How I started a data analytics startup and partnered with SEEK.
I’ve learnt a lot from starting a company from nothing. Most of it has come from the experience of working in startups and talking with the founders, some has come from books. But there’s certain things you can only learn by doing — moments of the startup journey become lessons, forming a filter for your future decisions.
I’d like to tell you a story about my first startup. Presumi was a resume-tracking and analytics platform for online employment marketplaces. It used a unique algorithm that allowed for detailed resume tracking all the way from the candidate to the employer and back again. Built on top of this was a bespoke B2B (business to business) analytics platform that used data collected from resumes and behaviour to help large online employment marketplaces make strategic decisions around marketing and strategy.
It wasn’t always business-facing though. From late 2015 to early 2017, Presumi was a B2C (business to consumer) platform focused on helping thousands of job seekers apply for their dream job. We designed a platform for candidates to understand what was happening with their resume and become a better job seeker with a data-driven approach.
Strangely, I actually had to reread all the articles published about Presumi to piece together what happened along my own journey. The shit that happens on a daily basis when you’re running a startup means you don’t really keep track. This story will likely be a bit lengthy for your average blog post, but I hope we’ll both learn something by the end.
I started working on Presumi after getting back from a Product Design internship at Palantir. I’d spent the last 3 months of late 2015 in Palo Alto and came back to the life of a university student, applying for part-time roles to support my studies.
I’d been a job seeker since I was 15 and I knew the pain that people felt when they didn’t hear back from an employer. It’s not just “not getting a job” — there’s a perpetual self-deprecation that comes with not receiving any response. Armed with some design and coding skills, I ventured out to make something that would ease the pain of applying for multiple jobs concurrently.
I started by assembling a library that hacked up my resume, injected all sorts of tracking mechanisms then sent it off to an employer as a test. I was blown away with how accurate it was so I started adding more analytics into the “job application pipeline” and gave the algorithm a simple UI help me apply for jobs like Facebook, Twitter, etc. I hosted it a spare domain I had so I could use it while I was away from my computer and someone posted it on BetaList and ProductHunt. We got about 700 users overnight.
This was a huge achievement for me at the time — a product that I’d built in a week now had hundreds of people submitting ideas for new features and sending me bug reports.
I saw potential in what I was building so I decided to try my hand in Startmate. I was a bit late as the applications had been closed for about a month. I contacted Niki Scevak directly and asked for his advice. He invited me into the Blackbird Ventures offices to pitch to him directly so he could give me some direction.
Looking back, it was a terrible deck but a half-decent pitch. I only had 4 slides and not much content but I really gave a shit about what I was working on. Niki, however, seemed to enjoy the pitch and invited me to be a bonus at the end of the Startmate pitches so I could get some advice from a greater audience.
So, I went there and delivered my pitch, met some other founders and got lots of great feedback around focus, automation, market and growth. Alan Jones was nice enough to invite me around to his office a few weeks later to have a chat about where to go next. I told him where I was at and he gave me some great advice going forward.
I asked around the university about where to start looking and was directed to the Hatchery program. I eventually joined the UTS Hatchery+ to try and take it to the next level. I’d worked at plenty of startups previously, so I assumed that I already understood the fundamentals of running a startup.
But running a startup is very different from working in one. You’re no longer distanced from the decisions that affect the future of the company and the implications fall on you alone. Mostly, you’re stressed for cash and time. Hanging out with friends is something you’d gladly give up to get just one more user. But I wouldn’t trade that feeling for anything.
Everything evolved from there — on the technology front there were a few platform rebuild and redesigns. I learnt React, Babel, ES6, Webpack, etc. just to make the interface feel amazing. Now it’s my main tech stack. I became a much better Product Designer, and I learnt how to communicate with users, understand our target market and hire people. We even got published on ProductHunt again — officially this time.
I was also able to enlist the help of a very talented illustrator to help me define the new brand identity for Presumi. It was a good working relationship — she craved for a creative outlet and Presumi didn’t have a defined style yet. So, over a few months, our entire brand was completely revamped. We got a new logo, a huge stack of illustrations and a new colour palette to match.
