HRMIM
Far Below Target: Bad Performance or a Badly Set Target?

Far Below Target: Bad Performance or a Badly Set Target?

Timuçin Diş · Strategy Consultant· 15 July 2026· 11 min read

The July scorecard is on your desk. One indicator at one property reads 34 out of 100. Far below target. Before you pick up the phone and call the department head, give it a minute.

That number can mean four completely different things, and each one calls for a different action. KPI outlier analysis is the work of telling them apart: diagnose before you intervene. This piece puts that diagnosis into a decision matrix, then covers how to be sure that when you compare 2025 with 2026, you are actually comparing the same thing.

An indicator far below target: why the first reflex is wrong

The first reflex is almost always the same — find the owner, demand an explanation. That reflex carries three costs, and all three accumulate quietly.

First, you dig in the wrong place. The department head spends two weeks combing through an operation and finds nothing, because the problem was never in the operation.

Second, you spend credibility. Raise three false alarms in a year and nobody runs for the fourth.

Third, and most expensive: you turn the scorecard into a blame instrument. Properties start managing the number instead of fixing the input, and your measurement system quietly corrupts itself.

The rule is simple. Classify the deviation before you act on it. Classification takes a day. The wrong action takes a quarter.

Three sources of deviation: operations, target, definition

Start with the term. An outlier is a value far enough from expectation that it demands an explanation. NIST/SEMATECH puts it the same way: an outlier is an observation that appears to deviate markedly from the rest of the data — which makes it a question, not a verdict (NIST/SEMATECH e-Handbook).

On a scorecard, that question has three structural sources:

  • Operations. You genuinely had a bad period. The indicator measured correctly and the result is poor.
  • Target. The indicator measured correctly, the result is normal, and the target was wrong from the day it was set.
  • Definition. The indicator did not measure what it was supposed to. The formula, unit, scope or data source is faulty.

There is a fourth case that is not a structural source but is the one most often mistaken for the others: a one-off external event. A regional power cut, a road closure, a three-day outage on a booking platform. The deviation is real, the measurement is right, the target is right — but something happened that will not happen again. Treat it as operational failure and you solve a problem that does not exist; treat it as a bad target and you lower a target that was fine.

All four look identical on the report: indicator below target. What separates them is the question you ask next.

The decision matrix: which question points to which source

The matrix is built on one principle: check the cheap and common causes first. A definition and data check takes an afternoon, and in practice it is the source that turns up most often. A root-cause investigation takes weeks. Run the sequence backwards and you do the expensive work for nothing.

SourceDiscriminating questionEvidence you expectFirst action
Definition / data faultDid the formula, source or collection method change this period?Change record; different formulas for the same indicator across properties; scope shiftFix the data, restate the period, do not void the result
One-off external eventIs there a non-recurring event that explains it?Dates line up; event scope overlaps indicator scopeAnnotate the period, leave the target alone, log the exclusion from trend
Badly set targetPresent since period one, and missed group-wide in the same direction?Continuity from period 1; single-direction miss across propertiesMove the target into the revision process (rules below)
Operational problemAre all three answers above "no"?Related indicators moving together; concentrated in one property or departmentOpen a root-cause investigation, assign owner and date

Do not reorder it. Ask the four questions top to bottom and stop at the first yes. These are the points where the sequence usually snags:

How do I tell a definition fault from an operational problem? Look at the raw inputs. In an operational problem the raw data is bad too — fewer guests, more complaints, more leavers. In a definition fault the raw data is normal and only the calculated number is broken. If the raw inputs and the score disagree, the formula is the suspect.

Is the deviation at one property or all of them? If it is one property, look at that property's operation or its data collection. If the same indicator misses in the same direction across the whole group, it is almost never operations — it is the target or the definition.

Has the deviation been there since the first period? If it has, the target goes to the top of the suspect list. A genuine operational decline has a starting point; a bad target is missed from the moment it is born.

Did the formula or the data source change this period? If yes, close that question before you look at anything else. And if you are not keeping a change record, invest in the change record before you invest in outlier analysis.

The signature of an operational deviation

A real operational deviation has a recognisable signature, and it carries at least three of these four marks.

Co-movement. Related indicators move together. If turnover is up, training completion is down and guest complaints are up. An indicator that swings hard without disturbing any neighbour is a measurement suspect.

Timing. The month the deviation starts lines up with a known operational event: a manager change, a refurbishment, a supplier switch.

Concentration. The deviation sits in one property or one department instead of spreading across the group.

Reversibility. Remove the cause and the indicator recovers. If it does not, the diagnosis was wrong.

Once the signature holds, go after root cause. What works at that stage is a structured conversation, not a spreadsheet — the discipline behind the questions you ask to get to root cause applies here too: open wide, and do not lock onto one hypothesis early.

The signature of a badly set target

A bad target is easier to read, because it is consistent.

