Subscribe to DSC Newsletter

In general, any expression of performance that applies to a department can, if the data system is configured properly, be stated in relation to individual workers.  For instance, if # of sales contracts / # of customer enquiries = success rate, the success rate can be given for the entire dealership and also for each sales agent in that dealership.  Due to the differences in performance between agents, it can be problematic to only make use of the aggregate.  Some agents might be blamed for poor performance even when in fact they are the strongest players.  It is questionable to use aggregates when developing and implementing plans of action to improve performance - even if this initiative is meant to bolster the entire department.  The ability to characterize a business problem in relation to its actionable components is an important aspect of managing a data-rich production environment.

 

Although I am not a statistician, I still recall the few courses that I took during my university years.  I found the examples, as one might expect from text books, rather general and generic.  Consequently, when a graduate enters a production environment for the first time, it might seem perfectly normal to analyze problems in terms of generalities and aggregates.  It is “in your face” analysis to evaluate operational concerns in terms of specific individuals, their particular behaviours, at exact points in time.  Yet it is important not just to focus on people.  It is necessary to consider the behaviour of the market, the capacity of affiliates, the feedback of clients.  Again, the underlying objective of any methodology is to bring actionability closer to the action.

 

Somebody who wishes to advance the cause science might make a statement like, “I have discovered that high levels of absenteeism immediately follow holidays - and this is the same time that demand for our services is high.”  A person more concerned about actionability might say, “Robert has demonstrated a high likelihood of calling in sick after holidays - possibly to avoid the high call volumes that we tend to experience on those days.”  What is the difference?  In the second case, the manager knows exactly how to deal with the problem.  I know what readers might be thinking.  They are thinking that I probably don’t have any friends.  It is not a matter of friendliness.  The numbers help people make informed decisions.  Of course, I believe that every person responsible for the data also has a responsibility to search for errors and to ensure that the data is ontologically sound - that it exists to honestly serve its intended purpose.

 

More often than not, performance is inconsistent at least in the short-term.  It is necessary to have ways of tracking its development - for each person.  It might be fun to follow the technicals for an entire department or company.  But who deserves a raise?  Whose techniques seem to work best?  The aggregate provides a lens - but it is unjust and imprecise.  If there are 100 sales agents on staff, it is necessary to examine their individual performance characteristics for each day.  Some companies have part-time staff.  A worker might work only a few hours each day.  It then becomes necessary to create hourly partitions.  A manager might ask, “Why is there such a big improvement in sales between 3 PM and 6 PM on Saturday?”  It would be comforting to have a concrete response like, “That is because of Janice.  She is an amazing part-time worker.”  Given enough resolution, large amounts of data can reduce inconsistencies while at the same time revealing performance trends and idiosyncrasies apparent in shorter time periods.

 

While on the subject of “large amounts of data,” consider the logistics that might be necessary to analyze data both on an aggregate and individual level - especially if the data system is not designed to do so.  One reason to explore phenomena on an aggregate level is because it is easy to present this type of analysis in text books.  But let us say in real life that there is great demand for more precision.  It is possible to generate spreadsheets showing the components of performance for all of the individual sales agents for any given day - not necessarily in the same row or column each day.  It is so tempting to run a report for each agent - thereby multiplying the time do the job by the number of agents.  If on the hand the daily data is all on the same spreadsheet.  It is so tempting to sort it - thereby adding a sort followed by a summation operation for each person.  Really, if the daily report contains 10,000 rows of agent data, that is a whole lot of sorting and summing.  I use what I call a FACT - a flagging-and-counting template - allowing me to handle an unlimited number of rows without manually entering a single sort or sum.  Who is the data scientist?  Well, I stand analysis-capable while the other person is taking up space in the office waiting for manicured tables.

 

It is not unusual for a person to be asked to provide an analysis comparing specific individuals - e.g. an “abnormal v. normal” performer - over specific time periods - e.g. during periods when business might be suspiciously poor.  It is a matter of great interest to try to determine what may have caused or contributed to lousy performance.  It is an angel’s quest to help protect those that are doing their best and identify areas of concern where people not at their best might do better.  It is also important to seek out external explanations that might be expressed in individual terms - e.g. if underlying computer system and network seem to interfere with or impede performance.  There might be lack of training or reference materials.  So breaking down a problem to its individual aspects helps to reveal areas of improvement pertaining to other levels of the operation; and these “signals” are not necessarily present on an aggregate or collective level.

 

It goes without saying that aggregates serve an important function.  Some people in the organization have no interest in specifics such as the identities of agents.  I believe the term “micromanagement” tends to be overused these days.  Essentially, if there is no interest in specifics, there is little evidence of micromanagement.  The focus might only be on the aggregates.  Aggregates do not fail at a general oversight level since operational decision-making occurs further down.  “I notice that this metric is starting to decline.  What have you done to correct this slide?”  The person or people who need operational guidance are not served at all by an indication that a metric is declining.  Rather they need to determine from the data how to prevent the slide.  If there is a persistent concern that must be rectified, there has to be “management intervention.”  I will suggest that a manager acting on an abundance of data seems most likely to directly change troubling metrics.  We should therefore link the use of data science in production mostly to operations rather than oversight - to enable constructive intervention.

 

Performance analysis is normally applied to individuals in relation to human resources management.  Executives are more likely concerned about aggregate results - unless of course they have great operational interest.  The new performance analysis extends from the technologies of the organization.  I have a graduate certificate in human resources management.  I would say that this function is poorly equipped to deal with performance as it relates to production and through the extensive use of databases.  It is simply not on the radar - not at this time anyways.  So on a practical level, the analytical capacity to deal with metrics on an individual level is likely to exist or emerge from some other organizational function - in my case quality control.  An organization can therefore have this unusual situation where the person checking the quality of work might later, purely due to analytical capacity, provide guidance on agent performance, patterns of market demand, production capacity, and future developments (in the form of projections).  I have encountered a great demand for “specifics” - for actionable and convincing guidance that can be attached to individual performance..

Views: 261

Comment

You need to be a member of Data Science Central to add comments!

Join Data Science Central

Videos

  • Add Videos
  • View All

© 2019   Data Science Central ®   Powered by

Badges  |  Report an Issue  |  Privacy Policy  |  Terms of Service