We’d been featured on Startup Daily, The American Genius, KillerStartups, the UTS Newsroom and the Salesforce IQ blog. We’d been listed on Land-Book (and their best landing pages article), the SaasClub Newsletter, LaunchingNext, StartupList, The Tech Portal and LaunchTicker. We’d been written about on Outilsveille, AllMyFaves, TrendHunter, What’s New, Clases de Periodismo, Come Recommended, Ivy Exec and Social Geek. We’d been recommended as one of the 14 Sydney Startups to watch in 2017 by CurrencyFair, next to Airtasker. We’d even been talked about on NewsTalk Radio. Things were going very, very well in terms of press.
About a month later, a representative from UTS Careers approached me at a booth at the university Careers Fair, trying to compete with a startup next to us that were giving out free donuts. “If you can tell a student why they didn’t get a job based on insights and data analytics” he asked, “does that mean the university could find out why their students aren’t getting jobs?”
The thought hadn’t crossed my mind at all, but it was a fantastic idea that might solve our monetisation issues. In hindsight, I realise I’d grown up with an entrepreneurial mindset that was extremely limited and narrow. I told him that we were working on something for him already… then I went home and started building the platform.
After a couple of weeks, the result were pretty impressive. A platform for the university’s Careers Centre to measure their student’s success in applying for jobs on a massive scale. Using data analytics to isolate cohorts of students that are having real trouble applying for jobs (frequent rejections) and proactively helping them to engage with the Careers Centre.
After long periods with lack of communication and difficulty getting sign-off from middle-management, I decided to pull back on the project and focus my efforts on something else. Besides, having to convince every student in the university to apply through a single platform, using a single method… it doesn’t really work in the real world. Turns out that someone saying “I can see people using this” is not remotely close to “I would use this”.
Basically, it doesn’t work because we didn’t have job seekers and employers on the same platform. I considered it, but it’d just be a low-rate copy of an existing service like CareerOne or Indeed. And the reason I started this in the first place is because those services didn’t offer something as unique as what I had. If I was going to produce a platform, it had to be different.
So, there went our chance at monetisation and becoming a “real startup” — the ones that had monthly revenue in a co-working space and all that cool stuff. We weren’t going to revolutionise the university through a platform that monitors and improves student’s job application success rates but hey, there’s always another bug to fix until you figure it out.
Soon after the university situation had ended, I decided that Presumi should go hard on the B2C front. I’d left it stagnant for too long, looking for an opportunity to expand or monetise, that I’d forgotten marketing and growth. With good fortune, there were a lot of students around campus that had heard of Presumi and wanted to get involved in its development, so I sought to bring on a few interns to handle the things I didn’t have time to do.
After a couple of weeks of interviewing, I found four amazing people that I couldn’t wait to start working with. What they lacked in experience, they made up for in passion and intelligence. For students, that’s pretty damn rare.
During our recent growth period, a large Australian-based recruiting company approached us and asked about how flexible our technology stack was — whether there was any possibility that our technology could work with theirs to produce something more.
It meant steering away from what I’d always known — namely building beautiful user experiences backed by creative technology. This platform would be either a bespoke product for that company, or maybe just an API. I wasn’t sure how to feel initially since I’d be hands-off in the design of the experience my technology is providing. More importantly, focusing purely on B2B meant shutting down the Presumi website to focus on corporate partnerships. This would be a massive change for us and our users, especially considering it would happen overnight.
But after a lot of critical reflection, it made sense. It turns out that, in trying design a product for job seekers to use directly, I had produced an algorithm that solves a massive issue for the existing world. It could shed some light on the issues these companies were facing with visibility of the job application process, using big data analysis to help the company make decisions.
Working and learning in the Hatchery+ program had taught me that you need to “find your target market and nail it”. I never thought that providing other companies with the technology they need to combat 10 years of ambiguity in the job application process would be my niche. But, I believed that in order to impact people on a global scale, we need to focus on integrating our technology into existing, corporate ecosystems.
I didn’t close the B2C part of Presumi earlier because I was afraid of leaving nothing behind — like Presumi never existed and we never impacted the lives of thousands of people. But I’ve realised there are different ways of leaving things behind besides memories and footprints. In this case, it’s in the form of a story. If we can help these big companies understand how you apply for jobs and why your resume got rejected, they’ll be able to improve the odds of you getting a job.
