Overcoming 6 Key Challenges in Establishing a Data Analytics Capability

[This article is taken from a presentation I delivered as part of a broader session on People Analytics for the Australian Human Resources Institute (AHRI) QLD Analytics Network on 7 October 2021. It uses the presentation slides and accompanying presentation script only slightly modified from the spoken word to fit the written form. thanks, Tony]

This article focuses on setting up a people analytics capability in your organization and thinking about what the key challenges are and how to resolve those. 

Let’s start by talking about data-driven insights, people analytics is the focus. But more broadly, there is an untold number of articles and research papers on the importance of data-driven decision making and my bookshelf is full of these books and papers. And I'm sure yours are too (or your virtual bookshelves).

Here is a nice example from Deloitte to set the scene.

Of the organizations they surveyed 39% of those have a strong analytics culture, and 48% of those were significantly exceeding their business goals. Compared to those that didn't display a strong analytics culture, only 22% were significantly exceeding their goals. There's a double whammy in terms of the proportional impact that Analytics gives to an organization. But importantly, also here, there's this angle of culture.  In the survey, most executives believe that they weren't really that insight-driven as an organization. So there is this challenge between the ability to pull data together, derive insights, and actually make decisions and make that a part of the framework for how business is done.

So why is this important to HR? If you can't connect people to business outcomes, then you're really just doing stuff because you think it's a good idea. Doing stuff that you think is a good idea is ok, but it will only get you so far - being able to prove it is a good idea and measure your impact is another thing entirely.  

The importance of data-driven decision making for HR 

Using data helps you prioritize your strategies. You can't do everything, so you need to focus your resources on HR. Metrics and data help you do that and this helps you build that culture of data-driven decision making. If you think about the people space, people related decision making and HR processes are all underpinned by principles like merit, natural justice, fairness and transparency. Without a data-driven approach to this, you're very kind of at risk of replicating the same diversity issues that you see in many organizations: pay equity issues, how promotions and pay increases are awarded,  or who should be hired.  Some of these examples are on a macro scale, for example, your whole company's diversity profile, and some at the micro-level, but the same general principles apply. 

Data is important for setting strategy and for tactical decision making 

At the micro-scale let’s take a specific example of an individual hiring decision.  We have selection criteria for hiring to ensure we get the right person for the role and use multi-source inputs to the process to base decisions on evidence and to avoid bias, nepotism, discrimination etc.  At the micro-level, you heavily rely on good processes, training and company culture.  Ideally, guiding these processes and strategies would be a great analytical understanding of your organization’s diversity profile, the skills and capabilities required for the next 2-3 years, market pay rates for similar roles, the complexity of the role, expected time to productivity… and on. Hiring strategies in the absence of this data are going to be much less effective than they would otherwise be.

The reality is that HR has been somewhat late to the party around the use of data and people analytics. If we think about this from a simple business accountability perspective HR teams are custodians of lots of systems. Not many organizations have just one system. And even if you have just one, we still have to curate and care for that information. It's a rich asset to the organization. 

Putting data in the hands of managers is critical for creating a data-driven culture 

Let’s look at some research from the Annual HR Systems survey. This survey provides a rich set of longitudinal research and here I’ve highlighted some insights they developed regarding the differences between organizations that are data-driven compared to those that are less so. This construct is similar to the Deloitte research we talked about earlier. The bars on the left of the chart are the results for organizations that are described as not being data-driven, and bars on the right are those that are identified as being data-driven.  As you would expect all segments on the right-hand side are higher than on the left, but by far the biggest difference and the thing that really stands out as being different is the deployment of information to managers, putting information in the hands of decision makers.  I have circled this segment in red on the chart.

So, this gives us something to think about in terms of what drives success.  Success isn't necessarily having a great dashboard, success is determined by whether or not people are using data and making decisions with it. 

The “maturity” of people analytics 

There has been a lot written on this topic across the decades, there are more books and research papers than you can imagine. Just a few examples here.  This is extremely well-trodden terrain and there is no shortage of great information to draw from. 

