Using operating models to bake Ethics into the People Analytics recipe
Introducing operating models into the debate around people analytics allows us to discuss infrastructure necessary to set up ethical and effective practices.
People Analytics is here … with all its flaws
In 2020, LinkedIn surveyed 7,000 talent professionals across the globe to identify the biggest trends in HR. The study reported a whopping 242% increase of “HR professionals with data analysis skills” (measured over 5 years).
However, data analysis is half of the recipe – for people analytics to be truly effective it is also necessary to draw meaningful insights and design actionable, impactful policies … the exact steps companies report having less mastery over:
Thus LinkedIn’s report suggests one flaw of people analytics is overstated usefulness in effective decision-making. The cleanest, most complete data means little if limited actions are taken.
Another flaw would be privacy threats – it’s easy to imagine people analytics regressing into just another excuse for creepy data collection.
Ethics and effectiveness can be operationalized
Two distinct operating models for people analytics are described in Richard Rosenow’s blog article “People Analytics: Platform Operating Model”:
1. A Service Model born from analytics experts responding to in-house requests to improve processes,
2. A Platform Model born from the adoption of data collection and analysis tools (e.g., Workday, Prism).
In the Service Model, Reporting and Research functions address data analysis and Partnership refers to “client-based decision support.”
In the Platform Model, people analytics processes separate into “platform data science” and “specialty research.” In Specialty Research, unique challenges are addressed while “platform data science” (Platform, Training, etc.) facilitates more automated, routine data analysis.
Going back to the LinkedIn report, we know that companies are struggling with decision support and implementation. From class, we know that even the most well-intentioned data collection can lose employee trust if there is no subsequent action.
Including Partnership as a standard function in the Service Model helps establish best practices – one that addresses both flaws: 1. New people analytics teams are reminded to allocate resources towards decision-support, increasing effectiveness, 2. Data collected is more likely to deliver impact, increasing meaning to employees and building trust.
The real privacy threat is purposeless data collection
What worries me is whether the Platform Model will become the dominating trend for future people analytics teams.
Rosenow states that both models are based on naturally occurring functions observed in growing analytics teams, with the Platform Model being a relatively newer phenomenon.
Currently, the Platform Model has no analogous role to the Partnership function. Instead of aiming to help teams maximize data impact team-by-team, the Solutions function is a pooled “technical support” team whose end goal is to help groups become self-sufficient.
This places the burden of meaningful impact on the platform itself. If the platform design is inappropriate, the company wastes resources (at best) and at worst, unintentionally becomes the big brother that tracks constituents for the sole purpose of watching.
By including Partnership-like infrastructure into operating model discussions, we can bake best practices into the People Analytics recipe, ensuring success.
Sources:
Rosenow, Richard. 2020. “People Analytics: Platform Operating Model – Richard Rosenow – Medium.” Medium. Medium. March 31, 2020. https://medium.com/@richardrosenow/people-analytics-platform-operating-model-57fecb0e7ea2.
Hi Vivien, I appreciated this article about the breakdown of Service vs Platform model and the flaws that you uncovered. I took a look into the Medium article and it seems to bend towards the idea that anything operationalized can be automated. It sounds like from your comparison with the Platform model and solutions function that you believe individualized feedback is more impactful than automated self-serving data. I think my question is that it mentions in the article that Partnerships require PhDs and heavy expertise from Data Scientists – how do we compare the ROI between the two models? maybe I’m missing something here. – Camlinh