myShoggoth

Seattle International Film Festival

SIFF is going on now through June 17th, don't miss the chance to see a variety of cool movies you won't easily get in a theater otherwise.

Minor TalentSpring site update

I just deployed a minor update to http://www.talentspring.com/ that has a bunch of bug fixes (66 tickets from our Trac system, to be exact).  This is our first real refresh to the site since launch, although we certain made a bunch of emergency fixes in the meantime.

What kinds of changes?  Well, lots of copy was fixed, the "download select_tab" bug is finally fixed, you can now successfully confirm email addresses, and we've changed the login block in the upper right to better explain what is your account name vs. what is your 'code name'.

Lots of little stuff, some that we knew about before launch but couldn't get to and others were found from user feedback (thank you!) after launch.

Lots more to come, please keep the feedback rolling in.  We're doing some major planning to address the bigger items that we need to fix up, but unfortunately it may be a few weeks before the results of that are visible on the site.

Remembering the Milk

Now that I'm refreshed and recharged after a nice long work-free weekend, I've been researching how to get control of my internal to-do list.  This is not nearly the first time this has happened, and I've never really found a satisfactory answer in the past.  Outlook tasks are... not good.  The pen and paper based organizer stuff was never my style.

For work items and bugs we use Trac at TalentSpring, and I'm very happy with that.  For things related to product scheduling, release load balancing, developer load balancing, and so on it is a great tool.

However, that's not the same thing as a to-do list.  Not everything I need to do for the business is a dev work item, so I'm looking for a way to organize my work and personal tasks that are outside the code jockey details of work items and bugs that Trac does well.

I just tried Backpack from 37signals, and by extension Ta-da List.  Nice tools, but too simple for what I need.  I know their philosophy is to keep things as simple as they possibly can, and I do like that, but it depends on slicing the market just right.

Let's say that you have a core piece of functionality that 90% of your overall target audience uses.  The simple thing is to do that core piece and leave it at that, right?  But how many of them use only that core piece?  If most of that 90% use that core piece plus something else, but that something else is only the same for 5% of that group each, you've now implemented something that isn't good enough for most of them but contains the core of what 90% of them want.

So you can say "We're focusing on what 90% of people want" but in reality, to really satisfy that 90%, you need to do a bunch of little 5% things to complete their scenarios.  The trick in that case might be to find the top N extra things that would help fully satisfy half of the 90%.

Anyway, the Backpack/Ta-da List stuff was nice and simple but too simple and I kept looking.  I found a TechCrunch article comparing the state of the industry for online to-do lists from a year ago and tried Remember the Milk, which so far is impressing me greatly.  They do some very nice subtle UI things that I want to study more and possibly steal (their login screen is particularly nice).  Not everything is perfect, they re-sort the task lists based on priority as you change them, which is a bad no-no.  After only a few minutes of adding new tasks I was already getting frustrated by clicking on the wrong item because things had shifted underneath me.

They have a little more functionality than I currently want, which I believe is ideal.  If they did just exactly what I currently want there's a good chance I'd want more quickly.

I'm going to stick with them a while and see how it goes.  So far, pretty damn slick.

Twenty Six Hours of Sleep

I'm taking a four day Memorial Day Weekend to try to recover from the push to launch and fixing of launch fallout.

So far that's two nights of 13 hours of sleep apiece.  I guess I've been a bit sleep deprived...

TalentSpring Voting

I've seen a number of comments and posts that make it clear that there is confusion about how the voting at TalentSpring works, here's my initial stab at clearing all of that up.  First I'll explain why the misconceptions are incorrect and then I'll talk in more detail about how it really works.

Misconception #1: TalentSpring is a Social Networking Site

I'm not sure how people got this idea, but we don't really have any social networking type features at this point.  It is true that members of the community vote on each other, but I think it is a stretch to consider that social networking.

The weird thing is that once some people decided to call it a social networking site, they then went off on a tangent about how social networking sites were so horrible and thus we must be as well because we're a social networking site...

Misconception #2: TalentSpring Uses Digg-Style Voting

A lot of excellent sites out there do their ranking by letting users cast a vote on an item, thumbs up or thumbs down, and items will bubble or or sink down based on those votes.  I think this is why people believe resumes on TalentSpring would become a popularity contest.

We don't do it that way, you don't say whether you like or dislike something, instead you compare two that should be comparable and say which one is better (even if marginally).