Facilitating the Utilization of HR Metrics – The Next HR Measurement Challenge; Irmer, Bernd E (Ph.D); Ellerby, Anastasia (MBA); Blannin, Heather, 2004

Early research on driving the adoption and use of people data 

In terms of the topic of adoption, this is a key theme for this discussion, and the focus is on the actual use of data in organizations. The image above is an extract from a paper published around 2004 by the InfoHRM company in partnership with the Corporate Leadership Council within the Corporate Executive Board (subsequently acquired by Gartner). This research identified the key phases of sophistication around the use of HR data for business impact. The phases were characterized as:

1. getting your house in order by automating reporting and reducing the load from ad-hoc queries by introducing self-service
2. starting to use more advanced metrics and multidimensional analysis, and then 
3. deploying more broadly into everyday decision-making and impacting business outcomes. 

Through a detailed survey and interview process companies self-identified into one of these categories regarding the maturity of their use of HR data. There was a big difference in what was required of companies in phase three and we will talk more about this.

Facilitating the Utilization of HR Metrics – The Next HR Measurement Challenge; Irmer, Bernd E (Ph.D); Ellerby, Anastasia (MBA); Blannin, Heather, 2006

Two years later, the study was re-run and the framework was updated based on more findings and longitudinal data.  There was an even stronger focus on understanding how to drive adoption and found that there was the dotted line after phase two was even more pronounced and that it was a really hard barrier for companies to jump across. The research provides a number of tactics and best practice advice to address this, but it was clear that having the technology to help scale, automate, improve quality etc. is necessary, but not sufficient for success and success takes good change management, cultural alignment and business impact orientation.  It is the latter topics that also drove the creation of the additional phase, i.e. those companies that were truly having an impact on business outcomes through the use of HR data.

This research was happening at a time when HR itself was heavily focused on the prevailing thought leadership of Dave Ulrich around HR and business alignment and other leading work by Mark Huselid, Brian Becker, Richard Beatty, John Boudreau, and others.  A big part of being a business partner and a business driver was the use of data and evidence-based decision making.

Maturity models enter the mainstream

Interestingly, around a similar timeframe, Gartner was building its own model for how companies could be more data-driven, and the use of business analytics across an organization [the earliest reference to this I can find is from 2009].  Gartner’s framework described this in the form of a four-phase model describing increasing difficulty for companies to move from descriptive analytics up to being able to deliver prescriptive analytics for the highest value. 

Bersin & Associates (since acquired by Deloitte) published this model around 2010. As you can see lots of similarities to what has come before and presents a maturity scale of using people analytics starting from operational reporting through advanced reporting, advanced analytics and up to predictive analytics.

Defining success 

While these models have helped companies and people analytics teams assess where they are and the opportunities to make more of a difference, I have a problem with all of these models. The problem is that they set up prescriptive or predictive analytics as the main destination everyone should be striving for and if you're not doing predictive analytics, then you're not really doing anything of worth and setting up an expectation that is hard to reach and not necessarily the right destination.  Something I'd recommend considering is how you see success and what matters to your organization is the most important thing, how you get there is just a part of the journey. 

Build a sustainable capability and avoid the key person dependency risk 

So, what do we need to do?  Many of you, myself included, may have had a role that could be characterized as “the Excel ninja” in your organization, or HR team. You are able to crunch through data, get data from lots of places, massage it, put it together, create some amazing reports and dashboards, and share them around. But then if someone wanted to see that data cut differently, that became a pile of work and maybe your weekend. 

This is all great for job security and feeling important and needed, but before long you get bored or burnt out, or both. And then you leave or move on to another role. You may have written some great handover notes, but there is an immense amount of tacit knowledge locked up in your brain and everyone likes to do things their own way, so the next person would invariably reinvent everything.  In between times, it is probably hard to fill the role, because people with the right skills are scarce in HR. Basically, relying on the Excel Ninja isn’t a great idea for any company as at some point all their people's analytics capability is going to walk out the door and they have to start again.

