littlelogs

Keep a social journal of your work progress as you make and learn things.

February 2017

peter
peter

Running some analysis on the current #talenthunch data, and teaching myself how to use D3.js at the same time.

This graph shows the candidates (red) and their selected skills (blue). The bottom left cluster of nodes are all software related, which demonstrates an interesting possible niche market idea for TalentHunch.

peter
peter

Spent the last few days trying to refactor the main #talenthunch job match creation script because it was taking 6 hours to run (yes, you read that right. 6 hours). After a lot of false starts and testing locally I think I’ve finally cracked it and hopefully it will now run quicker. Will find out tomorrow when the cron job runs next.

peter
peter

Sent out 2 #talenthunch emails this morning:

  • A notification to the 4 candidates that haven’t added any skills to their profile yet about how TalentHunch works (i.e. no skills = no matches)
  • A welcome email to all the new signups in Jan to say hi and asking them to share TalentHunch with their job hunting friends.
peter
peter

#talenthunch January 2016 stats

Business stats

  • 25 registered candidates
  • 25 new candidate signups
  • 133 skills added
  • 26 previous positions added
  • 14,332 job matches generated (220.50% of estimated 6,500)
  • 9,650 unique job matches (52.16% coverage of all jobs added in January)
  • 18,501 new job listings (68.52% of estimated 27,000)

Site stats

  • 154 uniques to landing page (signup conversion rate was 16.23%)
  • Top traffic sources: Direct, Hacker News, Twitter, organic search, Facebook

January 2017

peter
peter

I think I’ve finally worked out one way to increase #talenthunch candidate registrations; allow unregistered users to search for jobs from the homepage, but limit the search results shown to only 10. This, combined with some creative call to actions, should hopefully incentivise more candidates to register to see more search results (and generate their matches).

The quicker I can get more candidates onto the platform, then the more motivation advertisers will have to post on TalentHunch (when this functionality is complete).

peter
peter

Finished the #talenthunch password reset functionality today and will hopefully deploy it tonight or early tomorrow morning.

Stats for the first week:

  • Unique homepage views: 85
  • New candidates: 13
  • Total skills added: 45
  • Unique skills added: 36
  • Total positions added: 12
  • Total job matches: 2,402
  • Total jobs added: 5,206
peter
peter

New #talenthunch homepage is up. Still need to work out an easy/cheap way to drive more traffic to the site.

Catching up with a few HR/recruitment advisors in the coming weeks to pick their brains and demo the site to them to get their feedback.

peter
peter

Hit 1,065 matches from 2,315 jobs on #talenthunch overnight! #milestone

Other stats so far

  • Unique homepage views: 66
  • New candidate registrations: 9
  • Total skills added: 37
  • Unique (grouped by name) skills added: 30
  • Highest count for the same unique skill: 3
  • Positions added (work experience): 12
peter
peter

Was trying to hold off on today’s log because I’m really close to hitting the 1,000 #talenthunch job matches mark (and wanted to use the milestone tag), but alas it’s sitting on 980 matches.

I worked on revising the homepage design so it’s more descriptive, includes a product screenshot, and also has an FAQs section to answer questions like: “Why do you need my full name?”. Will hopefully push this live tomorrow.

peter
peter

#talenthunch stats for the first 24 hours:

  • Unique homepage views: 47
  • New candidate registrations: 6
  • Unique skills added: 19
  • Highest count for the same unique skill: 3
  • Positions added (work experience): 10
  • Job matches made: 320 (from 778 new jobs)
peter
peter

Finishing off the #talenthunch Edit Profile page and then turning my attention towards the Forgotten Password functionality (I didn’t make this part of the Outduo MVP and it came back to bite me). Not long until launch now.

December 2016

peter
peter

Just finished the final (yes, actually the final this time) candidate to job matching code for #talenthunch today. I can can successfully match a list of my skills and previous job titles to a database of jobs, and then return those matches with their associated match strength score.

Finishing off the site’s front-end tonight and tomorrow with the goal to go live before the end of the year (yes, I know, it’s another deadline but anything is possible).

peter
peter

A very late #laterlog:

  • My wisdom tooth operation went well.
  • Achieved very little #talenthunch dev during my work leave/recovery time.
  • Last Saturday I moved into the house I’ll be house/cat sitting for the next 5 weeks (within walking, instead of driving, distance to the CBD).
  • Achieved a good chunk of TalentHunch dev in November, but still a bit to do.
  • Will be finishing up at my real job on the 21st December, so will need to find something to do next year (which might be TalentHunch, depending on the timing).

