Two days ago I sat in on a webcast given by Google’s people analytics manager, Neal Patel. Yesterday I was a voyeur on HR Executive’s 4th annual Predict and Prepare webinar. What I got out of them? Refer to it in whichever way you choose – data, outputs, business-driven – it all surmounts to making decisions with data. Simply, HR analytics is the wave of the future.

To make data-based decisions, the data need to be transformed into relevant information. Data need to be collected, pulled from their source, analyzed, interpreted and reported in a way that the audience can make use of it. Speaking of reporting, an HR analytics group is not a reporting function; the staff goes far further than aggregating data and comparing groups’ scores in a benchmark-y kind of way. HR analytics is about having a question, translating it to a hypothesis, and testing that hypothesis empirically. We are talking about the scientific method applied to HR.

Both webcasts paused to touch on the obstacles to widespread use of HR analytics. Without a doubt, these obstacles are nothing to brush past. Between Google’s Neal Patel’s, panelist and HR tech soothsayer Naomi Bloom’s, and my own experience, here’s the list to watch:

• No data. As Neal talked about the very impressive “Project Oxygen” wherein they were identifying key managerial traits in relation to performance, he admitted that the data they used to conduct their analyses was “what they had”. In other words, he had to work with data collected for other purposes. Such will always be the way of analytics. (If Google doesn’t have a plethora of people-data lying around, I don’t know who will!) That being said, we can make each data source more analytics-friendly by assessing for universally ‘good’ measurement properties and storing data in a retrievable and organized way. We can also make sure we collect metrics each side of the equation – more on this later.
• No expertise. An analytics team needs to be powered by HR folks that work like scientists. In these two webcasts, a few roles were defined.
  • The data architect – Usually the CIO, this person devises the architecture for the entire company and integrates people-data with data derived from the business. Key word: integrates.
  • The data czar – This person knows where each datum is stored, how it is stored, and how it can be pulled out of the system.
  • HR analysts – The scientists of the outfit. Usually at the Ph.D. level (because that’s where you learn research methods sophisticated enough to handle HR issues and data), these people translate a business or HR question (“How can we lower supervisor-level turnover in our China site?”) to a “line of inquiry” or research program that examines the question every which way from Sunday. Analysts break down the line of inquiry into successive research projects, each hallmarked by a testable hypothesis given the data the organization has available. (An example of one hypothesis would be, “H1: Supervisors leave because the company-supported management style is at odds with the country culture of China, specifically the cultural value of “saving face”.) In short, you can’t do science without scientists: analysts can run through the pairings of data and questions in their mind, referencing the type and quality of data available. They churn through countless permutations of analyses in the blink of an eye – well, at least within a day or so. They are the machine of the Analytics outfit. In addition to their day job, HR analysts also act as consultants to IT and HR leaders, advising them on the type and quality of their people metrics.
  • Data jockeys – Supporting the analysts, jockeys run inferential, statistical analysis in order to answer hypotheses. These folks can mine data until the cows come home.
  • A marketing rep – Your analysts can do it, but it won’t be pretty (plus their time is expensive): I’d recommend assigning a marketing person to make the Analytics outputs digestible, whether written or PowerPoint.
• No audience. It still astounds me how many business leaders say they want the facts but balk at the data. Maybe they are just used to hearing opinions from HR, and are therefore skeptical. Regardless, the members of the people analytics team at Google felt compelled to put a sticker on the back of their laptops (“So executives can see it as we’re presenting”, said Neal) that reads: “We have charts and graphs to back us up – so f--- off”. The dashes after the "f" are more like fuzzy letters on the sticker, but you get the point. If analysts have to drop the f-bomb to get assert their empirical objectivity, we do indeed live in a world of executive doubt.

Combatting the above is reasonably straightforward, although not easy.

1. Hire, or although somewhat less likely, develop the talent to get the job done.
2. Collect the right metrics on both the “predictor” and “criterion” side of the equation – the metrics that relate to and presumably predict criterion like employee performance, turnover, attendance, and health; not to mention shrinkage, product quality, process yield, customer service, customer retention, and the big kahunas, business –level financial metrics. If you don’t know what “predictor” and “criterion” mean, I’ll refer you back to step 1.
3. Collect demographic metrics on employees, and keep them up-to-date.
4. Assess the metrics you do have for decent measurement properties – I’m talking normal distributions, skewness and kurtosis and their brethren; evidence of validity and reliability etc. – and improve them wherever possible.
5. Buy a stats package. Don’t ask your analytics group to do this stuff in Excel. That’s just mean. It's like asking a lumberjack to cut down trees with a wet herring – well, maybe a butter knife.
6. Work with vendors to get a data feed of raw data or put it all up in the proverbial “cloud”.  You might as well deal with privacy issues – like the anonymity of employee opinions collected in surveys or Germany's super-strict data privacy laws – while you’re at it.
7. Structure the data so that all sources can be integrated based on employees’ unique identifier.
8. Teach leaders and managers to trust the data – even if they don’t trust HR yet.

The road to data-driven HR is full of potholes – some minor bumps, others sinkholes perhaps only marginally smaller than the Grand Canyon. Don’t spend the next year wandering around the bottom of the gorge without a compass. Hire an HR analytics leader and get them talking to IT – it will take time to import data, create new SQL tables and collect data in a different way (or collect different data altogether).  I encourage you to take the leap: get started now and in six months you’ll be taking your first steps towards supporting a data-based HR practice: a practice that will support a more effective and efficient organization. What’s more, through the wizardry of analytics, you’ll be able to show exactly how HR’s data-based contributions have improved the organization’s efficiency and effectiveness. At whatever point you are on this journey, let me know if Bersin & Associates can help.