Data Scientists are amazing, but you need to build a broader people analytics team 

The lesson there is around building sustainable capability, not just relying on a single person. Now, get ready for a feeling of déjà vu.  We are in a very similar position today with the role of the data scientist. Everyone wants to be at the top of the maturity scale right? So, how to get there, just hire a Data Scientist! But, you are actually creating a much worse problem than you had with the Excel Ninja.  The data scientist is definitely a superhero and is able to do amazing things. But, as before, if you rely on only one person, you're at risk of not creating a sustainable capability for your organization. It is compounded here too, because 80% of the time the Data Scientist is cleaning and aligning datasets and curating predictive models. Most of the time this work is not repeatable and is designed for specific investigations, which can result in great insights, but pretty soon they get fed up and move on and you are left with a massive hole in your people analytics capability yet again.

https://www2.deloitte.com/us/en/pages/human-capital/articles/people-analytics-and-workforce-outcomes.html 

Deloitte has been doing some nice work around evolving this thinking to be more focused on capability creation, as opposed to an escalating pathway of sophistication. 

Peter Howes webinar discussing this and other related topics: https://www.onemodel.co/events/peter-workforce-planning-webinar

Reviewing all of this material I was reminded of a webinar I hosted in 2019 with Peter Howes. As many of you know, Peter is a giant in the industry. He was the founder of Infohrm and a pioneer in strategic HR, and HR systems, a speaker, educator, and author - a true thought leader in every sense of the term.  Peter created this model in around 1980.  His core principles are all still valid today and remain probably one of the best characterizations of people analytics done well that I have seen.  Essentially, if your team is wrapped up in administrative tasks, you should aim to shift the mix to include professional and strategic activities for greater business impact.  You still need to do operational and tactical reporting, that never goes away. Getting greater efficiency and automation for these activities frees you up to do work with greater business impact. 

The biggest challenges with People Analytics

It is pretty clear that the challenges for adoption of people analytics have been around for a couple of decades now, and while technology has caught up with our desires, there is a lot more to success in harnessing technology and developing a sustainable people analytics capability.

1. Data is spread across multiple systems

Even if your company has purchased an HRIS suite, you will still have issues pulling data together from across those different applications and invariably you will also have data in lots of different systems. You spend 80% of your time assembling data and probably no more than 5% of your time doing true insight generation. 

2. Data is not trusted by leaders

If someone doesn't like the message they are hearing from HR, they're going to attack the data. If you have any data quality issues, then that's going to show and it will undermine everything. Even if the inaccuracy is minor and doesn’t affect your message, it is an opening - a weakness.  People will start generating their own data, and use different definitions, resulting in a lack of consistency and trust.

3. Analytical tools are not being adopted

If your tools are too complex, then they won't be used. This is why many tools don't get used in most organizations, not just people analytics products. There are too many options, too many things to click, and that is a barrier to adoption.  Focusing the solution on the real needs of the different users and personas is critical. More is not better in this case, focused insights and fewer options for the end-user is what will bring success.

4. Data security & privacy is really complex

Obviously, in HR, security and privacy are critically important, and often a major reason why people data is not shared around organizations. Think back to the life of the Excel Ninja, they are probably generating hundreds of different spreadsheets and emailing them to managers. This is a lot of work, but it is also inherently risky.

5. High expectations for Data Science and AI/Machine Learning

Machine Learning (ML) and Artificial Intelligence (AI) is seen as being too futuristic for most despite the incredible amount of hype.  “How do we even get started?” is an all too common refrain.

6. Data and predictive models in HR apps are very “black box”

Predictive models and even basic data transformation models are often locked in the head of your Excel Ninja, or in a black box from your software vendor. This means you have a lack of transparency in understanding if there are any quality issues in the movement of data and calculations, or if you have an inherent bias, or how reliable and trustworthy those models are. Back to where we began this discussion, if you are making decisions that impact people’s lives, you need to have good reliable evidence.

Solving these challenges is necessary for success. So how do we actually do that? Let's talk through some tactics and ideas. 

Solving the People Analytics challenges

STEP ONE - Bring your data together

Naturally, bringing your data together is step one. Ideally, into a single data model, or if not, at least a repeatable process for merging your data together so you don't have hands involved in the process. This is really important, because if you have any manual processes you are again spending time on less value adding work taking you away from insight generation, and it's opening up opportunities for errors. 

STEP TWO - Create a set of key metrics and definitions

Creating a set of metrics and a set of definitions is really important, because then you've got consistency. You can then drive reliability and quality through that process. 

STEP THREE - Deploy simple, guided storyboards/dashboards & data exploration tools

Then with a set of defined metrics and storyboards (or dashboards, or whatever you call it) that are consistent and easily understood, you are able to start driving adoption. People get familiar with the frame you are presenting, the terms and the language, and the definitions. This brings a baseline of shared understanding and learning and the ability to then start adding to that through time.

