Hotel KPIs Can Drive the Wrong Behaviour

During a results meeting, I once saw a dashboard that appeared to describe an exceptionally healthy hotel. Housekeeping had exceeded its productivity target, Front Desk had reduced average check-in time, Food and Beverage was keeping food cost within budget, and Maintenance was closing more incidents than the previous month. Almost every box was green. Yet rooms were accumulating small defects, some guests had to repeat their requests, the team was arguing over priorities, and several operational decisions were beginning to be made with the report in mind rather than the guest stay. The indicators were improving while the hotel was quietly getting worse.

That contrast helped me understand that a hotel KPI does not merely measure behaviour. As soon as we link it to recognition, pressure, internal reputation, promotion or pay, it also begins to shape that behaviour. People quickly learn which figure is being watched, which deviation requires an explanation, and which result allows them to emerge well from a meeting. They then adapt their decisions to that system. No one needs to instruct them to manipulate anything; it is enough for them to clearly perceive which number protects their position.

The problem emerges when we turn a partial representation of reality into an absolute goal. Goodhart’s well-known law can be summarised in a highly useful way for Hospitality: when a measure becomes a target, it risks ceasing to be a good measure. If we reward only the number of rooms cleaned, we will end up producing speed. If we reward only cost reduction, we will produce savings. If we demand minimum times for every interaction, we will produce brevity. The uncomfortable question is what quality, information, collaboration or trust we may be sacrificing in order to achieve it.

A hotel intensifies this tension because it operates as a system of dependencies. A decision that benefits one department can transfer work, cost or frustration to another. A late check-out sold by Front Desk can compromise the Housekeeping sequence; a staffing reduction can temporarily improve labour cost while weakening service recovery; a hotel revenue management strategy can increase ADR and attract demand whose operational complexity has not been assessed. Departmental results and hotel results are not always the same thing, even if they share a building and appear together in the budget.

I have come to call the deterioration caused by an indicator KPI-Induced Harm when the behaviour required to protect it damages another relevant dimension of the business. That harm may affect the guest experience, quality, interdepartmental collaboration, safety, learning, workplace climate or hotel profitability. My intention in this article is not to discredit hotel metrics. Managing a hotel without indicators would be absurd. What I propose is more demanding: learning to measure not only the result we want to achieve, but also the behaviour we are creating in order to achieve it.

The hotel that emerges when the number starts to take command

KPIs are necessary because they reduce complexity. Every day, a hotel produces thousands of interactions, decisions and movements that no manager can observe directly. We need signals that allow us to identify trends, allocate resources, compare periods and detect deviations. The mistake is forgetting that every indicator is a selection. It shows one part of reality and leaves another outside. The simpler the figure, the greater the amount of context that has usually disappeared during its construction.

An average check-in time of five minutes, for example, does not explain whether the guest received the right information, understood the services they had booked, had to return to Front Desk, or whether the colleague assisting them identified an important need. Nor does it distinguish between an individual arrival without luggage, a family with children, a booking with issues, or a group arriving before the scheduled time. The figure appears objective, but it may be comparing entirely different operational situations.

360° Hotel Intelligence

Your hotel already generates the data. HotelGEX turns it into decisions.

A platform built from real hotel operations to connect guests, Customer Journey, operations, Revenue, F&B, Groups & Events and Management within one intelligent context.

Discover the platform

Discover HotelGEX → 30-day free trial

Explore HotelGEX

Customer 360 Revenue Intelligence Operational Intelligence F&B Groups & Events

Housekeeping productivity presents a similar challenge. Measuring rooms per person or minutes per room can be useful for organising resources, but it loses value when it ignores room type, previous occupancy, departure condition, extra beds, technical issues or the distance between areas. If we also link the indicator to daily pressure, rankings or hotel incentives, the team may learn that reporting a particularly deteriorated room amounts to admitting supposed inefficiency. At that point, the KPI stops revealing the problem and starts concealing it.

This response is too often attributed to a lack of commitment. My experience has taught me to be more cautious. People respond to the environment we have designed. If we ask for speed, recognise speed, publish speed and ask only about speed, we should not be surprised to receive speed, even when the guest needed something else. Before questioning a team’s attitude, it is worth reviewing the architecture of goals we have given it.

