KPI Weighting in Hotel Groups: Why City and Resort Scores Differ
The first question to ask about a scorecard is not "what did we score." It is: inside that score, how many points of voting power does each indicator hold? KPI weighting is the answer to exactly that question, and in most hotel groups the question was never answered — the weights were simply left equal.
Equal weighting looks harmless. What it hides is this: a training line worth TRY 45,000 a year and a profit line worth TRY 1,800,000 a year cast the same vote in the total. That is not a measurement preference. It is a budgeting error.
What follows builds weight on financial materiality. Every figure below is hypothetical and every calculation is left open so you can rerun it with your own property's numbers. None of them is an industry average or a benchmark.
Equal weighting is not a decision, it is a postponed decision
Leaving the weight set flat is the sentence "we could not agree on which indicator matters more," rendered as a spreadsheet. The decision was not made. It was made invisible.
The consequence stays perfectly visible. Say a hypothetical city hotel closes the period with these scores:
| Indicator | Period score (%) |
|---|---|
| GOP margin | 71 |
| Occupancy | 96 |
| Employee turnover | 118 |
| Guest satisfaction | 92 |
| Training hours per head | 130 |
Equal weighting gives: (71 + 96 + 118 + 92 + 130) / 5 = 101.4.
So a property that landed 29 points short of its gross operating profit target closes the period looking like it beat plan. Performance did not do that. Arithmetic did. Overshooting training hours per head is cheap, and equal weighting hands that cheap indicator the same vote as profit.
Weighting is the work of redistributing that vote. It is not easy work, but it is not deferrable either — defer it and the scorecard decides on your behalf.
What weight actually is: share of the scorecard total
Weight is the maximum contribution an indicator can make to the scorecard total, expressed as a percentage. The formula is unremarkable:
Scorecard total = Σ (indicator score × indicator weight)
Run the same scores through a weight set built on materiality. For the hypothetical city hotel: GOP margin 35, occupancy 25, guest satisfaction 20, employee turnover 15, training 5.
| Indicator | Score (%) | Weight (%) | Contribution |
|---|---|---|---|
| GOP margin | 71 | 35 | 24.85 |
| Occupancy | 96 | 25 | 24.00 |
| Guest satisfaction | 92 | 20 | 18.40 |
| Employee turnover | 118 | 15 | 17.70 |
| Training hours per head | 130 | 5 | 6.50 |
| Total | 100 | 91.45 |
Same property, same period, same actuals: 91.45 instead of 101.4. Ten points appeared without a single performance figure changing. The only thing that changed was how loudly each line was allowed to speak.
That arithmetic is not in the founding text. Kaplan and Norton's 1992 Balanced Scorecard article (HBR 70(1), pp. 71-79) does pull the scorecard out of being a list and into a deliberately small set of measures across four perspectives — but it never discusses weighting at all. Kaplan wrote years later that the omission was deliberate: collapsing the scorecard into a single weighted score was never the design intent (Kaplan, Conceptual Foundations of the Balanced Scorecard, HBS Working Paper 10-074, 2010). Which leaves this: if you are collapsing the scorecard into one total, the weights are yours to choose, not the literature's — and leaving them equal is a choice too.
Before setting weights, you need to know what each indicator hangs off, layer by layer, which is where the goal tree structure comes in. Distributing weight before the tree exists is budgeting without knowing which account the line posts to.
The materiality test: where does this indicator land on the P&L
In accounting, materiality is defined by whether omitting or misstating a piece of information would change a user's decision. The IFRS practice statement on making materiality judgements sets that test for financial reporting. The same test works on a scorecard with one translation: what is a one-point move in this indicator worth on the P&L, in money?
Answering that means knowing which departmental revenue or cost line the indicator attaches to. On the lodging side, USALI defines that line structure — the 12th revised edition carries an adoption date of 1 January 2026 (AHLA announcement).
