The ratio is a design principle, not a target

The 90/10 rule does not mean every LMS should achieve an exact percentage. It is a useful challenge: routine work should happen through dependable rules, while skilled time should be reserved for exceptions, decisions and improvement.

If administrators spend most of the week repeating the same enrolments, exports and corrections, the organisation is paying expert rates for manual processing. If nobody spends time on the remaining ten percent, the automation will slowly become unreliable.

The ninety percent should be quiet

Routine enrolment, reminders, completions, certificates and standard reports should follow agreed rules. Quiet does not mean neglected. It means the platform has been configured, documented and monitored well enough that repeatable work stays repeatable.

A quiet process has a defined trigger, an expected result and a way to reveal failure. A nightly user feed is not truly automated if nobody checks whether it ran. A compliance reminder is not dependable if ownership of the audience rule is unclear.

  • New starters receive the correct learning from an agreed data source
  • Role or audience changes update access without manual intervention
  • Completions and certificates follow one controlled definition
  • Scheduled reports use documented fields and recipients
  • Failed jobs, missing data and exceptions create visible alerts

The ten percent earns its keep

Good administration concentrates on exceptions and improvement: investigating a failed integration, changing an audience rule, preparing a new programme, analysing a reporting question or removing the cause of repeat support requests.

This is the work that requires context. It involves deciding whether the rule is still correct, understanding the effect on learners and other systems, testing a change, explaining the decision and leaving the platform easier to operate next time.

Find the manual work that has become invisible

For one month, ask everyone who touches the LMS to record repeat tasks, including work kept in personal reminders and spreadsheets. Capture the trigger, frequency, time taken, systems involved, error risk and consequence of delay.

Then group the list by root cause. Ten separate requests may all come from one missing audience rule. Three monthly spreadsheets may exist because a report definition was never agreed. Fixing the cause creates more capacity than processing each item faster.

  • Separate recurring requests from genuine one-off work
  • Identify hand-offs where data is copied between systems
  • Mark tasks that depend on one person's memory
  • Record corrections caused by unclear ownership or definitions
  • Prioritise by frequency, risk and avoidable effort

Automate in the right order

First agree the rule. Then clean the data needed to apply it. Next, test the rule against common cases and awkward exceptions. Only then automate it, with a visible failure route and a named owner.

Automating an unclear process makes inconsistency faster. Automating poor data makes errors less visible. The best early candidates are frequent, predictable, low-judgement tasks whose failures can be detected and reversed.

A falling ticket count can be a success

Managed administration should not create dependence. As repeated faults are fixed and routine work becomes reliable, capacity should move towards reporting improvements, cleaner learner journeys, integration resilience and work the internal team could not previously reach.

Review the balance every quarter. If routine demand is not falling, ask which recurring causes remain. If it has fallen, use the released capacity deliberately. The goal is not an empty support queue; it is a platform that needs less rescue and creates more value.