“Big data” is hot right now. The terms used to describe this trend are many: people analytics, data science, data-driven organizations, predictive modeling, machine learning . . . and the list goes on . . . I find myself wondering how these concepts fit together. Are these terms different lingo for the same general idea? Which concepts should my organization be pursuing? Here is my take: these terms do refer to different methods, some quite sophisticated and others quite simple, but they are all similar in that they aim to use employee data to better and more proactively manage talent. Using people analytic techniques, we can:
- Answer questions that help us leverage people to make the organization more productive, achieve strategic goals, and align talent with desired culture/values
- Identify, recruit, and select the best talent
- Evaluate training and programs
- Identify major skills gaps in the organization
- Uncover inequities or gaps in pay, promotions, and opportunities
- Predict and prevent attrition of top performers
- Understand how the employee experience impacts key organizational outcomes (e.g., employee engagement and customer experience; safety culture and accident rates; new hire perceptions and turnover)
The world of people analytics may seem overwhelming, but the basic premise is actually quite simple and it is worth understanding. In particular, people analytics can feel especially challenging for small to mid-sized organizations. Smaller companies may want to create insights for the organization using data, but find themselves lacking in ‘big’ data due to smaller number of employees or less sophistication in tracking talent-related metrics. All this got me wondering: What can big-data and people analytics look like for smaller and mid-sized organizations?
Challenges for small and mid-sized organizations
Garbage in, garbage out
More and more organizations are making data a focus and priority. Recent studies show that up to about 70% of companies have or are building a data warehouse (Bersin, 2017). Although this may be an overall trend, we often find that data management can present challenges for smaller and mid-sized companies. Many smaller companies don’t have the time, resources, or infrastructure to house their data in an easily accessible manner. Companies want to make data-informed decisions but often have data stored in multiple places that don’t talk to each other. Without some way of connecting metrics (e.g., tracking department data over time, or identifying all employees who completed a specific training), it is very difficult to answer people analytics questions.
What to do? It may be time to invest in a database, data warehouse, or data analyst to create some structure around the way your company collects and stores people data. Even slight changes may vastly increase your ability to leverage data. If you aren’t able to invest in this right now, consider who on your team may want to take on some of this responsibility. Is there a small group of people who would like to serve as the ‘data committee’ for the organization?
One major challenge of having a small organization (or small departments) is protecting people’s confidentiality. Ensuring confidentiality is especially important in smaller organizations or small departments where people may be extra sensitive about their individual feedback/data being viewed or shared.
What to do? Decide how many people you need, at a minimum, to look at a group’s data. For engagement surveys, we often use a threshold of five, meaning that we do not view or share any group’s data if there are fewer than five responses. Some of our clients use a more liberal threshold given their size, and are willing to share groups’ results if they have at least 3 people. Once you determine a threshold for sharing data, communicate it to employees and reiterate that it was put in place to protect confidentiality.
Comparing groups or observing change over time
When we make comparisons, either when comparing groups or looking at the change in an engagement score from one year to the next, having more data allows us to be more confident that differences are meaningful and are unlikely due to chance. However, when we have small groups, we are less confident (statistically speaking) whether differences are meaningful or not.
For example, if there are five people in your HR Department and we look at percent favorable scores to understand engagement (i.e., the percent of people who agree or strongly agree with engagement survey items), each person’s perception represents 20% of the group’s score. One person changing from an ‘Agree’ response to a ‘Neutral’ response can swing the group score from 60% to 40% favorable. People’s perceptions are their reality, so we should not dismiss feedback from small groups, but we may need to interpret their percent favorable scores with some caution.
As another example (see below), in a large region with 500 people, we can be confident that changes in engagement reflect a real change for the group because we have many observations contributing to the scores. However, in a region with only 10 people, we are less confident whether differences are true for the group or mainly due to perceptions of 1 or 2 people. For this reason, it is harder to find statistically significant differences with small groups than it is with large groups. For small organizations and/or small groups, we need to consider our confidence about changes (i.e., statistical significance), as well as what seems practically meaningful.
