The State of HR Analytics Today (2024)
Background
A friend recently asked me about the state of HR Analytics — what are organisations doing now in this domain. I had forgotten that I had introduced my friend, a data scientist, to the domain of HR Analytics years ago, and she had fallen in love with it. She’s now done several stints in major organisations as an HR Analytics lead. I myself was introduced into this domain by my consulting business partner (Sally Taher) who was an early pioneer and a passionate advocate of it.
I’ve written here before about HR being the weakest link in a knowledge economy. That article resonated with many, including those in HR — it was like holding a mirror up. This weakness in HR obviously impacts the progress of HR Analytics in very significant ways. And so I thought I would kick off 2024 with my 20th weekly article by talking about the state of HR Analytics, aka People Analytics, aka Human Capital Analytics.
(I write a weekly article on bad thinking and bad practices in data analytics / data science which you can find here.)
Wrong Focus
A quick google on what’s trending in HR Analytics in the last couple of years reveals almost no progress. It’s the same rhetoric. Predictive scoring for employee acquisition, predictive scoring for employee retention, measuring employee engagement and employee sentiment, improving employee experience. These are all the WRONG things to focus on when building out a discipline. Employee engagement and experience are complex ideas, and it’s arguable whether they really matter, and ultimately, aren’t really actionable and not a major driver of bottom line outcomes.
I’ve also written before that employee retention isn’t worth pursuing. Much like what we learnt in Customer Science, attrition management is not a high yield or low-hanging-fruit exercise. All retention initiatives are interventionary in nature, and as such, won’t benefit much from prediction scoring. But HR folks continue to be fixated with employee retention. Perhaps it stems from a misinformed view that saving one employee (or customer) is worth more than a few new employees (or customers). The 1990 Harvard Business Review article by Reichheld and Sasser claimed that companies can boost profits by almost 100% by retaining just 5% more of their customers. This has since been debunked based on real-world data and observations. But the idea remains persistently in the collective mind. It’s also worth noting that retention predictive models are one of the easiest to build; the signals are strong and fairly intuitive to understand. But having a precise model doesn’t mean you can change the outcome.
Pursuing Standards
HR shouldn’t be spending their time solving for retention. If they cannot connect attrition rate to productivity loss or organisation disruption (which they obviously can’t today), then why own that KPI? This speaks to the broader challenge — creating standards in HR Analytics practice. In customer and marketing science, many standards have evolved to both guide and strengthen the practice. For example, there is an approach to using customer lifetime value (CLTV) that helps shape prioritisation and engagement decisions. There is an understanding on communication fatigue that shapes database marketing action frequency.
Such guiding standards have yet to emerge from HR Analytics, and frankly, I’m not seeing efforts being made to pursue that. Instead, everyone’s trying to be innovative by creating their own spin. For example, what is the proper way to build an employee acquisition score, much like the way credit scoring has become standard. It’s not so much the input variables that need to be ‘standardised’ but rather, it’s the standardisation of what we are predicting. In credit scoring, we are always predicting likelihood to default in a given period. But for employee acquisition, are we predicting likelihood to perform? What does ‘perform’ mean?
What Matters
What matters is what employees care about — they care about equity, opportunity and growth. HR Analytics should therefore be focused on creating standardised knowledge on the measurement and levers of employee value which is the foundation for equity. HR Analytics should be focused on standardising the definition and measurement of organisationally-aligned skills and competencies to create proper ontologies which is the foundation for opportunity and growth. HR Analytics should be focused on understanding how work gets done in a networked and collaborative knowledge economy, and solve for the attribution of employees’ contributions. These are difficult but not complex work. They are also not ‘sexy’ work. But foundational work seldom is.
“Define your talent, know your talent, grow your talent.” That should be the principle mantra for HR Analytics. Everything else is just noise at the start of firming up the domain. Build the foundation so that the incremental ‘innovative’ stuff like engagement and sentiment have legs to stand on.
Back to Basics
Instead of wasting time being enamoured with fancy algorithms, HR Analytics needs to hunker down to address the data issue. (Having good data ALWAYS trumps good algorithm.) Few professionals talk about Data Management (curation and information representation) and Data Governance (quality and risk) for HR data. Much of the data quality is suspect. Performance appraisal data do not represent the truth on actual performance (you know how this game is played!). Job description data do not represent actual job scope. Skills and competencies are not properly documented (e.g. from resumes) and there’s no standard ontology (i.e. relationship between skills). What are the critical data elements for HR data and are HR professionals aware?
Recent surveys indicate that many organisations need to traverse an average of three HR systems to get basic analytics-worthy data. There seems to be a lack of attention paid to leverage age-old data warehouse design frameworks to stitch HR data together; many organisations still don’t have their HR data sitting in a proper data warehouse. The data doesn’t have to be perfect or clean, but it needs to be well-organised. By spending time on data curation and organisation, one learns very quickly about the over- and under-representation of certain types of HR data, and quickly understands the gaps. “Build your roads before you build your cars.”
Conclusion
Much like how Customer Science and Marketing Analytics mature, HR Analytics professionals, together with academia, should identify and converge on the same core set of foundational problems to solve, instead of trying to outdo each other with clever algorithms and rhetoric. The core tenets of the practice have yet to be well established, and in their absence, competing on ‘innovation’ won’t yield substantial outcomes.
