We Cracked Open Time-to-Fill and Found Some Weird Stuff

Apr 8, 2016
This article is part of a series called Wake-up Call.

Traditionally, the time-to-fill metric has been used as a scorecard for past performance. An organization may use time to fill to assess the performance of its internal team, or as a way to set service level agreements with outsourced recruiting partners. It has been used as a relatively blunt instrument for predicting success, as users apply historical data on simple time to fill per role to set goals for certain types of open reqs.

Today, however, we can better predict time to fill on a req from the day it is opened (and in the process, determine the probability that a req would exceed its target time to fill). We can also apply analytics technology to look behind the metric and reveal influencing factors. That means we can do much more than keep score and set goals. By understanding the factors, we can pinpoint areas of weakness, predict trends, and adjust our strategy to improve results.

In order to do this, we applied an approach called “survival analysis” to a significant pool of client recruiting data. Survival analysis is based on a process that enables doctors to predict, with increasing accuracy, the factors that most influence a patient’s ability to survive an illness. For example, we know that proper diet and exercise reduce the likelihood of heart disease. Along the way, strange things are also discovered, like the fact that marital status has a significant positive impact on cancer survival rates.

How does this relate to talent acquisition and recruiting operations? We adjusted the process to test the factors that contribute to recruiting — such as recruiter skills, location, time of year, type of position, and hiring manager involvement — and applied those factors to the history of positions being filled. Using this information, the process pinpoints best- and worst-case outcomes for speed and quality of result. This analysis can then be applied to help decision-makers make informed plans that maximize recruiting results in terms of speed and quality for filling new positions.

When we completed our survival analysis, our data scientists found many things that you would expect, and some strange things, too. The unexpected findings are the things that recruiter intuition doesn’t normally take into account. Here are a couple of odd findings that were particularly interesting:

Repetition Does Not Always Make You Recruit Better

We would think that a recruiter with the most experience would do best at hitting a target time-to-fill for a given req. In other words, the more you recruit for a certain role, the better you will be at filling that role on time. That’s not entirely true, but to see the variations we had to look beneath the obvious data.

The obvious data came from an overall analysis of total reqs versus the number of times a recruiter has seen that req in a particular location. Recruiters who had seen that type of req only one time hit their targets only about 40 percent of the time. Those who had seen it 2-4 times in a particular city fared slightly better; 5-10 times improved significantly; and 10+ times jumped significantly over that. No surprises here.


Now, we looked at performance based on job levels, with 0-3 representing lower level roles and 4 representing the more senior difficult-to-find roles. Once again, the 0-3 levels performed as expected. Recruiters who worked on that role 10+ times were most likely to hit their targets. The level 4 roles, however, showed a surprising variation. For these harder-to-fill roles, performance in hitting time-to-fill targets rose progressively for a frequency of 1-10 tries. For 10+ tries, however, results fell off significantly, with a larger percentage of time-to-fill targets being missed than for first-time recruiters on the role.


But wait, there must be some mistake, right? Nope. The truth is in the data. There are certainly many factors contributing to this result. For example, it is possible that higher-level positions require recruiters to develop and cultivate networks of people with those skills. The repetition and success in recruiting from that network may exhaust the supply, and the recruiter then has to rebuild. That’s one theory. In any case, it helps us understand what to expect, who to assign to a req, and how to better predict time to fill on the next opening.

Location and Time of Year

Testing this filter on other factors yielded more insight. For example, performance and frequency of attempt also varied by location. In Houston, we found an overall direct correlation between the number of attempts and improvements in targeting. In Phoenix we saw, once again, a drop-off in performance at 10+ attempts.

Time of year yielded some interesting results. Using historical data, we assigned a basic low-medium-high risk level to different reqs. Typically, we see a large number of reqs opened over the first quarter as new budgets arrive in many companies, but we also see a big jump in the risk factor. November may be quieter, but efficiency in hitting targets is much greater. We expected the opposite. Reqs over the holidays would intuitively take longer to fill. Efficiency in November seems like it should be lower, but it is not.

Weird stuff is why analysis is important. We may normally make assumptions based on common sense related to our business. If we use data to test those assumptions, we often learn that there is more to this business of recruiting than meets the eye.

This article is part of a series called Wake-up Call.
Get articles like this
in your inbox
Subscribe to our mailing list and get interesting articles about talent acquisition emailed weekly!