November 2016

peter
peter

Last day at my real job before I go on unpaid leave for a couple of weeks (wisdom tooth extraction, as bad as it sounds). Will try to get some #talenthunch dev done while I’m recovering, but my code quality may decrease.

October 2016

peter
peter

I’ve set myself a deadline in November for completing the next stage of #talenthunch.

Via the new API, I can now successfully register and login a user, and return their job matches (based on the information in their profile and previously saved job matches).

peter
peter

It’s a public holiday in southern Tasmania tomorrow, so I’m hoping to get some solid #talenthunch development time in. I think my release of the Jobs site recently has made me more motivated to get back into this project.

peter
peter

I deployed my latest side project #talenthunch Jobs this morning. Currently it’s just indexing a small percentage of Seek and Indeed jobs, but I will be adding other sources over time (including direct advertisers).

September 2016

peter
peter

Have been continuing to build the #talenthunch job search engine component this week. Have successfully worked out how to index jobs from GitHub Jobs, Indeed and Seek.

I wanted to get more done before I start my new job tomorrow, but I think I might just have to leave it for a bit unfortunately. #laterlog

peter
peter

Now that I’ve signed the paperwork I can almost make it official - I’ll be starting a 3 month contract on Monday with the Beacon Foundation in Hobart, Tasmania. I’ll be doing a whole lot of professional web development again, after an eight month break.

I’m sure I’ve rallied against ever getting another real job previously, but this is a fantastic organisation to work for, with a cause and mission I believe in 100%, and so it would have been very silly to turn down this opportunity.

I’ll still be continuing to work on both #outduo and #talenthunch on weekends and in other spare time, but they’ll no longer be my primary focus.

August 2016

peter
peter

Spent today getting back into my #talenthunch mindset.

  • Decided that the product, for now, will just be a job search engine that sources jobs from third parties. I will then need to gain a lot of user traction to be able to persuade companies to advertise their positions on talenthunch.com
  • Once this core job search works, I can add some more layers of functionality in the future, such as an aptitude questionnaire, to produce more relevant job matches for users
  • SSL cert installation and server hardening is underway
  • Activated the domain on my Mailgun account for sending job alerts
peter
peter

After my slump report earlier today, I actually managed to get a bit more into the groove with my #talenthunch dev. I think talking it out actually helped.

Finished a major set of functions that saves a user’s answer and returns the next question based on that answer. Using the methods I made the user will now only have to answer a minimum of 8 questions from the same aptitude category in the affirmative before they can see their results.

July 2016

peter
peter

While trying to configure the #talenthunch server, I remember that I wasn’t born to be a sysadmin. I always get far enough to break it, but then not know how to fix it again.

Called in my local friendly sysadmin (AKA “the guy”) to assist and he’s taking a look at it now.

peter
peter

Spent Friday evening writing out the questions for #talenthunch.

To get a more accurate understanding of the user’s job interests I’ve increased the number of questions from 35 to 56 (7 interest areas with 8 questions each).

Originally I’d planned that the questionnaire would assess the skill aptitude of users, but it has been easier/faster to write questions to assess interest rather than skill level, for now.

June 2016

April 2016

peter
peter

Having a small break from #outduo dev by spending this week mostly focused on #talenthunch. I’ve finally worked out how to convert user’s question responses into meaningful data that can be matched to a list of occupations.

In the process of drafting 56 questions, across 7 aptitude areas (such as Managerial and Mathematics), which will make up the core of the matching algorithm.

March 2016

peter
peter

Now that Outduo is in beta I can spread my time across some other work. Looking at my #talenthunch code again today to work out what the next stage is for this product.

I did have an interesting meeting with one of NZ’s main HR pre-screening software resellers a couple of weeks ago. Their product is incredibly advanced, and produces detailed reports for candidates, but takes up to 30-45 minutes for the candidate to complete.

Although Talent Hunch isn’t as advanced, I’m hoping that I can bring the maximum assessment time down to under 15 minutes.

November 2015

peter
peter

Renewed my #talenthunch domains today. I really like the concept, but the execution (and monetization method) are always lacking. Oh well, I have another 2 years before I have to think about it again.

September 2015

peter
peter

The biggest problem with #talenthunch right now is that it's too slow. Users are not going to want to wait for at least a minute, or longer, for their results to be displayed (even if those results are spot on!). Need to reduce the number of search queries hitting my source databases by filtering out keywords/phrases that are very unlikely to return any job titles. More testing to go.

Older