Incentives are not only financial either. A figure can shape behaviour because it determines who receives congratulations, who must justify their results, who is allocated better shifts, or which department is seen as responsible for a deviation. Even an informal ranking discussed every morning can exert more pressure than an annual bonus. The team quickly identifies which indicator brings prestige and which creates discomfort. In Hospitality, where coordination depends so heavily on everyday relationships, that perception has enormous power to change priorities.

I have seen this mechanism take many different forms in hotel management:

• Housekeeping productivity. When the dominant objective is to increase the number of completed rooms, inspections may be reduced, maintenance communications postponed, or tasks passed to the next shift. The indicator improves because it records the room as completed, even though another person must return later to correct it. Apparent productivity rises while rework increases, yet no one attributes that rework to the original KPI.

• Check-in time. Reducing waits is valuable, but excessive pressure can turn arrival into a rushed formality. The team avoids conversations that could anticipate needs, explain services or prevent complaints. It may also pass complex cases to colleagues in order not to affect its average. We achieve a shorter arrival process and, paradoxically, a stay with more subsequent contacts.

• Food and beverage cost. Keeping it under control is part of responsible operations. However, if it becomes the only relevant measure, it may encourage lower-quality purchasing, poorly judged portion reductions, a repetitive offering, or decisions that undermine perceived value. The cost percentage may fall while consumption, satisfaction and intent to return also decline.

• Upselling per employee. An upselling target can drive revenue when the recommendation offers genuine value. Poorly designed, it encourages persistent, irrelevant suggestions or recommendations made at inappropriate moments. The team learns to interpret every conversation as a sales opportunity and stops distinguishing between helping the guest and pressuring them. Immediate revenue appears in the report; the loss of trust rarely appears beside it.

• ADR and occupancy. Both indicators are fundamental in hotel revenue management, but neither alone represents the profitability of demand. A high rate may come with acquisition costs, inclusions, compensation or operational requirements that reduce contribution. High occupancy may fill the hotel with segments that displace more profitable business or overload services that lack sufficient capacity.

• Direct sales percentage. Improving the direct channel can be a sensible strategy, but turning its share into an isolated goal can lead to discounts, commercial investment or benefits whose cost exceeds the commission avoided. Distribution should be assessed by net contribution, not by ideological rivalry between channels. The complexity of this balance also arises when we explain why the same room can be sold at different prices without that difference necessarily being an inconsistency.

• Incident closure time. Maintenance can reduce its list by closing work orders after a temporary repair, grouping problems under the same cause, or classifying certain requests as non-priority. The data shows speed, but not recurrence, total downtime or guest impact. Closing an incident and resolving it are related actions, but they are not equivalent.

• Satisfaction score. When a survey becomes a departmental examination, understandable temptations arise: requesting reviews only from satisfied guests, insisting on the highest rating, or challenging the legitimacy of a comment. Instead of using feedback to learn, we begin using the guest to protect the indicator. In addition, knowing what the customer wants is insufficient if we ignore when they expect us to act.

• Labour cost. A reduction may look like a financial victory until it leads to overtime, turnover, absenteeism, errors, lost sales or compensation. The cost is contained on one line and reappears fragmented across many others. Because each consequence belongs to a different report, the initial decision may retain the appearance of success for too long.

The common pattern is local optimisation. Each department acts rationally within its own perimeter, but the whole achieves an inferior result. Housekeeping protects productivity, Front Desk protects times, Food and Beverage protects cost, and Sales protects conversion. No one wants to harm the hotel. Even so, each area may end up transferring work and risk to the others because we have defined success in a fragmented way.

This fragmentation creates a particularly dangerous contradiction: the better a team becomes at meeting its indicator, the greater its ability may be to damage what the indicator was meant to improve. A department with little discipline may not even reach the target. A highly organised one can discover the exceptions, shortcuts and decisions that protect it. Operational sophistication does not always eliminate metric gaming; sometimes it perfects it.

It is worth distinguishing intentional manipulation from defensive adaptation. Manipulation means consciously altering data or its classification. Adaptation means changing the work in response to what the organisation considers important. The latter is far more common and more difficult to detect. If Front Desk stops initiating useful conversations because it knows they will be interpreted as slowness, it is probably not falsifying anything. It is accurately obeying the message it receives.

KPI-Induced Harm usually takes six forms worth observing:

• Effort displacement. The team concentrates time on what is measured and reduces its attention to less visible tasks. Documentation becomes perfect while communication between shifts loses quality, or extras are sold while availability to resolve problems declines.