For our hypothetical city hotel, assume: TRY 180,000,000 annual revenue, 120 rooms, 180 employees.
| Indicator | Where it lands on the P&L | Annual value of a one-point move (assumed) |
|---|---|---|
| GOP margin | Gross operating profit, directly | TRY 1,800,000 |
| Occupancy | Rooms revenue → GOP | TRY 1,226,400 |
| Guest satisfaction | Lagged: repeat stays, channel cost | Not directly measurable |
| Employee turnover | Departmental payroll + recruiting + lost productivity | TRY 378,000 |
| Training hours per head | Training budget | TRY 45,000 |
The arithmetic, spelled out — the value of this table is not the numbers, it is that you can reproduce them:
- GOP margin: TRY 180,000,000 × 1% = TRY 1,800,000.
- Occupancy: 120 rooms × 365 nights = 43,800 available room nights. One point = 438 room nights. At an assumed ADR of TRY 4,000 → TRY 1,752,000 in revenue. At a 70% contribution margin → TRY 1,226,400.
- Employee turnover: 180 employees × 1% = 1.8 people. In this worked example, replacement cost runs TRY 210,000 a head → TRY 378,000. What actually belongs inside that figure is its own argument; the turnover cost lines that reach the P&L are longer than most groups assume.
- Training hours per head: 180 employees × 1 hour × TRY 250 loaded hourly cost = TRY 45,000.
"Not directly measurable" on the guest satisfaction row is not a dodge, it is honesty. For indicators with a lagged payoff, the materiality share is set by judgement rather than calculation — and that judgement needs a written rationale. Turning a score into a financial figure is not possible for every indicator, and saying so beats inventing a number.
City hotel: which indicators dominate in a fixed-cost-heavy structure
In a city hotel trading year-round, most of the cost base is fixed. Building, core team, energy, overhead — these stay put whether or not the house fills. In that structure, profit is acutely sensitive to how much of the capacity gets sold and at what rate.
Derive the raw materiality shares from the money values above. Assume guest satisfaction is assigned a judgemental proxy value of TRY 600,000 to stand in for its lagged effect. Total: 1,800,000 + 1,226,400 + 600,000 + 378,000 + 45,000 = TRY 4,049,400.
| Indicator | Materiality (TRY) | Raw share (%) |
|---|---|---|
| GOP margin | 1,800,000 | 44.45 |
| Occupancy | 1,226,400 | 30.29 |
| Guest satisfaction | 600,000 | 14.82 |
| Employee turnover | 378,000 | 9.34 |
| Training hours per head | 45,000 | 1.11 |
The raw share is the anchor, not the answer. Turnover landing at 9% in a city hotel reflects a labour market that stays open all year: when a front office agent resigns, the time to backfill is predictable. Training landing at 1% does not mean the indicator is worthless — it means it is being measured wrong. Training hours is an input indicator. It carries the weight of a budget line, and no more.
Seasonal resort: how revenue concentration changes the weights
Run the same calculation for a hypothetical seasonal resort: 400 rooms, open 184 nights from May through October, TRY 420,000,000 annual revenue, a season team of 600. Assume most of that revenue lands between June and September.
| Indicator | Calculation (assumed) | One-point move (TRY) | Raw share (%) |
|---|---|---|---|
| GOP margin | 420,000,000 × 1% | 4,200,000 | 48.0 |
| Occupancy | 400 × 184 = 73,600 room nights; 1% = 736 × TRY 5,500 ADR × 70% | 2,833,600 | 32.4 |
| Employee turnover | 600 × 1% = 6 people × TRY 260,000 | 1,560,000 | 17.8 |
| Training hours per head | 600 × 1 hour × TRY 250 | 150,000 | 1.7 |
Guest satisfaction is left off this table on purpose: its lagged effect behaves differently at a resort, and the proxy value assigned to it is a separate decision. The raw shares here are computed across the four measurable indicators only.
Two numbers deserve attention.
First, turnover's raw share moved from 9.34% to 17.8%. Headcount is not the whole reason. In this worked example I used TRY 260,000 per replacement instead of TRY 210,000, because backfilling a line cook in August is not the same transaction as backfilling one in February. Mid-season, labour supply tightens, time-to-fill stretches, and every extra day is paid for in room nights.
Second, concentrating revenue into a short calendar collapses the tolerance for error. In a year-round house, a bad week gets diluted across twelve months. In a 184-night season, a bad week is roughly a twentieth of the capacity and there is no November to make it back in.