STEP FOUR - Wrap everything in role based security from the start

In terms of security, you should think about security through the concept of user personas for which you construct roles, not thinking about security for individuals. Think about your Executives, GMs, People Leaders, HR Business Partners (HRBPs) etc. and what the different roles are, what data they need to see and then craft the security around the roles. This allows you to set your data free using security as a way of deploying content, not holding it back. Drowning in spreadsheets is often seen as a problem for data consistency, effort etc. but it is also a major security issue that can be avoided by taking this approach

STEP FIVE - Leverage technology and skills to enable the use of ML/AI & predictive insights 

The technology issues around ML/AI are completely solvable. There is lots of technology available and it is not really a technology problem anymore. It is more an issue of capability and understanding. The key is to leverage technology in a scalable way and not fall into the key person dependency trap.

STEP SIX - No magic allowed - make everything fully transparent & explainable

This leads into the last point, which is don't allow the use of black magic and closed systems.  Make sure that everything is explainable and understandable when it comes to metrics and predictive models or whatever kind of analysis that you're doing.

Some practical examples of People Analytics in practice

Let me share a couple of quick examples. Here is a storyboard that is structured around a specific topic and has the key questions your audience would be asking and these leading them through the data. So, it’s really easy to understand what's going on with layered complexity from the high level summary trends through to the details. Everything is interactive, you can click and drill. We are leading people through the topic pre-empting the questions that commonly arise when consuming this content. 

Below is an example using more of a classic KPI style Storyboard. Here you can assemble and browse through the KPIs from the simple to the more advanced, but the layout is consistent and easy to track from the big headline number through the trends and the detailed breakdowns. At any point, you can click and drill.

 One of the most important features here is this pervasive library of formulas, definitions and explanations.  As important is the ability to drill into the details and see who the people are for this analysis (naturally all this is seamlessly controlled by role based security).  The ability to drill down lets you validate the information, but also gets you into action. You are able to quickly dig into key employee segments, identify risks and target interventions. 

These are just a couple of these examples of what you can do to get started fairly simply, but quickly make a big difference in your organization. 

Building on the previous  examples, in the scatterplot below we have added a correlation, which is normally something scary for the average non-statistician, but if you look at the text above the chart you can see an automatically generated written interpretation of the results  using simple business language. Instead of just providing the numbers and expecting people to understand what a correlation coefficient is, or how to interpret significance, be explicit and explain whether something is significant or not – this goes a long way.

Here is a zoomed in view so you can see this more clearly - the chart heading is in the form of a question and the text is directly answering this question.

A Summary of Tactics to Build your People Analytics Capability

The slide above summarizes some of the tactics we have covered, with a few additions to help you build a people analytics capability in your organization.  If you don’t have the skills in HR, borrow from other disciplines, find the experts in the organization who can help you. Reach out to the broader people analytics community. There are lots of resources, networks and people ready to help. The People Analytics practice and network is bigger now than it has ever been. 

Also remember that it's not always just about the data, you're in HR, let's talk to people, be sure to check your findings, go around the organization and build your own network to better understand what's actually happening.

Some final thoughts

By way of some final thoughts. Focus on the questions that matter to your business, start with a small set of things that are repeatable and build trust. This will then give you time to do the more interesting stuff, find opportunities to drive success, and then market your successes. You can build a groundswell of people wanting to get analytics as opposed to you forcing it upon them. And again, it's about insight, not necessarily just about the data, but the actions you can take and the impact you can make. 

People Analytics is one of the hottest areas that organizations are looking to hire into internationally.

The above  framework is designed to help you put  all this into practice. You need to deal with the job of data orchestration to get all of your data into one place and one logical construct.  Focus on Storytelling, not just generating Dashboards.  Blend Predictive Analytics into this and some think of it as an add-on. 

Answer the questions that matter

If you are interested, One Model has heaps of assets that we can share with you. For example, contact us if you want some inspiration around the questions that matter. We have a great library of these and this is a really engaging way to talk to people in your organization about people analytics in a non-technical way.  We also have an e-book titled "Explore the Power of People Analytics" that’s a great resource to get started.