It is missed from the start. Below target in period one, below target in period two. Nothing changed in the operation; the target was never reachable.

It misses group-wide. If five properties out of five miss the same indicator in the same direction, five separate operational problems is a far less likely explanation than one bad target.

It rests on a wish, not on capacity. Ask how the target was set. "Fifteen percent above last year" is not a target-setting method; it is a way to end a meeting. If the target is not grounded in a capacity calculation, a resource plan or a documented benchmark, treat it as suspect.

The direction is wrong. The sneakiest form of a bad target: the number is right and the direction is not. Score a lower-is-better indicator like turnover with a higher-is-better formula and the scorecard lies to you systematically. The deviation that appears when target direction is set wrong looks almost identical to an operational one — but the action is entirely different. You fix the formula and speak to nobody.

The signature of a definition problem

A definition problem has a one-line signature: the raw data is normal and the calculated number is not.

A hypothetical but familiar example. Two properties both report "employee turnover." Property A puts headcount at period start in the denominator; Property B uses the period average. Property A counts seasonal staff among leavers; Property B does not. The operations are almost identical, the reported numbers are far apart. This outlier is not a performance finding — it is an accounting difference.

Three more faults from the same family:

  • Scope shift. The indicator covered three departments last year and five this year. The number moved; performance did not.
  • Source change. The data used to be collected by hand and now comes out of a report. The two methods do not produce the same value.
  • Unit and period shift. The indicator is defined monthly, the data arrives quarterly, and someone divides by three. Seasonality dies in that division.

All of this is preventable at setup, at the step where the shared definition set gets locked. If it was never locked, this has to be the first question you ask every single month.

A single-period deviation versus a trend

One period is a question. Three periods are a finding. A system that does not draw that line raises an alarm every month, and none of the alarms get taken seriously.

Set a practical threshold and write it down: a single-period deviation is investigated but never justifies a target revision; three consecutive periods missing in the same direction counts as structural and triggers a review of the target or the definition.

You can also run a formal outlier rule. Grubbs' test is a standard method for detecting a single outlier in a data set (NIST/SEMATECH e-Handbook). Use it in hospitality with two caveats: it assumes enough observations and an approximately normal distribution, and monthly hotel data carries heavy seasonality. Treating August and February as members of the same distribution is mathematically tidy and operationally meaningless. Where seasonality is strong, compare against the same period in prior years, not against a rolling annual mean.

Three conditions that make period-over-period comparison meaningful

Comparing 2025 with 2026 is not a matter of putting two numbers side by side. Miss any of three conditions and the gap between them is noise, not performance. The same logic governs comparison between properties — the question of when a comparison is actually meaningful breaks at exactly the same point in industry averages.

Definitions must hold

If the indicator kept its name but changed its formula, the two periods are not comparable. Formula, numerator, denominator, scope and unit — all five have to be stable. If the definition changed, the only honest options are to restate history or to break the series. Applying the change quietly and drawing a trend line over it is the most common reporting error there is.

Weights must hold

Indicator-level comparison is unaffected by weighting, but the total scorecard score is affected directly. If you changed weights mid-year, do not compare the 2025 total with the 2026 total; compare indicator by indicator. When the weight set moves, you cannot separate how much of the score change came from performance and how much came from the weights — and a gap you cannot decompose should not be put in front of a general manager.

Periods must be comparable

Three things have to match. Measurement frequency: move an indicator from monthly to quarterly and the series is broken. Scope: a property opened mid-year or a department taken over changes the total, so run the comparison on a constant scope — the units that existed in both periods. Calendar: convert to a rate rather than comparing a 28-day February with a 31-day July in absolute terms. In a seasonal business the safest comparison is always the same period last year.

Restating history after a definition change

Definition changes are inevitable — indicators mature. The problem is never the change, it is the undocumented change. The decision reduces to four scenarios:

SituationWhat to doHow it should read on the scorecard
Definition changed, raw history availableRestate history on the new definitionContinuous series, "restated" footnote with a date
Definition changed, raw history goneBreak the seriesTwo series, visible break, no trend arrow across it
Only the target changedNo restatementActuals comparable, scores not
Weights changedDo not compare totalsCompare indicator by indicator, report totals separately

One rule holds it together: the person who changes the definition keeps the record, not the person who runs the comparison. The change record has to say what changed, when, who approved it, and whether history was restated.

Who investigates an outlier: ownership and escalation

An outlier with no owner does not get investigated. "We'll look into it" is not ownership. Write down in advance who investigates what, within how long, and what they must produce. The thresholds below are an example set, not an industry benchmark — calibrate them against your own group's deviation distribution.

