We just shipped an update to Jobster.com that improves the quality and relevance of the job summaries displayed next to search hits; for example see the results of a search for search engine jobs in Seattle.

Google has trained users to expect hit summaries that are frequently fragmentary and hard to understand. A hit for “string theory” might include the summary “Interest in string theory is driven largely by the hope that it will prove to be a … It is not yet known whether string theory will be able to describe a …”

Vertical job search services, Jobster included, largely followed Google’s lead in their initial implementations of hit summaries. It’s a challenging problem to choose a few sentences that capture the essence of a long job description. Our next iteration is a good step forward but there’s clearly lots more we can do.

I believe that even for general web search it should be possible to greatly improve the quality of the summaries so that users have to do less pogo-sticking to find the results they want. One example is what Live is doing with their academic search. Notice how you can mouse over the hits on the Live results for search engine to view the title, abstracts, and authors of each hits.

When general web search summaries become good and comprehensive enough, it may not even be necessary in some cases to click through to the site hosting the content, which will result in increased tension between search engine and site authors and increased pressure to find ways to compensate content authors regardless of where their content is hosted.

In authentic conversations and recruiting, I asked whether the web couldn’t introduce a more real and meaningful conversation between employers and jobseekers.

Colin Kingsbury at HRM Direct replies with some well placed cautionary notes regarding the strong feelings inspired by hiring and firing, and the danger that cranks and negative commentary will come to dominate the conversation:

The problem here comes with the term meaningful. It is surprising how difficult it is to find out what your customers really think of you, whether you have five of them or five million. To the extent that “social computing” techniques help draw authentic and unfiltered customer opinion out, they will help businesses to do better. The problem is that many of the critics you may find yourelf engaging are not really honest brokers…

Recruiting is going to encounter an especially large challenge here because like dating, it is a process of rejecting people. No matter how nicely you do it, some people are going to take it badly, and a few of them are going to make it their life’s mission to cause you as much pain as you caused them. Unfortunately, it’s precisely these kinds of critics that take the most time and energy to deal with.

Of course, someone who got rejected for a job at Morgan Stanley has always had the right to carry a sandwich board on the sidewalk in front of the building and hand out leaflets. But this took energy, and reached very few people. With social computing, the gadflies can reach a global audience from the comfort of their sofas.

And contra Ms. Li, I think bringing these sorts of critics into your own forum lends them a credibility they might otherwise lack. Today any crank with an axe to grind can lash out at TGI Fridays on his blog and have it come up page one of a Google search for “work TGI Fridays”. But, the casual web browser will also play a little game of “consider the source” and perhaps conclude, “this guy is a crank.”**

To wit, MySpace and Blogger are like the sidewalk, and you can’t legally shut up someone who is determined to make a scene there. But, should you invite them into the lobby and offer them a refreshing beverage? And don’t forget, when you ask them to leave, all their friends may show up to join the protest. After all, it’s certainly not your best interests they care most about.

Is there no way to sustain a forum that provides value in the middle ground between “Careers at ShinyHappyCo” and f***edcompany.com? Perhaps unrealistically positive and depressingly negative forums about companies predominate because it’s so difficult to create a balanced forum.

Assuming a balanced employment forum could be created, would companies be better off encouraging participation, or attempting to limit employees to officially controlled spaces only?

Colin draws an interesting comparison with online dating services, which face similar issues:

If you really want to see where this is headed, I would keep an eye on the dating services. They are well ahead of the recruiting space in terms of sophistication in these areas, and the issues are very similar.

Sounds like it’s worth checking out the dating sites (with appropriate disclaimers in advance to our spouses, of course!)

Laurel has a
insightful comparison of the different software development philosophies of Joel on Software and 37 Signals.

In a nutshell, Joel says programmers are most productive in a splendid bubble of specialization and isolation (”A programmer is most productive with a quiet private office… a tester to find the bugs they just can’t see, a graphic designer to make their screens beautiful, a team of marketing people to make the masses want their products, a team of sales people to make sure the masses can get these products”, etc, etc.)

