Revenue Up, Turnover Down, Stock On Target: Scoring KPI Direction
Three indicators sit side by side on a departmental scorecard: revenue, employee turnover rate, stock days. Each has a target. Each has an actual. The system applies the same formula to all three — actual divided by target — and returns three percentages.
One of those percentages is right. The other two pull the scorecard in the wrong direction.
When revenue lands above target, something good happened. When turnover lands above target, something bad happened. When stock days land above or below target, something bad happened either way — one ties up working capital, the other interrupts service. The same piece of arithmetic cannot tell those three cases apart.
This piece covers target direction and the five scoring methods that follow from it. Every number below is hypothetical and written out in full so you can reproduce the arithmetic yourself.
Three indicators, three directions: revenue, turnover, stock days
Start with the June food and beverage numbers for a hypothetical city hotel. The figures are illustrative only — they are not drawn from any industry data set or benchmark.
| Indicator | Unit | Target | Actual | Variance | Is that actually good? |
|---|---|---|---|---|---|
| Revenue | TL | 10,000,000 | 11,000,000 | +1,000,000 | Yes |
| Employee turnover rate | % | 28 | 35 | +7 pts | No |
| Stock days | days | 12 | 18 | +6 days | No |
All three variances are positive. All three actuals sit above target. Only one of them means the department did well.
Run actual/target across this table and you get 110% for revenue, 125% for turnover and 150% for stock days. The scorecard hands its highest scores to the two line items where the property performed worst. The data is fine. The direction of the formula is not.
What target direction is, and why it is decided before scoring
Target direction is the rule that says which side of the target creates value. It is a property of the indicator itself — not of the period, the property, or the manager being reviewed. Revenue is higher-is-better in every hotel, every month. Turnover is lower-is-better in every hotel, every month.
There are three directions.
| Direction | Definition | Cost of variance | Hospitality examples |
|---|---|---|---|
| Higher is better | Above target creates value, below target loses it | One-sided | Revenue, RevPAR, training completion rate, guest satisfaction score |
| Lower is better | Below target creates value, above target loses it | One-sided | Turnover rate, complaint volume, energy consumption per room |
| Close to target is better | Variance in either direction is a loss | Two-sided, asymmetric | Stock days, staff-to-room ratio, budget variance, forecast error |
Direction has to be fixed before the scoring method is chosen, because direction selects the formula. Work in the other order — build the formula first, then pour indicators into it — and you have bent the meaning of the indicator to fit the geometry of the arithmetic. That is scoring what is easy to compute rather than what you set out to measure.
The classic balanced-scorecard argument makes the same point from the other end: it ties every indicator to a strategic objective, and once that link is drawn the direction is already implied. The Balanced Scorecard Institute's overview sets out how the link is meant to be made.
Method 1 — Actual / Target
The default almost everywhere. Score = Actual ÷ Target.
For revenue: 11,000,000 ÷ 10,000,000 = 1.10 → 110%.
Where it works: higher-is-better indicators with a natural floor at zero and no natural ceiling. Revenue, RevPAR, total training hours, upsell income.
The method carries one assumption that is rarely written down anywhere — that the indicator creates value as it grows. That assumption is false for turnover rate and false for stock days. The formula has no way of knowing it is false. It divides and returns a number.
One more caution. As the target shrinks, the score explodes. An ancillary revenue line with a 100,000 TL target and a 300,000 TL actual produces 300% and distorts the whole scorecard on its own. Small-target indicators need a capping decision, which I come back to below.
Method 2 — Target / Actual
The same ratio inverted. Score = Target ÷ Actual.
At first glance this looks like the fix for lower-is-better indicators: the smaller the actual, the larger the score. But the formula loses control as the actual approaches zero.
Suppose a department has a monthly guest complaint target of 10. Here is how the same set of actuals scores under the two candidate formulas:
| Actual complaints | Method 2 (T/A) | Method 4 (2 − A/T) |
|---|---|---|
| 10 | 100.0% | 100.0% |
| 5 | 200.0% | 150.0% |
| 2 | 500.0% | 180.0% |
| 1 | 1,000.0% | 190.0% |
| 0 | Undefined | 200.0% |
Cutting complaints from 10 to 1 is a genuinely good month. Recording it as 1,000% makes every other indicator in that department irrelevant. And at zero the formula is undefined — the system either throws an error or quietly leaves the cell blank.
Reserve Method 2 for indicators where the actual cannot realistically reach zero and where a ratio reading is meaningful: cost per occupied room, labour productivity ratios. For every other lower-is-better case, Method 4 behaves better.
Method 3 — Higher than target is better
Method 1 is a ratio. Method 3 is a rule. The difference is whether the target is a threshold or a reference point.
Two variants:
- Threshold: Score = min(Actual ÷ Target, 1). Hit the target and you get full marks; exceeding it earns nothing extra.
- Proportional: Score = Actual ÷ Target, floored at 0. Exceeding the target is rewarded.
