Separating a skill gap from a work-system fault
5 min read
Lesson objectives
- Distinguish a trainable knowledge/skill gap from workflow, tool, opportunity and incentive constraints using observed task evidence.
Please sign in to use all the tools
Once you sign in, the reading tools are yours: larger or smaller text, typeface, line spacing, focus mode, highlighting and bold, and notes saved to your account and gathered in My study.
Once a diagnostic brief is accepted, do not begin with “Which course should we design?” Ask which task failed and what prevented its successful performance. Learning activity can change what a person knows and can practise, but it cannot repair a disabled system field, missing access or contradictory instructions. Performance-gap analysis separates an outcome into behaviour and working conditions. The specialist specifies a task standard, compares it with a sample of actual work and tests alternative explanations before recommending learning. The explanation most convenient for the team must not be the only hypothesis.
Start with a clear unit of observation. In complaint handling it might be one closed complaint record; in a shift handover it might be one handover form. Write down the task standard approved by the process owner: required fields, timing, customer confirmation and permitted exceptions. If the standard is undocumented, do not invent it and then describe workers as failing it. Propose a provisional standard for discussion and mark it as unapproved. Sample ordinary and exceptional work. A sample containing failures alone helps reveal faults, but cannot show what distinguishes successful work or establish a general failure rate.
Build an evidence log with four columns:
- observation
- source and period
- possible interpretation
- and evidence that could disconfirm it
For example, “nine records lacked a reason code during a screen outage” suggests a tool fault; the incident log and a code-entry test after repair would help separate that from other causes. “Four agents could not explain the difference between two codes when asked” supports a possible knowledge gap, but first check whether the code definitions are available and current in a job aid. Do not treat “I do not know” as identical to a lack of time or permission to use what is known. Stronger claims need evidence of task performance under suitable conditions, not only general opinion.
Test causes in a practical sequence. First, does the worker know the requirement and demonstrate the task when given a fair opportunity? A short task trial or simulation probes knowledge and skill. Second, can the worker perform it in normal work with the available tools, time and authority? Observation and system logs reveal constraints. Third, are priorities and feedback consistent? If call speed alone is rewarded and longer clarification is penalised, clarification training by itself may not transfer. These questions neither diagnose personality nor prove an individual's psychological motivation; they test conditions surrounding a defined task. A required policy, incentive or system change goes to its authorised owner.
Causes may coexist. In the missing-record example the outage may account for nine cases, yet interviews and a short task trial might show that some agents confuse two codes when the system works. “No learning because a fault exists” is as hasty as “train everyone because errors exist.” Partition the problem: cases blocked by the system, cases involving uncertainty about codes in working conditions, and cases with unknown causes. Link each group to an intervention, owner and verification evidence. System repair belongs with operations or technology; a comparison job aid and coding practice may belong with L&D; unexplained cases call for another sample rather than a blanket programme.
A percentage from a single sample is not enough. Twelve incomplete records among 80 means 15% of those 80. If there are three among 40 the next week, that is 7.5%, but the comparison alone does not establish an intervention effect if case severity, sample selection or workload changed. Check that “incomplete” and the sampling method remain constant and note when the outage occurred. Report counts with percentages: one of four is 25%, but it is a weak basis for a large decision. Without a usable baseline, state the inference limit and make improved measurement the first improvement step.
The output is a bounded action decision rather than a decorative list of causes: which explanation the evidence supports, how strongly, what response fits, what remains unknown and when to review. Rank causes by impact and whether the proposed action can change them. Do not put “training” in every row. If authorised access to recordings or individual records is absent, use aggregated information or arrange new observation with the appropriate permission. Do not cross that boundary in the name of analysis. Escalate a conflict between a supervisor's standard and current written instructions before judging workers against either. This protects fairness as well as organisational time.
Lesson summary
Distinguish a trainable knowledge/skill gap from workflow, tool, opportunity and incentive constraints using observed task evidence.