37 signals, in contrast, says developers are most effective when roles are blurred (”When the edges are blurred, and one thing is many things, you can achieve so much more with less time, effort, and people.”)

I love startups precisely because of the chance to blur roles and to experience first hand the end of end process of understanding and meeting the needs of users.

At Jobster, as Laurel explains, we are working to retain this level of ownership and experience even as we grow from a small company where everyone has to do everything to a medium sized company where’s the luxury of specialization is possible:

At Jobster, in the past few months or so, we’ve been gradually growing our development team past the point where it makes sense for everyone to work on everything. At this point, we could decide to slice things horizontally – define everyone’s roles more strictly, have core developers, library developers, project managers, product managers, program managers, UI developers, UI designers, graphic designers… But we’re taking the approach of slicing vertically. Spinning off parts of our product into independent chunks worked on by independent teams. This wouldn’t work if everyone was stuck in their roles. We simply don’t have enough people to fill a small independent team with specialists. Sure, I’m not a good UI designer. But our good UI designer is busy with other projects right now, and we can’t hire another one before I want to demo my project next month. So this project won’t work if the 3 of us “software developers” can’t come up with a UI design that’s adequate enough that the decision makers get what we’re trying to do and believe in it enough to give ours priority over the designer’s other projects.

At a medium sized company like Jobster instead of a really small one like 37signals, we have the luxury of having people who specialize in certain areas (we certainly have a lot of sales people who call people at Fortune x00 companies and try to get them to sign big contracts). But we’re not big enough to be able to move people around efficiently. This is an opportunity and a challenge – at a small company people have to do everything; at a large company people can’t do everything. At our size we could do either, and we have to figure out what makes sense both for the group and for each individual.

Alan Steele continues the thread in it’s not all about productivity:

This is why optimizing for the ability to type ’svn commit’ makes no sense at all. If you break down a software project by elapsed time, it usually looks like this: 70% figuring out what the problem is, how best to solve it elegantly, efficiently and in a manner that delights the customer, and getting everyone to agree that this is the best way forward; 3% writing the core code; 27% getting the code you just wrote to work as intended, fleshing out the supplementary features and closing in down with a minimum of bugs.

Not only does the up-front part represent the vast majority of elapsed time, it’s also the part that affects the end-result most dramatically. And it benefits enormously from people talking to each other, which is rather more difficult when those people are ensconced in their private, temperature-controlled offices hooked up to their caffeinated carbonated beverage IVs.

If you don’t believe me, try this: go buy a really, really nice new computer; load it up with the newest bad-ass development tools on the planet; learn how to program (ok, this part could take a little while); and then write some great software. Right between steps 3 and 4, you’ll find yourself facing something that feels remarkably similar to writer’s block. In fact, it’s pretty much the same thing as writer’s block, and it happens to organizations just as it does to individuals.

The good news is that designing software, unlike writing, benefits enormously from having a small team of people working together to overcome that block. Getting a few people together to think about a problem produces far superior results than putting someone in an office and saying “Think!”, particularly if each has a slightly different take on the problem (say, one with more of a business/customer focus, another with a design/usability focus, and a few techies). The myth of the lone wolf programmer is crap: great software is built by teams of people. The reason Excel is such a kicking piece of software is that literally hundreds of people have worked collaboratively for many years to make it great. Software development is a team sport.

The secondary argument is simply against Joel’s attitude of coddling his programmers to the ridiculous degree of comparing the rest of the company to the servants of the Roman army. This kind of attitude doesn’t do much for collaboration with those supposed servants, many of whom have very good ideas about how the software should be built. It also encourages a distasteful kind of laziness where developers expect everything brought to them on a silver platter. Any software manager who has witnessed the electrifying effect of a customer visit on a previously insulated developer knows what I’m talking about: there is nothing so powerful as getting out in the world, watching and talking to people to learn what is actually needed from your software. It’s also a lot more fun that way.

There is one sentence of Joel’s recent article that I agree with, which is that an abstraction layer is needed between development and the rest of the organization. But he’s got it backwards: a software manager needs to create for the rest of the organization an abstraction (more like an illusion, really) that makes product development look like a predictable shipping machine, producing regular deliveries of software containing bright new innovations, when the reality behind the scenes is considerably messier. Otherwise, it gets very hard to justify the fancy computers, comfortable salaries and free soft drinks…

I feel lucky to be part of an industry new and vital enough that it’s still possible, meaningful, and impactful to have conversations like these.