Suppose mandatory food-hygiene training has a completion target of 85%.
| Actual | Threshold | Proportional |
|---|---|---|
| 92% | 100.0% | 108.2% |
| 85% | 100.0% | 100.0% |
| 70% | 82.4% | 82.4% |
Which variant is correct depends on the economics of the indicator. Overshooting a mandatory training target has no marginal return — everyone who needed the certificate has it. Threshold is right. Overshooting an upsell revenue target lands directly on the P&L. Proportional is right.
Method 4 — Lower than target is better
The linear mirror. Score = (2 × Target − Actual) ÷ Target, which is the same as Score = 2 − (Actual ÷ Target), floored at 0.
Back to turnover. Target 28%, actual 35%:
Score = 2 − (35 ÷ 28) = 2 − 1.25 = 0.75 → 75%
Under Method 1 the same indicator scored 125%. That 50-point gap is not a calculation error. It is a direction error.
Now reverse it. With the same target, suppose the actual had come in at 22%:
Score = 2 − (22 ÷ 28) = 2 − 0.7857 = 1.2143 → 121.4%
Method 1 would have scored that genuine improvement at 78.6%. So the wrong direction fails in both directions at once: it rewards deterioration and it penalises improvement. On a line manager's scorecard, that means losing points in the month they finally managed to keep their team.
I have written before about why a rising turnover rate is a bad signal and why the same rate does not mean the same thing at every property.
Method 5 — Closeness to target
For some indicators the target is not a ceiling or a floor. It is the centre of a band, and variance costs money on both sides — though it lands on different lines of the P&L.
Score = 1 − ( |Actual − Target| ÷ Target ), floored at 0.
Stock days is the textbook case. In our hypothetical property, monthly F&B stock consumption is 1,200,000 TL, so 40,000 TL a day. That gives:
| Stock days | Inventory tied up | Variance from target | Method 5 score |
|---|---|---|---|
| 12 days (target) | 480,000 TL | 0 | 100.0% |
| 18 days | 720,000 TL | +6 days | 50.0% |
| 8 days | 320,000 TL | −4 days | 66.7% |
Running at 18 days locks an extra 240,000 TL of working capital into the storeroom. Apply a purely hypothetical 40% annual cost of funds and that is a 96,000 TL carrying charge over a year, landing in finance costs.
Running at 8 days releases the cash but takes the hit somewhere else: menu items running out, emergency purchasing premiums, guest complaints. That cost spreads across food cost and lost revenue, and it is considerably harder to measure.
Both variances are losses. They are not the same size and they do not appear on the same line. Method 5 at least recognises both of them as losses — which is precisely what no one-sided formula can do.
In practice a tolerance band makes this workable: full score if the variance sits within ±10% of target, linear decay outside it. On a 12-day target that means full marks anywhere between 10.8 and 13.2 days.
Which method for which indicator: a mapping table
Take the three original actuals and run them through all five methods. The inputs do not change. Only the formula does.
| Indicator (T → A) | Method 1 | Method 2 | Method 3 (threshold) | Method 4 | Method 5 | Correct method |
|---|---|---|---|---|---|---|
| Revenue (10,000,000 → 11,000,000) | 110.0% | 90.9% | 100.0% | 90.0% | 90.0% | Method 1 |
| Turnover rate (28 → 35) | 125.0% | 80.0% | 100.0% | 75.0% | 75.0% | Method 4 |
| Stock days (12 → 18) | 150.0% | 66.7% | 100.0% | 50.0% | 50.0% | Method 5 |
Methods 4 and 5 return identical numbers in this table. That is a coincidence of the data: all three actuals sit above target, so every variance is one-sided. The moment an actual falls below target the two methods separate, and the separation is the whole point.
| Indicator | Target | Actual | Method 4 | Method 5 | Which is right |
|---|---|---|---|---|---|
| Turnover rate (%) | 28 | 22 | 121.4% | 78.6% | Method 4 — a real improvement |
| Stock days | 12 | 8 | 133.3% | 66.7% | Method 5 — understock is also a loss |
Method 4 awards 133.3% to a property that ran its storeroom down to eight days. It scores an empty warehouse as outstanding performance. That row is the price of failing to notice that an indicator belongs to the closeness-to-target family.
When you make the mapping decision, factor in how much of the scorecard total the indicator carries; method and indicator weighting compound each other. The heavier a wrongly directed indicator is, the larger the error it produces.
Should scores above 100% be capped
The answer is not "always" — it is "per indicator". Capping is an accounting decision: do you let overperformance flow through to the scorecard total?
Take the same department in a good month. Weights: revenue 50, turnover 30, stock days 20.
| Indicator | Weight | Raw score | Uncapped contribution | Capped at 100% |
|---|---|---|---|---|
| Revenue | 50 | 110.0% | 55.0 | 50.0 |
| Turnover rate | 30 | 121.4% | 36.4 | 30.0 |
| Stock days | 20 | 100.0% | 20.0 | 20.0 |
| Department score | 100 | 111.4 | 100.0 |
Both numbers are defensible. Uncapped scoring makes exceptional performance visible; capped scoring stops one indicator from papering over weakness elsewhere. What is not defensible is leaving the decision unmade — because then the cap rule varies with whoever happened to configure the indicator.
