Wait – who did we employ again?

Modern work management tools, such as Fintertip, create massive amounts of data about their users as a convenient side product. This information can be utilized multiple ways. Many are afraid of an Orwellian surveillance workplace in which labour is minutely measured and the salary is brutally adjusted based on the cold-blooded numbers. For example, Amazon has gotten bad publicity about how they track their warehouse workers working in their fulfilment centres, some losing their job due to insufficient performance metrics.

But data about one’s employees can also be used in more benign ways. One is to identify the different employee profiles you have and for example plan training accordingly. You can check whether people with similar roles or working in the same team have similar or diverging digital footprints. And naturally, it can be used to create a quick-and-dirty picture of your organization and the different roles within; a classical organizational chart created dynamically from data.

Fingertip’s gamification module is not only a tool to make working more fun and inserting good-spirited competition in your employees’ daily lives; it is also a tool which can be used to assess the different roles in your team. Fingertip scores and records a wide variety of actions a user can take inside the system, from logging a call to a customer to creating a chatter message to a decision.

And besides gamification, there are obviously how people go about doing tasks and making decisions within the system. Do they create a lot of decisions or do they finish tons of tasks? Are their decisions visited often? Or even, how many characters do they use to describe their decisions? Obviously not all work done is not documented in Fingertip but with the data it collects, you can get a lot done.

As an example, I gathered data from Fingertippers who have been active in the past month, in order to remove past employees. The data set consists of 18 different variables. This comprises of data from a variety of sources, such as gamification (for example the proportion the actions classified under five different classes, such as Tasks or Chatter), decisions and tasks. But by no means are the chosen variables the only way to represent users in a mathematical form.

The users were assigned to three distinct clusters using Partitioning Around Medoids (PAM) algorithm. The clusters produced relatively distinct groups. For plotting purposes, the initial 18 variables were reduced to two using t-sne algorithm which is a dimensionality reduction method akin to principal component analysis which naturally tries to group similar observations together in the new space. The users are seen below in a two-dimensional space with coloring by their cluster membership. The X and Y axes cannot be interpreted in relation to the original variables, so they do not yield any information in themselves.

One group consists of the core users, shown below in blue as cluster 3. They naturally show a lot more activity in completed and created tasks and decisions and their gamification activity is prolific in Task-related actions.

Then we have a cluster, shown in red, which has most of the board members and some people who don’t work 100 % of their time for Fingertip. Their infrequent activity focuses on decision and chatter, they barely bother with completing tasks.

Finally, in green, we have people who don’t really make decisions but instead get assigned a lot of tasks; many of these people are software engineers whose job is to improve the software.

For privacy purposes, names of the employees are not shown here. But for a CEO it can be very revealing to see which employees cluster together and therefore show similar working patterns.

A leader who does not know his employees is essentially half-blind. Data can never replace actual contact but it can also reveal stories that would otherwise remain hidden. Use it wisely!