Jobster is unusual (or perhaps not, I really don’t know) in being a company that does extensive development in both Rails and J2EE/Spring/Hibernate, and that has developers who move fluidly between the two worlds. (If your company has experience with both worlds, we’d be interested in sharing learnings.)

Our flagship employer application is built on Spring, and a new consumer application (now in development and soon to be released in a limited beta) is written on Rails. The consumer application was developed rapidly by a small team of developers and is one that we intend to iterate rapidly on.

There is a tendency in most companies, I think, to believe that the whole company needs to standardize on a single technology stack, or at least that specific developers need to specialize.

Given the very different audiences and requirements of the two major applications we are building, I think the “one stack to rule them all” argument is false. For the end user application, rapid iteration and learning is essential, while we have to be more deliberate and careful about how we change the experience for recruiters using the application. Conversely, our employer application plays more to the strength of J2EE through requirements for things like multi-column primary keys, distributed transactions, and so forth.

Our experience has been that a team of strong developers grounded in the MVC model of Spring finds the transition to Ruby and Rails easy and enjoyable, and that they do enjoy greater productivity. We’re very excited about what Rails allows us to do.

Seattle Mind Camp 2.0, a “a self-organizing, digitally minded, entrepreneur-driven, overnight Seattle confab” will be April 29-30th.  I had a lot of fun talking with Andru Edwards, the founder of the event, on the Gear Live podcast.
I’ll be there this year along with Charles Porter who works on search and search  relevance at Jobster.

Tickets are sold out already, apparently, but if you’re reading this and planning to go, please leave me a note; I’m looking forward to meeting new folks.

One of the things I enjoy about working at Jobster is that there are so many interesting problems on our roadmap, far too many in fact for us to accomplish given the current size of our development team. If you’re someone who is passionate about working on interesting problems in areas like search, we’d like to talk.
Two of these interesting problems are job and people search. I’ve written on job search before, so my focus here is people search.

The internet is rapidly becoming a place that holds a vast amount of meaningful information about people (mostly early adopters these days.)

But when it comes to being able to search that information across the entire internet (vs. a walled garden) and return people that are relevant to a particular intention, for example hiring, internet search services have only begun to scratch the surface of what could be accomplished.

Resume search is an interesting and easier subset of the full people search problem. For the relatively small number of people who post resumes online, it’s a useful but far from exhaustive subset of useful information about that person.

The first and easiest step is finding pages that are likely to contain resumes. Rather than simply crawling sites that are known to contain resumes (like thejobspider.com), a fairly simple full-text query suffices to return a set of pages that with pretty good likelihood contain resumes.

I’ve put together an extremely simple resume search demo that shows this approach.

You’ll notice that I used Windows Live Search rather than Google for this simple demo.

[It’s interesting to note that the sophistication of Google’s query language is lagging behind some of the competition; consider for example the prefer: qualifier and MSN Live Search Macros . This is probably because end users, by and large, do not use sophisticated queries and because Google has shown mixed interest in being a search platform that others can build value on top of.
I think the experiments Windows Live is doing to replacing paging with an Ajax scrollbar are pretty interesting– the place I would really love to see this functionality in web email inbox.]

Thanks to services like the Alexa Web Search Platform plus their search API, this first stage can leverage comprehensive full text indexes of the web that have already been constructed.

The next, more interesting step is crawling the hits, extracting and indexing structured or semi-structured data from those pages, and using that data to improve the relevance, presentation, and searchability of the data.

Ziggs extracts names, locations, and companies, but this information alone is not enough to determine with a resume is relevant and a lot of pogo-sticking is required to interpret the result.

Pagebites improves on Ziggs by also extracting education, objective, and skill fields and promoting those in the search results.

In both cases relevance remains a challenge. Neither understands that nursing is a profession and that the search results should favor actual nurses over congressmen who happened to include nursing as a keyword.

