<img height="1" width="1" style="display:none" src="https://www.facebook.com/tr?id=353110511707231&amp;ev=PageView&amp;noscript=1">

What is Data-Driven Decision Making? A Guide to Getting More from Your Data

July 13, 2017

Data is everywhere! Twenty years ago, people would say "you’re living in the information age.” With technology advancements, pieces of data are thrown at us more than ever. Your “likes” on social media are tracked and stored in a “data lake” somewhere so some sophisticated algorithm can determine what ads appear on the side bar in your internet browser. You can install a device in your car to monitor your driving habits, such as your speed and how hard you hit your brakes. Then, the insurance company can use this data to adjust your insurance rate. And, if you're like me, you probably have considered buying a fitness tracker to monitor your daily steps, heart rate, and sleep patterns to help you reach your health goals. With or without your consensus (and awareness), we are in an era where data is being created and used constantly.

If data can help companies market products based on your interests, give you reasonable car insurance rates, and motivate you to live healthier, wouldn’t it be cool if data could help you find the right people to work with you? What if the data could help you optimize your talent pool? It can!

Data, regardless of size, format, and quality, has been used for centuries to aid in all kinds of decisions, from whom you’re going to hire to babysit your kids on date night, to personnel selection in the pre-employment phase, to mapping out talent management strategies and increasing employment engagement. Data points in the hiring and talent management process include:

To get the most out of your data, we have compiled five steps to make better data-driven decisions:


1. Identify the Right Data

Ask yourself – what is the question you’d like to answer, and what kind of data will answer it? Clearly define what you need – is it individual level, group level, or business unit level? Are you leaving any groups out? Are you considering possible extraneous factors? Are these data “job relevant?” Does law prohibit the use of this type of data for this purpose (e.g. promoting from within, recruiting and hiring external candidates, etc.)? Don’t skimp on this step. Take the time to identify the right source of data.
I can’t stress enough the importance of this step. Even if you had all the data in the world, it wouldn’t help with your decision-making unless it was the “right” data. You can build out a beautiful, perfect fitting model to predict variables, but if your data isn’t appropriate or representative to begin with, the algorithm will be problematic down the road. You might find yourself with a lawsuit on your hands because certain minority groups are not accounted for in the data.

2. Collect Data Wisely

Now that you have determined the type of information that will help you make better decisions, you will need a thoughtful plan to collect the data, if it does not already exist somewhere in a data warehouse. This is where things can go south quickly. You might find that the data exists but it is stored in multiple independent systems, or possibly in non-digital format. 
For example, let’s say you want to build a predictive model of driving behaviors to reduce incident rates. You have years of incidents reports with hand-written investigation notes on file and paper copies of a drivers’ log. But, you also have digital data, as you recently implemented a system on each truck that automatically captures driving data. What would you use? How would you integrate all those data into a manageable form for analysis? Could the data easily be replenished on a regular basis?
In the case of no existing data, how do you go about collecting it? Have you gained buy-in from the relevant parties and identified your data champions? Have you identified a person to be your data campaign manager who can educate key players on the business value of investing time and resources for this data collection? If you are doing an organizational survey, are the participants ready to provide honest responses?


3. Analyze Data Responsibly

Once you have overcome any barriers and gotten your data into one place, data cleansing should come next, before you start any real analysis.  Ask any statistician or data scientist - they’ll echo the sentiment that the data cleaning phase takes time. Bad data leads to erroneous conclusions, and this step ensures that you are working with only valid data. In a recent blog on outliers, we illustrate how some data points could be extremely influential in analysis outcomes. This also can happen in validation research to understand how employment assessments predict job performance. 
Whether you are working with industrial/organizational psychologists, statisticians, or data scientists to analyze your data, ask them what steps they have taken to ensure the quality of the data. Don’t be intimidated by jargon and methods they might employ, like hierarchical linear regression, random forests, or machine learning. Just remember, statistical models are simply tools that can summarize observed trends in data.  There isn’t necessarily a superior method. The choice of analytic technique should be grounded in theory.


4. Interpret Results with Visual Aids

Next, you’re likely to be tasked with communicating the results to leadership. How can you communicate the results in an impactful way? Use visual aids. Beyond the usual Excel and Tableau, there are many more tools available for you to create a story-telling data visualization.  To just name a few, RAW, Plotly, Datawrapper, INKSCAPE, and Power BI are among some of the popular resources.
Consider your audience when creating these charts. If you have company branding guidelines, use your company’s color scheme consistently. Experiment and consult with your colleagues – if you find it hard to explain the visualization to others within 30 seconds, it is probably too complex. Simplify it further so your target audience can easily grasp the most important message from your findings.


5. Cross-validate, Cross-validate, Cross-validate!

You’ve gotten your results and developed a new strategy for making data-driven decisions, but your strategy should be adaptable. Change is constant. What worked yesterday and today might not work tomorrow. What you find in one set of data might not generalize to another job, position, or under new leadership. Jobs, positions, organizations, industries, and economy will change over time. Keep in mind that you’ll want to devote a little time to reevaluate and restart the data analysis cycle with any organizational changes. 


Remember: to be a savvy user of today’s abundant data, you still need to exercise your judgment. Statistics and data are double-edged swords that can steer you in the direction of higher productivity and profitability, or could dangerously mislead you if executed with little care. Data-driven decisions are fortified when used with assessments and behavioral based interviews to evaluate things like safety, culture fit, and motivational fit.

HR Analytics

Mavis Kung, Ph.D. Mavis Kung, Ph.D. is the Manager of Research and Development based in the Pittsburgh office of PSI. She focuses on conducting validation studies, acting as an internal technical expert on selection for project consultants and clients, analyzing assessment data to determine selection system effectiveness, validity, and fairness and providing recommendations for system improvement and development.