detection is great, but I’ve had an occasional mail or two marked as spam; when they were really meant for my reading pleasure! For anyone who gets a few thousand spam messages a week (I do), its a pain to have a glance at all of them and try and make sure nothing non-spammy has gotten in. What I’m suggesting is a section within the spam label, that lists emails to me, that were on the borderline for being marked as spam. A 0-10 ranking would be even better, but I guess a binary label would be better, so as not to give the secret sauce (some advanced Bayesian Filtering) behind Gmail spam filtering.

Viewing big messages: By big I really mean long. For anyone who gets emails with thousands of lines in them (code reviews anyone?) this is a great feature. If your email is too long, Gmail gives you a link at the bottom saying “View entire message”, clicking which opens another window with the entire email contents. The problem with this is that if you want to reply to that email with inline comments, you’ve to open it its entirety, copy the email contents, paste them back in gmail and then type your reply. You of course lose all the nice indentation etc., plus this is a real pain cause the new window that was opened has a the Google logo and other HTML and selecting a few thousand lines without that is a pain! Hence I suggest having a “no clipping” option for long emails (if I’m willing for the nifty gmail AJAX to take a while loading my message… so be it!). Or possibly a simpler solution (from a development perspective) would be to have a reply option in the new window thats opened for viewing the entire message.

Attach notes to emails: I’d like the ability to write notes associated with email threads/specific emails in the thread. This would be in addition to being able to label/star messages. These notes should be part of my Gmail index and automatically appear in my Google notebook (with a special gmail label) with links back to the appropriate email thread/email in my gmail inbox. I’d say notes should be allowed on more Google apps (currently on Calendar only… but these don’t appear in notebook), hence Google notebook becoming an even more central part of my daily routine, a one stop place to see all my notes, not only ones I explicitly typed in there!

Quoted text: Gmail automatically recognizes text that has been quoted from a previous email in an email thread. This is a great feature since it saves me time, in terms of being able to see new text written by someone and not see old contents. But sometimes I’d wish all the quoted text be shown with a single click (instead of me having to click “Show quoted text” 10 times in an email). This definitely is low hanging fruit cause it would take 2 minutes from a JavaScript developer to code!

I’m a heavy user of Google mail/calendar/notebook etc. This is a part 1 on some of these Google products, and how I think they can offer a better user experience, with what I think are potentially low hanging fruit.

I have a bit more calendar feedback than others, hence devoting an entire post on it. So here goes:

Multiple reminders: be abled to have multiple reminders for each of my calendar events (eg. 1 week, 2 hours). NOTE: Seems this feature got implemented sometime in the last 2 weeks or so… Google can read my mind!
Repeat Whenever: Be able to repeat a event at $n$ future dates. For example festivals happen on different days (Chinese new year etc.) Not all events happen at regular intervals.
Reminders for non-default calendars: Be able to set reminders for events not on default calendar? NOTE: Even this got implemented as of May 3, 2007. Read here.
Events at non-day/time resolution: Putting events at the month/year/week resolution? For example year of the pig, month of excitement, week of joy? I can’t describe these events without making an event that has start and end dates of appropriate length. You can imagine the messy look on my calendar if I describe the year of the pig as a year long event!! This feature would be on the lines of an “All day” event.
Better repeats: I’d like the capability to be able to put events at the level of “last day of month / middle of the month” etc. I know last friday or the month is possible, but I don’t happen to be able to use it (yet). The repeat whenever feature I mention above is slightly different than this one.
Better reminders: Reminders at $n$ days before an event? Currently its maxed out at 1 week?
GTalk integration: How about showing me as busy on GTalk based on my calendar events, if I’ve marked myself as busy on my calendar? Not sure if everyone would want this, hence this could be an optional setting.

Hopefully someone on the Google Calendar is reading this (and my mind!)

UPDATE: How could I forget to request a J2ME app for Google Calendar, similar to Gmail Mobile. Am sure Google is working on this, given the popularity of their current mobile offerings!