More importantly, users do not choose who they vote on, we have algorithms that calculate which votes are needed to properly order the job categories, and we do random sampling so users cannot game the system to make sure they get to vote on who they want to.

Misconception #3: I Can Vote People Down To Vote Myself Up

This is incorrect both for the reason I just mentioned (you don't choose who you vote on) and a more subtle point: You don't know how the decision you're making will affect you.  Which vote would help your placement depends on other people's votes that have been made as well as ones that have not yet been made.

The voting decision is between two resumes, you cannot "push one down" without lifting the other up, and without knowing the full relationship of all the existing members of the category as well as those that are in the process of being ranked you cannot know which one would help you out, or if it would affect you at all.

How It Really Works

Well, at least at a high level.

First, you enter either a mini-resume (the three accomplishments thing that allows you to try it out without going to the trouble of entering a full resume) or your full resume and request ranking.  We then have other users vote on your resume vs. others in the same job category.  Any individual user won't get to do more than one or two votes on the same resume, and it will take a number of users making redundant decisions before we decide on your real placement.

If users are disagreeing on votes we bring in those we have identified as more skilled (more often correct) to break ties.  We also have people do extra redundant votes in some cases to get more verification.  From all of that we can figure out who are better voters, which votes should be tossed, and which votes should be kept for ranking calculations.

Once you're placed with a specific version of your resume you can update the contents and get re-ranked with the new version.  Heck, if you think it was placed poorly you can request re-ranking with the same version.  We also have employees who scan the rankings and if they see someone clearly out of place they can force a re-ranking.

We want this to be a feedback loop that doesn't exist in the industry right now.  You get a bad score?  Look over your resume, look at those around you in the rankings, and those much higher.  Are they just that much more experienced, or are they describing themselves differently in a way that's getting them the votes more often?  Maybe your resume is full of typos or bad grammar.  Today you post that resume and the only ones who are going to tell you it needs work are your friends who take the time, not employers and certainly not other employees you're competing against.

As with the resume version post, please let me know if you have further questions on how it all works, your feedback is greatly appreciated.

Resume Versions on TalentSpring

I had an IM conversation with a friend where he made it clear that not only did we not make it clear what was going on with resume versions on the site (I knew that), but that lack of clarity was causing more frustration and angst than I had expected.

We need to both make it more clear in our UI as well as having some Help that explains it as well.  This post is not instead of those changes, but rather a chance for me to do a braindump on how things work so that those who make the UI changes and write the help can use it as a reference, plus a reference for those of you who are more technically minded and are curious.

So, why do we need versions?  The biggest reason is that when you get ranked on http://www.talentspring.com/, if we allowed people to make changes afterwards they could easily game the system.  A simple version would be to get ranked with incredible (and wrong) accomplishments or work history, then once you get a wonderful score you change the contents of the resume to be correct and never ask for a re-ranking.

More importantly, when people are voting they're voting one set of contents against another.  If the contents have changed that decision is no longer valid, which makes the rest of our calculations bad at best.

When you request ranking at TalentSpring, we freeze the contents of your resume and save it as a version, creating a new version that you can then edit.  People will vote on your resume versus others, but you cannot change the contents, all edits are to the new version.

You can only be ranked in a job category with one version of your resume, and you can only be in the process of being ranked in a job category with one version.  Once the new version is ranked, the old ones are removed.  The data remains, however, because the decisions people have made about whether the contents are better or worse than other contents are necessary to make correct calculations of the overall rankings in the job category.  However, votes on previous versions do not affect rankings for new versions, so you aren't at a disadvantage if you get ranked with less than ideal resume contents at some point.

We need to tell the power user for each job category:

• Are they ranked, and if so what is their score?
• Which version of their resume is the one that is ranked (and a way to view it).
• Is that version now out of date, that is has the user updated their resume since that version?
• Are they in mini-resume (the three accomplishments version) or full resume mode for that job category?

For the normal user, we want to tell them that they're out of date in a job category and that they should get re-ranked for it.  The same basic decisions but in a simpler and more action-oriented format to help those who don't necessarily know or care to know what is going on under the hood.

If you have any other questions about this topic let me know, I want to make it clear to our users what is going on and why.

Under 200 lbs.!

I have no clue how this happened under the circumstances (I'm a bigtime stress eater), but today I weighed in under 200 lbs. for the first time in a few years.

