Four steps to People Analytics excellence

Despite a lot of attention and activity currently devoted to the topic of People Analytics, I find that there is still a struggle to articulate crisply what the purpose of the field is and how to make meaningful progress within it.

Much of this struggle, I believe, comes from the fact that the path forward is cloudy. Data and technology in this space is still nascent and there is a hectic scramble for share of mind. People Analytics and HR tech conferences are popping up all over the place, and although they are brimming with ideas, they don't really bring the clarity of purpose to the space which is sorely needed.

Along with this, we see a massive growth in organizations establishing dedicated People Analytics teams. In some cases these are just rebranded HR reporting teams, and in some cases they are genuine investments in advanced analytics and data science talent. Whatever they are, many of them struggle to put strategic direction to their work because of the lack of clarity around the purpose of People Analytics, which makes it unclear where they want to go to and how to get there.

Over the past few months, based on a range of work that I am involved in in the People Analytics space, a framework has become clear to me that summarizes the progressive layers of attainment and insight that People Analytics can bring to organizations and enterprises. There are four 'steps', each one enabled by, or a progression from, the previous step. Many existing People Analytics groups will be at the bottom of this staircase, still trying to grapple with how to progressively move up. Other groups will be attempting work at higher steps of the staircase but possibly struggling because of a lack of progress on earlier steps (for example attempting predictive models without appropriately refined input measures).

Wherever you are on this journey, I hope this framework is useful and helps towards that clarity that we need in this space.

Step 1: Automate traditional HR reporting

Freeing resource is a key unlock to generating next generation people insights. Many of the people analytics leaders I have spoken to express frustration at the lack of capacity to engage on genuinely advanced analytics because all of their resource is tied up in multiple highly manual people reporting processes. This can range from tedious checking of data to repetitive manual processes in generating reports on that data.

Mastering technology that can automate some of these tasks is critical. Not all of them can be automated, but depending on how well data is captured, substantial chunks of work can be set up in automatic self-updating processes which can substantially reduce repetitive, uninspiring tasks which offer no learning potential to people analytics talent.

Harnessing tools like R and Python, and to a lesser extent Alteryx and Tableau, is critical to moving up onto step one and achieving greater levels of automation in your reporting. Examples of this automation at various levels include:

• Developing scripts to process reports 'in the background' on local computers, allowing analysts time and capacity for other tasks
• Automating the presentation layer, by using scripts to populate Powerpoint (see Ben Teusch's tutorial here), or utilizing online data sharing tools like R Shiny or Tableau
• Fully automating HR reporting by generating ETL jobs and presentation layers which push and pull from databases (see my tutorial here),

Step 2: Apply advanced methods to critical people-related questions

Freeing up analytics talent through greater automation allows more time to be spent on analytics that drives genuine insight on critical people related questions. The first step is to ensure a strong understanding of the people priorities of the organization, so that a focus can be placed on those priorities.

Knowledge of, and an ability to apply, a range of advanced analytical and statistical techniques is critical to move from Step 1 to Step 2. Here are some techniques which I view as 'must haves' to make genuine progress at this level:

• Basic statistical modelling and testing (correlation, regression, linear and non-linear modelling, significance testing). This allows basic, early stage examination of data to determine its relevance and importance to the problem under consideration.
• Dimensionality reduction techniques, particularly Principal Component Analysis. Many people related problems involve too many data dimensions to realistically analyze separately. These techniques allow reduction to a manageable number of dimensions, usually two or three, and the formulation of insights that are easier to digest across the organization
• Unsupervised learning techniques such as clustering, anomaly detection or signal separation techniques. These help clean and sharpen large sets of data points and identify how to simplify into meaningful subgroups.
• Time-series analysis and basic epidemiological techniques such as survival analysis. This allows the tracking of populations over time, a critically important element to people analytics.
• Text analytics techniques like TF-IDF and Topic Modelling. So much people data is text. The ability to process this text analytically is very important in extracting value from it.

Success on Step 2 does not just involve applying techniques such as these. It is also critical to communicate the insight in a way that can mobilize action where needed, and to build operational tools within the organization that take advantage of such insights. For example, if biases are detected within a decision making process, building a tool where decision makers can benchmark their decisions against other indicators or historical statistics can help self-correct these biases.

Step 3: Improve people measures

An inevitable consequence of Step 2 is that new insights will highlight the need for new measures. In some cases, new measures are needed because there simply is no existing metric for something that is believed to be critically important. In others, existing metrics are found to be unsatisfactory due to poor statistics and measurement properties, and so a redefiniton is needed. Finally, latent metrics can be discovered in Step 2 which can now be made explicit in everyday tracking and reporting. For example, it could be discovered that a certain response level to a certain survey is a powerful indicator of likelihood to leave the organization in the next 6 months, and therefore it becomes important to introduce this measure as a risk indicator in future reporting.

Unearthing and solving measurement problems is, in my opinion, the most challenging step. Often the introduction of new people measures involves a lengthy process requiring changes to systems, new technology and behavioral change. A recent piece of mine outlined some of the most important current challenges in People Measurement. Tackling these challenges effectively offers up the reward of progression to the highest step of the staircase where you can build a deep understanding of how all parts of your employee journey piece together.

Step 4: Build a longitudinal view of the employee journey

Only when you have all the pieces can you start to put the jigsaw together. Enabled by strong data capture (Step 1), improved insight into critical people outcomes (Step 2), and better, stronger measures which are closely tied to those outcomes (Step 3), the possibility now exists to make decisions and predictions which are informed by a long term understanding of the employee journey.

This is where you can start to use your data to generate accurate predictions on the most critical outcomes in your people processes, for example recruiting, performance, promotion, retention. Most attempts to generate Step 4 predictive insights prior to making progress on Steps 1-3 will likely fail or yield obvious insights (e.g., the strongest predictor of promotion is performance rating).

Critical to extracting value at Step 4 is to have strong technical expertise in predictive analytics and machine learning. For a more comprehensive treatment of this see my series on Machine Learning in HR.

Reaching Step 4 is a long term game. At McKinsey, for example, we are only now reaching Step 4 in some elements of our people analytics - elements where a decade or more of development in systems, analytics and measurement is finally paying off. We still have a long way to go in many areas.

As much in People Analytics as in any field of endeavour, having a clear path is critical as you develop out your capabilities and expertise. Its helps greatly with choices around resource, direction and focus, and provides a 'big picture' where everyone can see how their piece, big or small, fits in to the longer journey. I hope this approach and framework helps bring some clarity and direction to People Analytics as a field, and to the many talented people working within it.

I lead McKinsey's internal People Analytics and Measurement function. Originally I was a Pure Mathematician, then I became a Psychometrician. I am passionate about applying the rigor of both those disciplines to complex people questions. I'm also a coding geek and a massive fan of Japanese RPGs.

You can message me on LinkedIn or engage with our People Analytics group within McKinsey's Organization Practice.

All opinions expressed are my own and not to be associated with my employer or any other organization I am associated with.