That being said, welcome to the second chapter of the Presumi story. I hope you enjoyed the first part because this is where things get real interesting. For the next 9 months, I was working closely with SEEK — Australia’s no. 1 jobs, employment, career and recruitment site.
Working with SEEK, or any company really, wasn’t planned — I originally approached them with a request to use their API is for an experimental feature I wanted to run. In order to use the SEEK API, you have to get in touch with them and discuss becoming a partner, since the most important asset they have is owning job listings. The reply I got wasn’t awesome…
Thank you for your recent enquiry into integration opportunities with SEEK. […] We have reviewed your request and at this stage your application does not meet our requirements. Thank you once again for expressing interest in SEEK integration opportunities.
… but to be fair, I didn’t have much to offer in response for borrowing their competitive advantage. Not long after though, I received another email:
I hope you are well. I work with SEEK in the international strategy team. […] I came across your startup Presumi and would love to find out more. Would you have time for an informal chat early next week?
After a couple of conference calls, I was invited down to the SEEK headquarters in Melbourne for a chat about how we could work together. While I thought Presumi’s algorithm could be implemented in the main SEEK website, their international strategy team had other ideas in mind.
The problem they had was relatively complex: when a user wants to apply for a job on any job listings website, they upload their resume, find a job and send their resume to the employer through the ad. All user behaviour within this ecosystem is able to be tracked and you can bet it is. But once the resume is sent out, SEEK and the user are unable to see what’s happening on the hirers side. Beginning to see where this is going?
I was introduced to the SEEK Asia team from Hong Kong to discuss a new experiment — creating an analytics system around the Presumi algorithm and using it to process and analyse job applications for one of their partner companies, JobsDB. The hypothesis was that applying with a trackable resume is a concept that candidates prefer and hirers will still engage with, allowing them to capture more and better signals of reciprocated interest.
To prove this, we would need to validate a hypothesis from each side of the platform: that candidates prefer to use a trackable resume because they can get feedback and interest signals from hirers, and that hirers can access and review resumes easily and are willing to give feedback to candidates. This presented us with a range of questions we needed to answer:
To understand and quantify this detailed level of information on such a large scale, I’d have to build a new analytics system around the main Presumi algorithm that calculated things like the number of interactions per visit, hirer click-through rates and candidate dashboard engagement metrics.
For the period of the experiment, users would be offered an option on the JobsDB website when applying for a job. To go the “normal way” or to use the new, experimental “Status Tracker” feature. If the user opted in to the new feature, we would direct their request over to the Presumi algorithm to process their resume and send a trackable version to the hirer.
On the back-end, we skinned the Presumi “application” pages using the jobsDB UI Library so users wouldn’t know they were changing platforms. When users arrived on these pages, they would enter the same information we collected on the Presumi B2C website (name, email message, etc) but the hirer email address and resume were already pre-filled from their profile.
Once a user had submitted a job application to the hirer through the Presumi platform, we’d notify them whenever something happened with their resume — when the hirer received the email, opened it, said whether they were interested or not and send a reply to the candidate.
This platform ran in the back-end of JobsDB for about 6 months. The analytics from tens of thousands of applications poured in so quickly that we started to encounter performance issues. I’d never built something on this scale before and I’m less of an infrastructure developer than I’d like to admit so naturally there were a handful of crashes I had to deal with.
Eventually though, through working remotely with the Hong Kong team, I got the platform stabilised and had it happily chugging along for the rest of the duration of the experiment. A couple of months ago, we closed the funnel to the platform and the international strategy team sat down to decide whether Presumi was worth pursuing in the next financial year.
As it turns out, as keen as the international strategy team was, they didn’t quite have the budget to pursue it. With that, I decided to hang my hat up on the 18-month side project move on to explore something new.
This is the end of the Presumi story. It was a project I loved working on and I’m so happy for the opportunity to help tens of thousands of people apply for jobs. It was never about analytics, or resumes, or data, or insights… it was to bring you hope.