Both Ziggs and Pagebites rely primarily on resumes. An alternative approach is seen in ZoomInfo which crawls sources like SEC filings and press releases and uses natural language parsing to attempt to extract and collate a virtual bio of the person involved from multiple sources. (See for instance this automatically constructed bio of Guy Kawasaki.) The limitations of this approach are currently accuracy– the natural language technology is not perfect– and comprehensiveness– many people do not appear in press releases or SEC filings. The open ZoomInfo service only allows searching by name, so it’s difficult to evaluate relevance, but in principle they should be able to good things with relevance.

As I was saying, all of these sites just begin to scratch the surface of constructing a complete, accurate, and easily searchable picture of someone based on their online presence. If this is the sort of problem that attracts you, please get in touch with me at the email address listed on this blog, or check out labs.jobster.com/jobs.html.

Jobster Trends is a new feature that gives you a look at the who, where, and what of job search. It includes the most popular searches, biggest gainers and losers, as well as the most popular job search locations.

Combine this data with information about what jobs are being posted where, as Jobster obtains by crawling the web, and you start to have a pretty interesting map of supply vs. demand across the country for different types of positions.

We look forward to presenting more of this data over time; it’s a pretty fascinating topic. Eventually, I can imagine economists using search data as one of the things they factor into their forecast.

Jobster and Six Apart are pleased to announce the new Jobster widgets for Typepad blog authors, which make it easy for any blog author to provide relevant jobs to their readers.  To promote the new widgets, we are including 400 chances to win free Typepad service. These widgets reflect Jobster’s philosophy of enabling a more distributed, social approach to finding and advertising jobs.

(If you’re a Typepad user, you should also check out some of the other new widgets in the Typepad Widget Gallery)

sample job feed widget

Below are the full details from Jason’s post on the Jobster Blog.

---

Typepad is the most powerful yet easy to use blogging service around.  period.

hundreds of thousands of individuals — like myself — who can’t write a single line of computer code to save their life, rely on typepad to power their blogs.  typepad has made it so easy that anyone can do it.

today, jobster and typepad are teaming up to enable typepad bloggers to easily add a sidebar with relevant jobs and/or a jobster search box to their blogs.  (read more about this exciting news) 

think of it as an important first step in “distributed job search.”  we take the most contextually relevant job listings where the readers are vs. expecting them to go to a job board.  as i like to say, in this day and age, that great web services developer (for instance) is more likely to be reading a blog on a topic like “ruby on rails” than they are to be searching on a job board on a given day, so we need to get our jobs where they are.

Blogging about recruiting?  add a jobster feed of recruiting jobs to provide useful content to your readership, kinda like jason davis does at recruiting.com (see right column sidebar).

Blogging about marketing?  add a jobster feed of marketing jobs to provide useful content to your readership, kinda like  harry joiner, the marketing headhunter is doing.

but wait, there’s more.

FREE TYPEPAD FOR A YEAR!  As part of this launch, Jobster is offering 400 individuals free typepad service for a year.  If you are already a typepad user, all you need to do is install a jobster widget on your typepad blog before may 25th, check a box, and you can win free typead blogging for a year.  that’s right, jobster will pay for your typepad subscription for a year. 

what, you don’t have a typepad blog yet?  click here to get a typepad account and then add the jobster widget to your blog , enter then enter our “free blogging” promotion, so we can start paying your blogging fees for you.

it’s easy!

pretty cool, huh?

Michael Cote has some good thinking on what a professional might want out of a “platform” for his or her online professional identity.

Some of the ideas include:

• An RSS feed of events in your network
• A JSON or RSS feed of endorsements you could include in your blog.
• the ability to export your profile as PDF or XML.
• support for microformats in the public profile pages.
• support for FOAF for contact lists.
• integration and synchronization across different social networking sites.

We’re heading for a world in which users will no longer tolerate having their profile data locked up in walled gardens. There is something compelling about a site which actively helps users make their profiles available in the places and formats they want to.

Mark Maunder is once again living the life of an early stage entrepreneur. I’ve enjoyed working with him and will be watching his next venture with interest.