Matt Wallaert is Lead Scientist at GetRaised.com.
Most of the time, when we talk about “policy by the numbers,” we are talking about using data to identify the existence of a problem. But I want to tell a somewhat different kind of story: that numbers can be not just identifiers, but solutions.
The story still starts with identification. A few years ago, I was the Head of Product at Thrive, a personal financial management site (like Mint.com) that was sold to LendingTree. Our main job was to help people change their financial behaviors: spend less, save more. But when we looked at our data, we found women were falling far behind in savings simply because they made only about 75% of what men did. This meant that even if we created the best budget program imaginable, women would still fall behind unless we could raise their income level to that of men.
Enter data as the solution. We created GetRaised.com, a free, entirely data-driven product to help women ask for, and successfully get, raises. On the surface, the process is relatively simple: figure out if someone is underpaid, help them generate a letter to do something about it, track the progress and process around that letter, from handing it over to a manager to meeting to discuss.
But the true engine of change in GetRaised is the data that sits behind it. Figuring out if someone is underpaid means pulling from a mashup up of data from the US Bureau of Labor Statistics (BLS) and user-contributed data. Indeed, one of the hardest parts developing GetRaised was translating the Standard Occupational Classification (SOC) used by the BLS into job titles that match up with the titles that women are likely to enter on the site.
Generating a raise letter means more data. Besides the salary numbers, we scrape job postings, and match similar open jobs in a user’s market that pay a higher salary than what the user currently receives. And we used large survey data from our interviews with human resources professionals to figure out how much is the right amount to ask for in a raise. (Statistically speaking, 8% is the sweet spot.)
The headline value of GetRaised, of course, is not the math that drives it but the math that comes out: to date, approximately 70% of women who turn in a GetRaised Raise Request letter receive their raise, and the average raise is ~$6,500. But for data scientists and policy makers, perhaps more important GetRaised's success is that it exists at all.
As was well said at the recent DataGotham conference in New York City, analysis is useless without action. While the majority of data analysis in academia is used to drive larger policies, products themselves can be based on data, not just as methods of problem detection but as actual solutions. We need more systems that need increasingly smaller amounts of human intervention to introduce increasingly larger amounts of change. Indeed, I can imagine a GetRaised that already knows your job title and salary data, and that monitors your work output to quantify it in a raise request. Data can be more than just looking at the patterns and problems; data-driven solutions can be developed for problems we already know exist.
Post a Comment