Automatically refreshed job board

Hi all

I posted a link to a Reddit convo for this project in my introduction, but if anyone’s game, I’d very much like some feedback here. I started a job board side-project a couple of years ago, and this year I’ve done about six months full-time on it. I now have to switch to something income-generating, so it’ll get incubated on the side for a while :smile:.

There’s quite a few job boards out there, so this is what (hopefully) sets mine apart:

  • Search engine robots operate in the background to retrieve job data every day, one for each employer, direct from employer’s careers pages
  • All jobs are geolocated and plotted on a map as a primary UI

It’s a tech board at present but if I can get some traction here, there’s no reason why it couldn’t be used in other professional areas (law, finance, etc) within any geolocatable area (I basically need a lat-long database for the area covered).

Fancy kicking the tyres on the prototype? The popup tab UI notes all sorts of interesting project info at the moment, for a fellow entrepreneur audience.

What you can see on that site probably represents 10% of the work that’s gone into the project - the bulk of the effort is the scraper engine and its administrative interface. It’s built so that a new scraper can be brought online in about 15 minutes. I tend to clone an existing one and tweak it using a point-and-click interface. I’ve got 40+ employers and 200+ jobs, all refreshed every day. The scrapers are pretty stable, though part of my play is that this will hold true for 1000 employers - that might be a different matter!

The project does not have a name yet (I have about 100 ideas on a Trello) and I need to do a complete frontend refresh (not my forte). It also needs a brand identity, which I’ll probably get someone else to do (definitely not my bag).

The purpose of the project was to scratch an itch a few years ago, when I saw that a lot of roles were being handed to recruiters, and thus the quality of online job search was very poor. I sense there has been some backlash against recruiters recently, with many smaller firms deciding to do their own acquisition process, and many niche sites popping up that cater for particular technologies (e.g. Rails), working formats (remote) or audiences (Stack Overflow). The quality of recruiter data IMO tends to be harmed in two ways: (1) duplication and error in feed systems that push to Monster, Indeed, CareerJet, etc, and (2) being unable to publish the employer or exact location for fear that the recruiter will be cut out of the deal (or they’ll be usurped by another recruiter).

The first two parts of my monetisation strategy is to use a plug-in ad partner initially, and then to offer manual/enhanced job postings for direct employers. Far in the distance, I envisage offering an API for the data, and a few other things.

For those of you who use the big recruitment sites (Monster etc mentioned earlier), do you see the data freshness as a convincing USP? Do you find being able to see office locations on the zoomable map as useful? (One great bit of feedback I’ve had so far is to add nearest rail stations to each job detail view - I think this is open data, so it would be quite easy to add). What attracts you to the job boards you use now? Do you prefer the “real employers only” format, or do recruiters offer you an advantage?

Any other feedback is welcome, of course.


1 Like

Speaking as a current user of job boards, the features and UX on them leave a LOT to be desired in my opinion! When I have some time (and if BT fix my phoneline this week so I have broadband again…) I will definitely have a look at this.

Splendid, thanks - good luck with the BT monster!

Bear in mind the UX on my prototype will all change - so let me know what features you think are most important. The search system needs to always be visible (it’s rather hidden at the moment) and I’ve yet to add in a feed from the scraper (to show it is active). I need to add location search and, well, a million other things really :smile:.

This is good timing, as Silicon Canal were talking on Monday about putting together a jobs board for use in the local area - I’ve directed them to this thread.

Great, thanks @LimeBlast.

I think the biggest problem with any on-line jobs board is the mistaken belief that managers in large companies have that all agencies provide some service beyond keyword-matching. If you can find a way of filtering out time-wasters who exaggerate their skill level on their CV then you could be onto a winner. Starting in a ‘startup community’ where the people employing techies are more likely to still be in touch seems a very good idea. Good luck.

I have worked at companies that will only deal with a limited number of approved agencies. I think the way to deal with that is to collect data.

Yes, I would imagine that would be a difficult problem to solve, and I’m not really trying to address that (yet, anyway :smile:). I think of my project as similar to Indeed (a fairly traditional aggregator) with Zoopla’s emphasis on location - with much better data quality and recency. I’m additionally solving the chicken and egg problem with boards too - this attempts to break out of the “no users no hirers” loop.

Clearly there’s demand for recruitment agency services, but my own experience of them is really only as an introducer. I spoke to one very nice woman today but she couldn’t answer my technical questions about a rather brief advert, and I’ll wager that when the key contact for the role calls me back tomorrow, he won’t be able to either. There’s no surprise here - they can only know so much about each hirer’s projects, and on top they are not technical either. By this stage if I’d called the hirer, we’d have a much better idea of whether we can do business - i.e. they’ll still have to do their own filtering.

Building online/offline communities is certainly one way of resolving the difficulty, as long as it doesn’t become a clique that makes it harder for industry entrants to get on board (everyone starts with zero recommendations!). I quite like Stack Overflow, as new people can start build up a profile easily, but this suffers from a rich-get-richer problem that has unfairly benefited early adopters (and it isn’t free of accusations of cliques either).

There’s some interesting job systems that do matching based on detailed job and candidate profiles (e.g. WorkShape) - but I don’t know how that could weed out skill exaggeration, really. It’s a toughie!

I was thinking of some sort of feedback loop from interviewers, weighted by how negative they usually are. Also feedback from interviewees on whether job descriptions were accurate, salaries exaggerated etc.

My experience of agents when I was contracting was that friendliness, efficiency, honesty and technical understanding were all to be valued and I didn’t meet many who had them all.