DeviationInvestigated byDeadlineOutput
10-25% below targetIndicator owner5 working daysShort note: source class + action
25-50% below targetDepartment head5 working daysOutlier form + root cause
More than 50% below targetProperty general manager10 working daysFormal review + group notification
Three consecutive periods, or group-wideGroup performance owner10 working daysDefinition / target review

Escalation runs sideways as well as up. If the indicator owner classifies the cause as a definition fault, the case leaves operations and moves to the owner of the definition set. Put that handoff in writing, or the outlier disappears between two desks.

What an outlier report needs in order to produce a decision

Most outlier reports get read and closed. A report that produces a decision carries these fields:

  • Actual, target and deviation — in absolute and percentage terms
  • The prior period and the same period last year
  • The same indicator's value at the other properties
  • Source classification: operations / target / definition / one-off event
  • The reasoning behind the classification — one sentence is enough, but it cannot be blank
  • Owner, action, due date
  • A record of any definition or target change

Apply a test: if the report does not end in a decision sentence, it is not a report, it is an archive. What Ridgway wrote in 1956 still holds — a performance measurement system built without accounting for the behaviour it produces will distort the very thing it set out to measure (Ridgway, Dysfunctional Consequences of Performance Measurements, Administrative Science Quarterly 1(2), 1956, pp. 240-247).

Which is why an outlier report should never stand alone. To classify the source of a deviation you need the prior period, last year and the other properties in front of you at the same time; seeing outliers and comparisons in the same report set takes diagnosis from days down to hours.

Rules for turning a deviation into a target revision

Target revision is a legitimate tool and a very easy one to abuse. Four rules cover it.

1. The justification follows the source class. Only a badly set target and a structural change in capacity justify a revision. An operational deviation does not change the target — it changes the operation. A one-off external event gets annotated on the period, nothing more.

2. Revisions apply forward. Rewriting a past period's target to repair a score is cosmetics, not measurement. A new target takes effect from the period the decision was made.

3. The old target stays on the record. If you cannot see that a target was lowered twice in one year, you cannot audit your target-setting process at all.

4. Approval sits one level above whoever set the target. Where people revise their own targets, a target is just another word for a forecast.

One rule on timing: keep mid-year revisions exceptional and push the real correction into the annual target-setting cycle. Your outlier records are the most valuable input you have when next year's targets get set — they have already told you which targets were wrong to begin with.

Conclusion

An indicator far below target is a question, not an answer. Close it in four steps: check the definition and the data, check for a one-off external event, test whether the target was set correctly in the first place, and only then look at operations. That is the order. Reverse it and you do the expensive work for no reason.

On the comparison side the rule is blunter. If the definition, the weights, the measurement frequency and the scope are not all stable, the gap between 2025 and 2026 is not performance. If any one of the four moved, restate history or break the series — but never draw a trend line over it in silence.

The best time to build this discipline is before you see the deviation. Write the thresholds, the owners and the deadlines during a quiet period, so that when the July scorecard reads 34, the only thing up for debate is the diagnosis and not the process itself.


If you carry measurement responsibility across a multi-property group and you are tired of relitigating the same outliers every month, a conversation with the HRMIM team may be worth an hour. We can walk through your indicators, how your targets get set, and how your period comparisons hold up — get in touch and we will start where it hurts most.

Get New Posts by Email

We will email you when a new blog post is published.

Loading the form…

Related Posts

KPI - Process Performance Assessment: Every Property, Every Department, One Scorecard

KPI - Process Performance Assessment: Every Property, Every Department, One Scorecard

Everything the previous six posts covered — the goal tree, indicator validity, weighting, measurement frequency, target direction and outlier analysis — can be solved one at a time. Holding them in one structure is the hard part. KPI - Process Performance Assessment evaluates company performance by property and department, monthly, across a defined period: a four-layer goal tree, weights that can differ by property, per-indicator frequency, five scoring methods and a 10-page report set.

29 July 2026 · 8 min read

Revenue Up, Turnover Down, Stock On Target: Scoring KPI Direction

Revenue Up, Turnover Down, Stock On Target: Scoring KPI Direction

Revenue above target is good. Turnover above target is bad. Stock days should sit close to target, in either direction. Three indicators, three scoring directions — and no single formula scores all three correctly. This piece works through target direction and the scoring methods that follow from it, with hospitality examples: actual/target, target/actual, higher-is-better, lower-is-better and closeness-to-target. Including why an indicator with no target should not be scored at all.

1 July 2026 · 12 min read

Your Budget Is a Measurement Instrument: What Targets Do to Behaviour

Your Budget Is a Measurement Instrument: What Targets Do to Behaviour

The moment a budget target is set it stops being only a measurement and starts changing the behaviour it measures. Goal-setting research shows specific, difficult goals raise performance — and that the effect reverses when commitment lapses or ability limits are reached. Attach a bonus to a budget target and what you get is negotiation, not measurement. This piece treats the budget as a measurement instrument: the behavioural effect of targets, budgetary slack, and what a variance captures.

23 September 2026 · 14 min read

Let's work with HRMIM

Explore our platform free of charge.