Same indicator, two properties, two weights — this is not an inconsistency
Put the two final weight sets side by side. These are not the raw shares; they are the raw shares adjusted by management judgement.
| Indicator | City hotel (%) | Seasonal resort (%) | Why they differ |
|---|---|---|---|
| GOP margin | 35 | 30 | Resort profit rides harder on season demand nobody controls |
| Occupancy | 25 | 25 | First-order revenue driver in both structures |
| Guest satisfaction | 20 | 15 | The city hotel's repeat-stay cycle is short and observable |
| Employee turnover | 15 | 25 | Mid-season replacement cost and time-to-fill are materially higher |
| Training hours per head | 5 | 5 | Input indicator; low in both structures |
| Total | 100 | 100 |
Corporate's first objection is always the same: "then how do we compare properties?" The answer is that you were not comparing them anyway. Two properties reported under one weight set look comparable without being comparable. This is the weighting-side twin of the problem where the same rate means different things at different properties.
Comparability comes from a shared method, not from identical weights. If every property runs the same materiality test, weights that come out different are not an inconsistency — they are evidence the test is working. Making that practical means holding property-specific weights inside a single measurement structure where weights can vary by property.
When department weight and indicator weight collide
So far we have talked about one layer of weighting. Real scorecards carry two: the department's weight, and the indicator's weight inside that department. The two multiply, and what comes out is the effective weight.
Say the hypothetical city hotel weights departments like this: Rooms 40, F&B 25, Sales & Marketing 15, Engineering 10, HR 10.
| Department | Dept weight (%) | Indicator | Weight in dept (%) | Effective weight (%) |
|---|---|---|---|---|
| Rooms | 40 | Occupancy | 40 | 16.0 |
| Rooms | 40 | ADR | 30 | 12.0 |
| Rooms | 40 | Employee turnover | 30 | 12.0 |
| HR | 10 | Employee turnover | 40 | 4.0 |
| HR | 10 | Time to hire | 30 | 3.0 |
| HR | 10 | Training hours per head | 30 | 3.0 |
Two things fall out of this.
Employee turnover sits in two places on the scorecard, for a combined effective weight of 12.0 + 4.0 = 16.0. But three-quarters of that weight lives on the Rooms scorecard. Read the org chart and you would say HR owns turnover. Read the arithmetic and Rooms owns it. When those two answers disagree, the ownership argument surfaces in December, in the bonus meeting.
The second point is quieter. HR's heaviest indicator carries an effective weight of 4% at group level. Pushing training from 100 to 200 — an operationally absurd overshoot — adds 3.0 points to the group score. Pulling GOP margin from 71 to 85 in the same period adds 4.9 points at a 35% weight. An extravagant overshoot on a cheap line is worth nearly as much as a serious recovery on an expensive one.
Effective weight is the number that shapes behaviour. Not the number printed in the weight column.
Holding the total at 100: normalisation and rounding
Raw materiality shares rarely total exactly 100, and rounding to whole numbers usually leaves you at 98 or 101. Discovering that after the scorecard goes live is a bad afternoon.
Work it through with the city hotel's raw shares: 44.45 / 30.29 / 14.82 / 9.34 / 1.11.
| Step | GOP | Occupancy | Satisfaction | Turnover | Training | Total |
|---|---|---|---|---|---|---|
| Raw share | 44.45 | 30.29 | 14.82 | 9.34 | 1.11 | 100.01 |
| Round down | 44 | 30 | 14 | 9 | 1 | 98 |
| Remainder | 0.45 | 0.29 | 0.82 | 0.34 | 0.11 | |
| +1 to the two largest remainders | 45 | 30 | 15 | 9 | 1 | 100 |
The largest-remainder method takes the arbitrariness out of deciding which line gets the spare point. It reads like housekeeping. From an audit-trail point of view it is not.
But 45 / 30 / 15 / 9 / 1 is not an operable set. Training at 1% is noise — whatever it scores, it cannot move the total, so nobody spends effort measuring it. Turnover at 9% cannot raise a signal at a property heading into a second-half staffing crisis.
So the final set deviates from the raw set deliberately: 35 / 25 / 20 / 15 / 5. That deviation is not an error. It is a decision, which is precisely why it has to be written down.