I’m looking at this purely from the view of a candidate, but location is a particular bugbear of mine. I am currently looking for jobs which are commutable from Coventry by public transport. So, areas of Birmingham/Solihull are ok, but some bits of Solihull are a pain to get to because of the way the buses run.

I can commute to Euston in an hour from Coventry station, but somewhere only a few miles from me as the crow flies is impossible if there’s no public transport nearby. Milton Keynes by train is fine; Leicester by train is awful.

I wouldn’t expect you to be able to do the donkey work in figuring out the transport links, but I would want to be able to input something that includes central London, or anything walking distance from Birmingham station, but excludes nearer less accessible locations. Job boards usually have a basic ‘within x miles of’ feature, or a picklist, or a free text field. I’ve been contacted by someone for a job in Newcastle because I’d typed in “north London” and her search had matched “north”!

That’s an interesting idea - a centralised record of employers and candidates, and a rating system for each. It might get tricksy from a libel perspective though - if one side feels the other has unfairly assessed them, and they believe that would hurt their future employment/hiring, can that sort of disagreement be resolved and still remain scalable?

Yes about (some) agents! Let me relate a scary story. Back when I was in my first badly remunerated job, fresh out of university and wet behind the ears, I finally left after five years. A large IT recruiter (we’ll call them Computer Fortunes to avoid any embarrassment) placed me successfully with a new firm. He asked if he could send me CVs to replace me in the old firm once I had gone.

I consulted with HR and they said there was no harm in receiving some, but we already had some CVs from another agency, and they would be seen first. Unfortunately the quality of both batches was low - the salary was terrible and it wasn’t attracting the right candidates. My recruiter phoned me often, which in my naivety I thought was just good post-sales support - in fact he was pumping me for information about the quality of CVs we were looking at, and using that information to bolster his position in price negotiations.

In the end HR came in and asked me to stop talking to him entirely, at which point I realised I’d been played like a fiddle. Oops! :violin:

Thanks @Judes… you may have saved me the hassle of applying for a contract role in Leicester, as it happens! (I calculate 65 minutes each way, plus 20 minutes cycling, plus whatever walk on the Leics side - and I don’t know where the role is exactly. This remote-working flexi-timer would have to leave the house at 7am. Ouch).

My first approach in tackling the problem you describe would be to allow the user to enter their home postcode (or use a draggable pin). All search results are then indicated as a distance from this point (the search options in the prototype show how this might work - draggable marker only atm).

Secondly I’d use open rail data to show the nearest stations either side. So, if you view a specific role, you’ll see a map with your home location, a couple of nearest stations, and the job location and a couple of nearest stations there. That’s quite easy to do, and is probably a good start.

Working out the easiest commuter routes would be interesting to look into, great idea. Google Maps can work out timings between stations, so there may be an API available. It’s just a cross-reference table of station A to station B (involving C changes and D minutes total). So it’s a question of getting that data, which might be available from the train operators as a feed. I’ll look into this, thanks!

Cool - sounds fascinating! Cov to Leics is a particular pain because you have to get a once-an-hour shuttle train to Nuneaton, which gets in a couple of minutes after the half-hourly train leaves for Leicester. Integrated public transport FTW!

Also, a job board has two completely different (and possibly conflicting, in some cases) functions - to work for recruiters and also for candidates. A really interesting project!

Thanks @Judes! Yes, different audiences are looking for different things. Seeker-recruiter tends to be quite oppositional since the seeker wants maximum searchability, but the recruiter’s business model is predicated on secrecy. (This is why recruiter involvement in this project won’t really work, for the first phase at least).

The seeker-hirer relationship is a bit less conflicted IMO, since both parties want to be put in touch, and both want maximum searchability and data openness. (Hirers seem also to want cloud HR systems these days to streamline the application process, but seekers don’t really care about that. I don’t plan to look at that, since getting search right is still quite a lot of work).

I’ve been sketching an improved front end over the last few days, so watch this space. :eyes:

Because my CV mentions my (currently very small) experience of PHP, I keep getting offered PHP dev jobs, presumably because recruiters just do a blanket search and fire off a load of emails without actually reading anything. Not sure that any search algorithm can fix that!

Yes, that’s one fundamental problem with the “email me jobs like these” feature. I think it’s fixable though - it would need to capture a detailed profile from the hirer and each seeker (e.g. 0-5 stars for various skills) to do good quality matching. I can’t really do that here though, since I’ve only a job description to go on - and parsing that into skill levels is a rather tough Comp Sci problem :smiley_cat:

If I can attract hirers directly (in a second project phase) then that would be possible.

Self-selected skill level is a horrible way of judging someone. I always rate my skill levels quite low, but then I’m very self-critical. Also, who am I rating myself against? If my mum is 0 (because she’s never used a computer), then that bumps up my relative skill level up a lot as oppose to if 0 is a beginner dev. And is 10 Douglas Crockford?

You could have the system learn what are the best matches by presenting you with, say, 5 options and getting you to say yes or no to them, so would learn that ‘PHP developer’ wasn’t a suitable match even though you have the word PHP in your CV. Like you say, this is a hard problem!

I’m not so sure about that. It’s easy to be self-deprecating (it’s rather English really, isn’t it?) but I don’t think CVs and so forth is the right place for it. I’m prone to imposter syndrome like the best of them, but self-assessing oneself fairly is actually good practice.

If we’re using a star system, 10 stars would generally be “expert level, can teach”. One star would be “have used a little in the past”. I agree it’s subjective, but it’s only a rough estimate to get an interview anyway - it doesn’t matter if your “real” score is a star above/below what you’ve given, as long as it is done honestly.

(Douglas Crockford can click on a special easter-egg that grants access to an 11-star rating! :star:)

Proudly sponsored by Bytemark