A workable middle position: cap at 120%. Overperformance still counts, but no single indicator can carry the scorecard on its own.
Why an indicator with no target should not be scored
If the target field is empty, actual/target is undefined. Systems typically do one of three things here, and all three are wrong.
| What the system does | What the scorecard shows | Why it is wrong |
|---|---|---|
| Writes a score of 0 | Failure | There was no measurement, so there was no failure |
| Writes a score of 100 | Full achievement | An unearned score that inflates the total |
| Drops the indicator silently | Nothing at all | The weights no longer sum to 100 |
The correct behaviour is a fourth: flag the indicator as not scoreable, redistribute its weight for that period, and disclose that in a scorecard footnote.
This should not be confused with an indicator whose measurement period has not yet arrived. A not-yet-due indicator and an indicator that never had a target are different problems with different fixes. Both trace back to the same gap, though: the indicator's definition was written incompletely.
How the wrong method makes the scorecard lie
Turn the original three actuals into a department score. Weights again 50/30/20.
| Indicator | Weight | All Method 1 | Contribution | Correct method | Score | Contribution |
|---|---|---|---|---|---|---|
| Revenue | 50 | 110.0% | 55.0 | Method 1 | 110.0% | 55.0 |
| Turnover rate | 30 | 125.0% | 37.5 | Method 4 | 75.0% | 22.5 |
| Stock days | 20 | 150.0% | 30.0 | Method 5 | 50.0% | 10.0 |
| Department score | 100 | 122.5 | 87.5 |
Same month, same data, 35 points apart. The single-formula scorecard reports this department at 22.5% above target. In reality it closed the month with a turnover rate seven points above plan and 240,000 TL of surplus stock sitting in the storeroom.
This is the arithmetic version of a very old warning in the measurement literature. Ridgway set out the dysfunctional consequences of performance measurement in Administrative Science Quarterly in 1956, showing how one-dimensional measures bend behaviour toward the dimension being measured (Ridgway, 1956, Administrative Science Quarterly 1(2), pp. 240-247). Strathern's 1997 study of audit in British universities describes the same mechanism and is the source of the sentence now widely quoted as Goodhart's law (Strathern, 1997, European Review 5(3), pp. 305-321).
There is a difference worth drawing, though. Both of those works describe the indicator itself degrading under measurement pressure. A wrong target direction is an earlier failure than that: the indicator is sound, the measurement is accurate, and only the scoring arithmetic is broken. Which makes it harder to spot. Nobody is gaming anything. The formula produces the wrong answer entirely on its own.
Writing the method choice into the scorecard definition
The scoring method belongs to the indicator's definition, exactly as much as its name and its formula do. Every indicator in the definition record should have these fields filled in:
| Field | Example — Revenue | Example — Turnover rate | Example — Stock days |
|---|---|---|---|
| Unit | TL | % | days |
| Target | 10,000,000 | 28 | 12 |
| Target direction | Higher is better | Lower is better | Close to target |
| Scoring method | Method 1 | Method 4 | Method 5 |
| Tolerance band | — | — | ±10% |
| Cap | 120% | 120% | 100% |
| Owner | F&B Manager | HR Manager | Purchasing Manager |
| Source system | POS | Payroll | Inventory management |
Having this record moves the argument from month-end to the start of the period. "Why did turnover score 75?" is answered by the definition record, and nobody renegotiates a formula on the day the numbers close.
Running five different methods on one scorecard, indicator by indicator, is possible in a spreadsheet but hard to maintain — each new indicator means copying a formula by hand, and eventually somebody drags the wrong cell. Running all five methods on one scorecard is far safer in a structure where the definition is held separately from the data.
Conclusion
Target direction is the cheapest step in scorecard design and the one most often skipped. Assigning a direction to an indicator takes five seconds. The error a wrong direction produces accumulates quietly for an entire period.
What to keep:
- Direction belongs to the indicator, not to the period. Set it once, write it into the definition record.
- There are three directions: higher is better, lower is better, close to target. The third is the one most often missed and the most expensive to miss.
- Methods 4 and 5 agree while variance is one-sided and diverge when it flips. The point where they diverge is the only place the scorecard will tell you that you classified an indicator wrongly.
- An indicator with no target does not get scored. Not a zero, not a hundred.
- The method choice is part of the definition. Until it is written down, it gets relitigated every month.
Next time you open your scorecard, run one check: how many indicators do you not want to see above target, and how many of those are scored on actual/target? The gap between those two counts is a measure of how far your scorecard is off.
The pattern we run into most often when setting up measurement structures for hotel groups is not badly chosen indicators — it is well-chosen indicators scored in the wrong direction. If you would like to run that check on your own scorecard, or review your indicator set together, the HRMIM team is reachable here.