Enough of that, time to get back to ironing out the bugs in the site.

Too old for a startup?

Interesting article in the New York Times about the ages of successful startup founders, as you can imagine it caught my interest.

Unlike my previous startup in the late 90s (a small half dozen person consulting firm that made people money but didn't otherwise go anywhere), we have had to recruit and find people in a situation where our money is tight and a steady revenue was not expected for some time.  Hopefully it is sooner now that we've launched, but our beta is free for recruiters right now.

Talking to local developers especially, it is very difficult to get someone who is experienced, skilled, and willing to jump out of a job with a nice cushy salary to take a chance on a startup.  Often they have mortgages, families, are starting to use their medical benefits more often, and if they've been at the same company for more than five years they've built up a lot of skills that are purely useful in navigating the structure and politics of that company.

Someone moving from Microsoft to Google, or Amazon to Microsoft, etc. is not nearly as big of a transition.  They're all big, successful companies that are developer-oriented.  All have great benefits, all will pay competitive salaries, and all have very interesting problems to work on.

Enter a startup.  We went to a University of Washington career fair last fall to recruit, and our pitch was:  We can't tell you what we're doing or in what field or anything specific about the problems, but we have big interesting problems to work on and we promise it will be fun.  Also, we can't pay well but equity is part of our compensation package.  Uh, and no benefits yet.

The amazing thing is we got someone (Lincoln Ritter, the top notch developer who has damn near single handedly built a professional web UI from a sort-of working ugly skeleton I had built from scratch before he joined us) and he's turned out even better than the high expectations we had for him.  But it isn't an easy sell, and you know that the people who join are taking a big risk doing so.

But Lincoln had just finished his Masters and was starting his PhD program and felt like he was in a perfect place in his life to take the chance on a startup.  What about senior programmers who had a cushy deal at a large company?  The sell is much harder, and in a way you figure that the majority of people working at big companies are already self-selected to be more risk-averse.

They're in a big, safe company.  Chances are they wanted to be there and at least part of the reason was because they didn't want the ups and downs of startups, free-lance consulting, or small companies.

So why are so many more successful startup people young?  There's less downside and more upside for them, and I think that's a big part of it.  Another thing is that they don't have some bad experiences that have told them stuff isn't possible, when really it is just hard or not possible in the situation it was tried.  That doesn't mean they'll succeed, but they will try.  You can't succeed without trying, and if enough young people try some of them will succeed, and some will succeed huge ala Microsoft, Apple, Google, etc.

Launching with bugs

Well, we launched.

And with the launch we had a few bugs (and a ton of typos) that weren't caught beforehand.  I wish that weren't the case but I've done enough software releases to know that releasing something completely bug-free that is actually relevant in the marketplace isn't really possible.

Big bug #1: We were using a Ruby on Rails plugin for Google Analytics, and it was adding a dozen or two calls per page, including our AJAX requests.  I have no idea why the plugin was behaving so badly, and I'm very annoyed at myself for not finding the issue before launch.  The fix was to blow away the plugin and add a single call directly into our main layout template.

The drawback is that we don't get Analytics information on some of our pages that don't use the general page template, but we can fix that over time.  That change has made the site much faster and usable.

Big bug #2: Going to the main page is supposed to take you to a landing page where we can tell you about ourselves (http://www.talentspring.com/welcome), but due to aggressive Ruby on Rails action caching it wasn't happening.  We needed to take that logic out of the action itself and put it into a filter so that it would be run despite the action results being cached.

Generally speaking the caching is causing us a lot of problems of varying sizes.  From text showing when it shouldn't to at one point showing the completely wrong page, although it is giving us a huge perf improvement overall it is at a big cost to our sanity and complexity.

Yeah, the caching is going to be one of the hardest parts to get exactly right.

TalentSpring.com launches!

Finally, the startup I co-founded with Bryan Starbuck and have been working on for over half a year has launched our beta, please check us out at http://www.talentspring.com/!

TechCrunch just posted their writeup on what we're doing at http://www.techcrunch.com/2007/05/18/talentspring-aims-to-disrupt-resume-marketplace/, and I'm very excited.  Their review is very positive, I just hope we live up to the expectations.

I expect to be very busy for the next week as we handle all the assorted problems that we weren't able to account for beforehand.

If you have feedback on the site please leave it there or here, now that we're public and no longer in Stealth Mode I can't wait to talk about it all!