SLM: Introduction
We start this introduction with some examples of traditional metrics for measuring performance.
Examples
In what follows, we give some examples of metrics that are traditionally used in SLM in a number of contexts. First, some callcenter metrics:
Average time elapsed for calls on hold
Average number of interactions with customer before problem solved
Number of escalations in a given period
The second set of examples are related to IT helpdesk services:
Average time taken to acknowledge a problem ticket
Average resolution time for an incident (with priority 1)
Percentage of incidents with resolution time above threshold
The third set of examples has to do with IT operations:
Percentage systems uptime
Unexpected systems downtime
Mean time between system failures
We will refer to these examples in later posts. We now turn to some characteristics of the way most service levels are implemented.
Consider the following (fake) recommendation for a hotel room:
The guaranteed average room temperature is 20 degrees Celcius.
Even with a money back guarantee, I would not rent the room. Suppose its 40 degrees during the day and 0 degrees during the night? The average temperature may still be 20 degrees…
Averages
When dealing with customer interactions (calls), IT incidents and problems, etc., one usually deals with a large number of instances that are observed. Hundreds or even thousands of calls may be logged in a call center per day. In order to aggregate the information on these calls, one usually takes averages over a given period of time. This is reflected in the examples given before.
Averages, however, are only a first order representation of a set of instances. One bad instance can easily be compensated by a good instance resulting in a good average. In other words, if the average on-hold time for a callcenter is 2 minutes, it may well be that I have to wait 10 minutes. If only at another time 5 calls are answered within a second, the average is back to 2 minutes.
Averages and Sums
One often deals with lots of data points, for instance when considering incidents or service calls to a helpdesk.
Usually, a (small) number of important metrics are selected and calculated on a regular basis (monthly, quarterly). The metrics are defined as averages or sums, such that one number is characteristic for the set of incidents, calls, problems, … that it covers.
An important metric is often called a Key Performance Indicator.
Service Level
Traditionally, a service level is either defined as the result of one KPI or as the (weighted) average of a small set of KPIs, most often expressed as a percentage. This way, it is at best an average of a limited number of KPIs.
In practice, one often measures the service level and takes a baseline which is then used as benchmark.
Service Level Agreement
A contract with a service provider typically includes several services. An agreement about each of these services and their respective targets and quality parameters is called a Service Level Agreement (SLA). Often, though, the term SLA is used not to denote the contract as such, but rather the Service Level calculated for all the services is in the contract.