Changing weights mid-year: legitimate when, manipulation when
One question separates the two cases: was the change made before seeing the result, or after?
| Situation | Legitimate | Condition |
|---|---|---|
| A department opened or an outlet closed | Yes | Prospective, from the next measurement period |
| An indicator's definition or formula changed | Yes | Old and new sets reported separately |
| The accounting line structure changed (e.g. a USALI edition change) | Yes | The transition is noted on the scorecard |
| An indicator is running badly, so lower its weight | No | That is score correction, not weighting |
| A retroactive fix after the period closed | No | It makes the scorecard unauditable |
| A property called it unfair, so rebalance | No | Invalid unless the rationale comes from the materiality calculation |
One rule holds all of this up: weight changes are prospective only, and they are approved by someone who is not scored on them. A department that sets its own weight is not running a scorecard. It is running a budget negotiation.
Documenting the weighting decision
The most fragile part of a weight set is that six months later nobody remembers why that number was chosen. A rationale record fixes it, and it fits on one page.
| Field | Content | Example |
|---|---|---|
| Indicator | Exact name from the scorecard definition | Employee turnover |
| Raw materiality share | Value out of the calculation | 9.34% |
| Final weight | Value in force on the scorecard | 15% |
| Reason for deviation | Why it left the raw share | Replacement cost is lagged; the raw share reflects one period only |
| Basis of calculation | Assumptions used | 180 employees, TRY 210,000 per replacement |
| Decision date / approval | Who, when | 2025-12-12, GM + Finance |
| Period of validity | Which periods it covers | Full year 2026 |
| Next review | Calendar | 2026-10 |
Without this record, the weighting argument restarts from zero every year, and the loudest department wins it.
Testing how sensitive the scorecard total is to the weight set
The last job before a weight set goes live is running the same actuals through several sets. Take the period scores above (71 / 96 / 92 / 118 / 130) and score them four ways.
| Weight set | GOP | Occ | Sat | Turnover | Training | Scorecard total |
|---|---|---|---|---|---|---|
| Equal | 20 | 20 | 20 | 20 | 20 | 101.40 |
| Seasonal resort set | 30 | 25 | 15 | 25 | 5 | 95.10 |
| City hotel final set | 35 | 25 | 20 | 15 | 5 | 91.45 |
| Raw materiality set | 45 | 30 | 15 | 9 | 1 | 86.47 |
Not one performance figure moved. The total travelled between 86.47 and 101.40 — a spread of roughly 15 points.
A workable threshold: if that spread is wider than the band you use to award a bonus or trigger an intervention, then the weight set is running the scorecard, not performance. When that happens, either widen the band or make the weight set defensible. Do not ignore both.
An annual weight review calendar
A weight set is a living document, but not a permanently open one. It is revised once a year, tied to the budget calendar.
| Period | Task | Owner |
|---|---|---|
| September | Draft next year's revenue and cost structure | Finance |
| October | Refresh the materiality calculation per indicator | Finance + HR |
| November | Normalise, write the deviation rationales | Scorecard owner |
| December | Sensitivity test and sign-off | General Manager |
| January | New set in force | — |
| Mid-season | Only for structural change, prospective | GM approval |
The single purpose of this calendar is to confine the weighting argument to the right month of the year. A group debating weights in August is not debating weights.
Conclusion
Weighting looks like spreadsheet work. It is budget work. Three sentences carry it.
Equal weighting is not neutral. It puts the cheap indicator level with the expensive one and pulls the scorecard systematically toward optimism.
Weight is anchored in financial materiality — what a one-point move is worth on the P&L. That calculation is not available for every indicator, and where it is not, the decision is made by written judgement instead.
The same indicator carrying different weight in a city hotel and a seasonal resort is correct. What has to be shared across the group is not the weight. It is the test that produces it.
The work that follows is joining weighting to scoring: the right weight multiplied by the wrong scoring method still produces the wrong total.
Deciding which indicator to start the materiality calculation with is usually a day's work in most groups; the hard part is not the arithmetic, it is writing down the reasoning. We work on the measurement side of hospitality and we enjoy talking to teams opening this argument for the first time — if you would like help pinning down where to start, get in touch.


