Although I deal with many different types of metrics, I believe they can be generally classified as follows: 1) time use; 2) alignment; 3) production; 4) performance; 5) service; 6) and market. In this blog, I will be providing some comments pertaining to each. Although I have yet to encounter any myself, I am certain that there must be text books on the issue of operational metrics and how to make use of them. However, I personally developed nearly all of those that I use. Although I do not feel that I can freely share elaborate details, techniques, or methodologies, I recognize that young people entering the field should probably have some familiarity with metrics. I also feel that data science specifically cannot survive in a purely academic setting; it has to be highly applicable to business and real-life problems. So to promote the industry I would like to offer some personal insights and experiences.
Time use - or I sometimes call it time contribution - is not a measure of how much a person works. It is a measure of how long a person occupies a place in a production environment. Rarely are people paid by piece these days: e.g. $5 for each scarf produced. Instead they receive a salary or hourly rate. Perhaps many small business owners think of time use primarily in relation to the costing of labour. But there are many good management reasons to follow time use. Time use allows for the strategic placement of people and machines. By “placement,” I mean placement in time: e.g. “Jeffery will spend 3 hours doing job #1, 2 hours on job #2, and 1 hour on job #3.” Conversely, time use answers the question, “What was Jeffery doing today?” From an analytical standpoint, time use can be considered a production input; this makes it possible to evaluate how well time gives rise to desirable production outputs.
Closely related to time use is alignment. Conceptually speaking, further on the issue of time, it is possible to determine how resources are distributed over different time periods. The distribution of resources should coincide or be aligned with the demands of the market. Alignment represents a challenging undertaking if the market is unpredictable. Further, there is usually a limit to how well or often internal systems and people can be reassigned to cope with sudden changes in the market. So it is genuinely worthwhile to anticipate the market. It is rewarding and humbling “trying to figure people out.” I am not saying that people are strange and therefore difficult to figure out. I mean they are behaving perfectly normal; so it is important to gain the ability to predict their behaviours. Metrics pertaining to alignment would be of great use to call centres.
Production is quite different from performance. Production is an outcome - the things produced. Performance is more related to the processes involved in production. An extreme example would be like having 1,000 people working 10 hours each to produce 10 lamps. Production is the 10 lamps. Having 1,000 people working 10 hours to reach the production outcome is a sign of poor performance - unless these are extremely fine lamps. When I started on my job, in terms of the data that I gathered and used, I was highly production oriented. I kept track of what people produced over different time periods. Although I checked trends and studied seasonality, my need for analytics was limited. To emphasize my point, I will mention that I inherited a mostly paper-based system. Among my main tools were pens, notepads, paperclips, and a couple of rulers.
Production isn’t too meaningful without operational attribution - i.e. the knowledge of what went into production. This is the essence of what I said when I proposed a paperless environment focusing on the accumulation of electronic data. If attribution is important, then paper is restrictive. The largest chunk of metrics that I handle deals with performance. Performance metrics can help a data specialist protect performers. Without these metrics, critics are more likely to form conclusions that are biased, nonsensical, destructive, and of course unproductive. Literally, productive people might be purged from organizations that fail to maintain appropriate performance metrics. Further, these organizations are likely to produce products that are disassociated from the needs of the market. Efficiency is a type of production metric. Quality can be a type of efficiency metric. If 50 percent of Bob’s production is deficient, this is highly inefficient since it necessitates additional production.
Service metrics are interesting in that they don’t necessarily lead to greater production or improved performance. They lead to conformance to a preset standard. Quality can also be considered a type of service metric depending on the situation. (As I mentioned earlier, improved service does not necessarily lead to more production.) The term that I use to describe conformance that is unrelated to production is compliance. Service is about being compliant to specific standards of service. Data is needed to confirm compliance. In relation to call centres, a service standard that I found on the internet is as follows: 80 percent of calls should be picked up in the first 20 seconds. I feel that anyone who has called a business and waited for a human voice would likely agree that 20 seconds is a long time. So service is about improving the customer experience and bolstering a company's base of support. (Not to mention that improved service can positively affect sales.)
Finally, there are many metrics relating to the market. This can include feedback from clients. But I am thinking about something more immediate and pressing. Recall my explanation about alignment. Alignment is most useful when it is possible to predict the market. Those that have read my blog on deconstruction are aware of my suggestion to divide market stats by different days, weeks, and months. A retailer for example wants to be adequately staffed - not understaffed or excessively staffed - to ensure smooth operations on a particular Saturday or Sunday. A call centre wants to have just enough agents to handle the load from 2 PM to 5 PM on a particular Friday. The challenge is specific - and the consequences of failure are usually apparent - putting into question staffing levels, management skills, and even hiring practices. Don’t believe that an analyst is free of responsibility just because he or she isn’t at the front line.
Are there other types of business metrics? This blog provides an incomplete list of categories. In particular, there are many metrics associated with finance, accounting, and investment that I haven’t mentioned: e.g. allocations, expenses, and revenues. I also left out risk measurements. Most people that analyze business data know that sales is a big issue. I don't study sales per se but market demand for product - an aspect of the market. It might be difficult for those that specialize in dollar-unitized metrics to focus strictly on operations - e.g. for a management accountant, just due to the predisposition to express and analyze everything in relation to costs. I suggest however that a person is “efficient” not because this employee is paid less but as a result of fewer mistakes, greater reliability, and superior pace. Something outside my current scope of responsibility would be disability costs of workplace accidents.
Employee compensation and costing are best kept outside operational analytics. It is like a competitive sport - in that salary negotiations should take place outside the game. This is because the game is “mission critical.” Each loss leads to encroachment by the competition. Moreover, compensation should be more a matter of operational success or failure rather than assignment. Or perhaps this is my rather harsh perspective on an employee’s contribution. An employee isn’t paid to do a job but rather what that job does for the company. If the job is done well but he company goes into the gutter, it is a matter of accounting that this individual be paid the agreed amount. If the company goes into the gutter, to me the job failed to be evaluated in relation to the best metrics - namely, those that could have helped the company win the game.
I try to deliver visibility and control. I don’t have control that I give to others. But rather, I do my best to ensure that people’s actions lead to their desired outcomes. It is control in the sense of a motor action conforming to one’s intent. However, I am also responsible for “posing” the metrics so that business problems are expressed in the proper manner. For me, this means being fair and honest. Fairness and honesty can be difficult if the underlying metrics are unfair and dishonest. The construction and selection of metrics can influence the decisions that people make. Who gets blamed if they make bad decisions in such a scenario? That is right. There is such a thing as choosing metrics to ensure balance. Mistakes are less likely if there are many metrics taken from the ecology of metrics available - and ensuring that systemic placement of these metrics in the production environment is understood.
It is important not to get caught in a mindset that effectively imposes one’s biases over the ecological system. Everything is connected to everything else unless one is willing to ignore those connections in favour of a skewed perspective. A professional dealing with metrics knows that organizations do not necessarily question or critically examine their metrics. Sometimes this is due to lack of familiarity; it might be inexperience; or maybe there was never a focus on critical thinking skills in the education of data specialists. Living within the phenomena can mitigate the risk of using a faulty lens. For instance, a lens preoccupied with salaries will probably reduce the cost of salaries - but not make the game easier to win. A lens focused on cost reduction is problematic in that costs are most reduced if the company stops operating. Some entrenched businesses lenses can lead an organization astray. Living in the ecology - being engrossed in and sensitive to multifarious metrics - can help to maintain an open perspective.