With Help Of Data Science Community, Adzuna Can Now Predict Salaries For Job Ads Missing Salary Info

Next Story

Bidgely Cleans Up $5M From Khosla Ventures To Make Household Energy Conservation Fun And Easy

Adzuna, which operates what it calls a “next-generation classifieds search engine”, with a heavy focus on the job ads market, is rolling out a pretty interesting feature in the U.K. today: ‘Jobsworth’ promises to predict the salary attached to any job ad in the search engine’s index when that crucial and decision-making information is missing.

The U.K. startup is also announcing that it “quietly raised” a new funding round, which I understand closed in April this year, bringing its total funding to £2 million. The new funds — just over £1 million — come from existing investors Index Ventures, The Accelerator Group, and Passion Capital, making this effectively a follow-on round after the same VCs funded its seed round in January 2012.

Co-founder Andrew Hunter says that the new capital will be used to fuel Adzuna’s international expansion. “We will be putting significant investment behind our core Adzuna product with particular focus on social job search, data and mobile,” he tells TechCrunch.

Adzuna’s job search works by indexing job vacancies found on classified ads and job boards around the web, and mining that data for its search engine and to provide free access to labour market trends, such as tracking salaries for a particular vocation and region. It’s also built a social layer on top to enable it to offer a feature called Adzuna Connect whereby users can sign in to Adzuna via Facebook or LinkedIn to leverage their social graph to help them get an in at a particular company with a vacancy.

Screen shot 2013-07-22 at 15.14.04However, of the 500,000 or so U.K. job ads that Adzuna indexes daily, about half of those don’t display a salary. That causes frustration for users who don’t know what salary to expect and can’t tell which jobs fit their aspirations. “Ultimately, not showing a salary results in poorer matching of candidates to employers and we want to improve this,” says Hunter.

Thus, with the help of the data science community, Jobsworth was born. It works by analysing each job ad in Adzuna’a index to understand current salary trends based on keywords, categories, companies and locations. The startup’s “neural network-based system” then produces a Jobsworth salary estimate for every ad with no advertised salary.

“One of our core goals as a business is to arm users with useful data to help them make better career decisions and solving the problem of missing salaries was at the top of the list,” says Hunter.

But, of course, pulling off something like Jobsworth, requires crunching a lot of data and some decent data science chops to turn the results into something reliable, including taking into account job title level, sector and location. It’s here where Hunter says Adzuna turned to the wider data science community. “We ended up running a competition on Kaggle to help us come up with the best possible algorithmic solution,” he says. “The competition had over 2,800 entries from some of the brightest data minds in the world, so picking through all of the great entries was fairly challenging too.”

As a result, the company says it evaluated over 100 different algorithms and machine-learning approaches to find the most accurate approach, claiming to be within 10% accuracy “the vast majority of the time – and getting better every day”.

Adzuna is also celebrating its 2 year anniversary this week. The startup, whose founders hail from Gumtree, Zoopla and Qype, now sees 2 million monthly unique visitors, indexing over 100 million classified ads. Since its original U.K. launch, its jobs search has expanded to 6 countries — adding Germany, Australia, South Africa, Canada, and Brazil — and revenue is said to be growing at 50% a quarter.