How Visier Works: Predicting Time-to-Fill For a Job
This post is part of our Data-Driven Talent Acquisition Strategies series.
One question every recruiter faces is: “How long will it take to fill this position?” Seasoned recruiters have had long practice in figuring out how to deflect this question by speaking of the tough hiring market, sharing insight on the current candidate pool, or turning the question back to validate just how important timeliness really is for a specific hire.
But what if you could turn to an executive and show them the result of a predictive algorithm that had analyzed all your hiring data and come up with an answer? One that provides a given likelihood, or confidence, of making a hire within a defined timeframe. For many Talent Acquisition leaders, this would be almost like the holy grail. With this information, they could better plan recruiting activities and staffing and help the business make more successful business plans through optimized staffing.
Unfortunately, there are so many details and moving pieces when it comes to hiring that trying to analyze and predict hiring timelines has been relegated by most to the “too hard” bucket. To make reliable predictions, you need to analyze historical trends, current pipeline activity, hiring targets, applicant flows by source, and more. Even if you had the analytic capability, you first have to source this information from many different systems and the data extraction and integration is enough to keep most from trying.
However, with Visier, it becomes much easier to accurately forecast how long it will take to fill a specific role with a quality hire. Below is how Visier works.
Visier’s approach uses a combination of predictive analytics and an in-memory analytic engine to deliver trustworthy time-to-fill predictions in real time. The pipeline forecast feature of Visier Talent Acquisition calculates historical conversion rates and transition time distribution between recruitment stages to project time-to-fill at a set confidence level.
Visier’s forecast is based on four observations that are all measured from your historical data:
- Conversion rates: The ratio of applicants that move from one stage in the recruiting process* to a specific other stage.
- Duration in stage distributions: The list of times that applicants remained in a stage before moving to a specific other stage.
- Rate of new applications: The number of new applicants entering the first stage per day.
- Currently active applicants in each stage: The number of applicants currently in each stage at the moment the analytic is run.
*Note: Visier supports multiple recruiting processes, each of which might have its own stages. For instance, many large organizations will have a standard process as well as a campus hiring process. As a result, the algorithm must take into account the recruiting process for the role or group of roles being considered.
The forecast then proceeds by calculating the movement of the current applicants through the recruiting pipeline using these measures and inserting simulations for any randomness that occurs in the hiring process.
Using this process for all applicants and all stages, Visier simulates the movement in the pipeline in a defined time range, repeating it many times. Visier gives every applicant a large number of chances to progress to higher stages (based on the historic trends) and calculates different duration times between the transitions (again, based on the historic trends). The forecast results are then calculated from the set of simulated outcomes.
Since providing a single answer is unreliable and sets unrealistic expectations, Visier runs the calculations at three different confidence levels. Ian Cook, Visier’s Director of Product Management, does a great job explaining this process in the video below:
How Much Risk is Too Much Risk?
As any Data Scientist will agree, the accuracy of a forecast depends on:
- The number of simulations that were available to calculate the forecast results
- The quantity and relevance of historical data (in the case of hiring forecasts, the quantity and relevance of historical data used to calculate the conversion rates and duration time distributions)
Put simply, a large number of simulations are necessary to generate high quality forecast results. The forecast must be based on historical recruitment data, and results can only be considered reliable if the historical dataset of movement through the pipeline is large enough to allow insights on realistic conversion rates.
Visier automatically determines if there is sufficient historical data and in addition, more advanced users can specify preferences on the risk they are willing to accept to the accuracy in a worst case scenario.
Of course, the historical data used to analyze the pipeline movement has to be relevant with respect to the open positions that the forecast is generated for. For instance, recruitment data for Sales Executives can not be used to generate reliable forecasts for the procurement of Maintenance Engineers.
This is self-evident and Visier allows the user to focus the analysis on the relevant section of the historical data by defining a specific context — information that describes the conditions to consider such as roles, timeframes, geographic regions, etc. I’ve mentioned previously that predictive analytics can’t replace human intervention as there are details that only a human being can connect back to data in order to bring out the true insights.
Predictive Analytics For Everyone
Every day, we all use technology that, if we were challenged, we could not fully explain nor develop ourselves. From the microwave oven to the the apps on our phones, these technologies provide us with tremendous value — even if we only know how to apply them.
Data science and predictive analytics are hardly different. When done right, they become indispensable in providing us with insights that can make us better. While we may not be able to do the statistics or data management to recreate these solutions ourselves, we can apply the insight to shorten hiring time or increase the quality of our candidate pool. Doing data science and predictive analytics right means empowering as many users with the advantages they bring.
Everyone in your HR department needs the insights from analytics, and the real mission your HR department should be taking on is empowering all people leaders to make them better at hiring, managing, and leading the workforce.
True self-service analytics should make it easy to deliver analytics to every user, while still respecting that not all users are the same. This is how we approach user experience here at Visier, and the above example of forecasting time-to-fill dates (as well as other important talent analytics) is just one example of how we are continuing the pursuit of innovative ways to answer questions — in a way every user can use.
The Future Becomes More Certain
The daily decision that recruiters and talent acquisition leaders face is where to focus their energy and resources. Should they be sourcing more people? Should they be spending time qualifying candidates? Should we work harder to convince the hiring manager to extend an offer?
When you know that you’re going to miss a hiring deadline, the answer to these decisions becomes all the more urgent. The data will help you understand where to speed up the process, thus introducing real, solid efficiency without guesswork.
This leads to better hires and hiring success, and more than any other activity within HR, it is talent acquisition that impacts business outcomes.
Be the first to know!
Never miss a story! Get the Outsmart newsletter.