What to do? The more data we have, the more confident we can feel about making comparisons. This doesn’t mean that we can’t make comparisons with small groups, but need to think carefully about the conclusions we draw. Ideally, we are able to find statistically significant differences, meaning differences are unlikely due to random chance. We can also use our common sense and think about what size of difference is meaningful for the group. Increasing engagement from 25% Favorable to 50% Favorable may not be statistically significant in the scenario shown above, but is a large change and may be meaningful, especially if we continue to see this increased level of engagement over the next several years.
Predicting infrequent events
In a small company, events like turnover and hiring typically don’t happen on a frequent basis, which makes it more difficult to understand the commonalities of what contributes to success. Individual differences and the organizational environment at the time may be more apparent to us than patterns across people or over time. As an example, if we want to evaluate an onboarding program for a company of 50 people, it will take some time to collect data about new hires’ experiences. On the one hand, we could wait and collect feedback until we have a fairly large sample of new hires; on the other hand, if it takes us five years to get a reasonable sample, we risk our data being old and outdated.
What to do? Make predictions when you can – some information is typically better than no information – but keep in mind that the information isn’t perfect. Supplement your data by talking with other organizations about their experiences, looking for information from other sources/outside research, and using as many types of data as possible (that are relevant).
Lack of data strategy
One of the biggest reasons we see organizations (large or small) struggle with people analytics is because they lack a data strategy. Many organizations don’t identify what questions they want to answer or what business outcome they are trying to understand. Rather, they have data and want to do something with it.
For example, let’s say the organization is struggling to retain its best people. A relevant question would be to ask: Does the experience we are creating for employees contribute to employee exits, and if so, where could we focus to make improvements?
What to do? It is time to create a data strategy. Start with the business outcome or strategy that is most critical right now and work backwards. What types of data are important to understand the business outcome (make sure to consider multiple sources of data)? Is that data currently available or something you need to attain? How can you go about collecting the data you need? Who do you need to collect data from? Once you ask the right questions and have a plan in place for answering those questions, the right kinds of insights will emerge.
Other strategies for talent analytics in small/medium-sized organizations
- Mix in qualitative data. For employee engagement surveys, where quantitative data allows for quick and easily comparable feedback, open-ended questions allow employees to provide rich explanations, examples, and context that helps understand their experiences. It’s best practice to use both quantitative and qualitative feedback to create a well-rounded, holistic story with data. And, for smaller and mid-sized organizations, fewer people means it will be manageable to read through open responses and pull out themes. Consider including open-ended questions on your next survey, facilitating focus groups, interviewing select employees, or observing employees to collect qualitative information about your workforce.
- Go to the research. While your organization is no doubt unique, external research can be a great way to is to identify factors that are important across workplaces, locations, contexts, and industries. When you feel at a loss with your data (or lack of it), hit the books, read your favorite journals, and search credible sources. You will likely find answers or at least some information that will serve as a good starting point. Some research and sources are better than others, so make sure to read with a critical eye.
- Check your confirmation bias. As humans, we use heuristics and rules of thumb to make decisions all the time. This allows us to function in our world (imagine if you had to think carefully about everything you do, including walking across the room, answering the phone, or smiling back at a coworker). Subconscious guidelines, while helpful, can lead to poor decision making, biases, and misguided action. For example, confirmation bias is our natural tendency to look for information that supports our viewpoint. When interpreting data, it is especially important to keep our confirmation bias in check. Often, data can be a nice confirmation of what we already know or expect, but not always. Challenge yourself to look for information or data that supports the opposite of what you expect. You may uncover a new perspective, or, can feel confident that your initial conclusion is correct.
- Get creative in how you collect data. If there is certain data you want but don’t have, get creative in how you might be able to capture the information. As an example, one Newmeasures client was interested in understanding the engagement of its top performers, but did not have a standardized metric to identify those people. We guided this client to ask managers to provide a quick rating of each employee by asking: 1) Would you hire this person again? 2) Do you want to work with this person in the future? 3) Is this employee and ‘A,’ ‘B,’ or ‘C’ (we defined these buckets) performer. With minimal effort the organization was able to get “good enough” data to understand that high performers were indeed experiencing the workplace differently from others and what to do about it. The conclusion – when data is lacking, ask your team to put their heads together to brainstorm how you may be able to get some useable metric.
Bersin, J. (2017). People analytics: Here with a vengeance. Retrieved from https://www.forbes.com/sites/joshbersin/2017/12/16/people-analytics-here-with-a-vengeance/#60ef3f8032a1