• Artificial denominator reduction. Some cases are excluded, reclassified or referred elsewhere to improve the average. A complex query no longer counts as first-contact resolution; a problematic room is assigned to another category; a complaint is logged as an observation. The report improves because we have changed the boundary of what counts.

• Temporal transfer. Costs, incidents or tasks are postponed to protect the weekly, monthly or annual close. The KPI reaches its target at the expense of the following period. Operations then enter a dynamic of accumulated debt that eventually appears when it is already difficult to link it to the original decision.

• Interdepartmental transfer. One area improves its indicator by sending work, complexity or conflict to another. This practice damages collaboration because each team feels others achieve good results at its expense. The dashboard begins to operate as a system of territorial defence.

• Selection of the convenient customer. The team pays more attention to guests who help achieve the goal and avoids situations capable of damaging it. It may favour easy sales, predictably positive surveys or straightforward complaints, leaving unattended those interactions that require greater judgement.

• Purpose substitution. The figure stops being a signal and becomes the purpose of the work. We no longer clean in order to deliver an impeccable room, but to meet the ratio; we no longer resolve issues to restore trust, but to close the ticket; we no longer recommend in order to help, but to increase average revenue.

This final form is the deepest because it changes the meaning of the work. A good indicator should help the team understand its contribution. A poorly designed one reduces that contribution to a figure that does not capture its full value. When this happens, people may feel obliged to choose between meeting the target and doing what they consider professionally right. Maintaining that contradiction for too long erodes judgement, pride and trust.

A dangerous obsession with uniform outcomes can also emerge. Every arrival, room, event and guest has different complexity, yet the KPI seeks comparability. To achieve it, we remove nuance until we construct a standard that appears precise. We then penalise exceptions for failing to behave like the average. This tension is closely related to the point at which demanding too much begins to make the hotel worse.

The final consequence can be disconcerting: a property meets its departmental objectives and yet loses coherence. Every area presents a defensible report, but the guest experiences the seams between departments. That is why I consider it so important to remember that the coherent hotel is worth more than the apparently perfect hotel. The customer does not assess our KPIs; they experience the combined result of the decisions those KPIs create.

Designing indicators that protect the complete outcome

The solution is not to multiply metrics until we create a dashboard that needs its own interpretation department. Adding indicators without hierarchy can increase noise and give everyone the figure that best supports their narrative. Nor is it advisable to remove goals and rely only on intuition. The key is to design a system in which every KPI has a clear purpose, known limits and counterbalances capable of revealing its side effects.

Before approving an indicator, I use a simple question: what would a reasonable person have to do to improve this figure, even if they did not share our intention? The wording forces us to abandon the ideal behaviour we imagine from a meeting room. It puts us in the position of someone working under pressure, with limited time, conflicting priorities and a guest in front of them. If the easiest route to reaching the target harms the hotel, the problem is not the person who discovers it, but the design.

Starting from that question, I suggest reviewing every metric using a KPI-Induced Harm Matrix. It is not intended to assign mathematical truth to human behaviour. Its value lies in forcing us to discuss dimensions that often disappear when we select indicators:

• Outcome we want to protect. The purpose must be described in operational, not merely numerical, language. If we measure check-in time, the outcome we seek may be a smooth, informative arrival without unnecessary effort. This definition prevents us from confusing fewer minutes with the real goal.

• Desired behaviour. We must specify which decisions we expect the indicator to encourage. This may include preparing arrivals in advance, simplifying administrative tasks or allocating resources more effectively. A KPI without explicitly desired behaviour leaves too much room for defensive interpretations.

• Plausible wrong behaviour. It is worth imagining how the figure could be improved by sacrificing quality, concealing exceptions or transferring work. We are not seeking to accuse anyone, but to anticipate rational responses. The easier it is to describe that shortcut, the more attention the design will require.

• Scope for manipulation. We review who enters the data, who classifies exceptions and who validates closure. An indicator may be technically correct yet highly sensitive to changes in criteria, cut-off times, samples or denominators.

• Cross-impact. We identify which departments, guests, suppliers or periods bear the cost of improving the figure. This dimension is essential because many hotel KPIs appear positive only while we observe the area receiving the benefit.

• Time until harm emerges. Some consequences are immediate while others take months. Reducing staffing may improve current results, while turnover, burnout and quality loss emerge later. The longer the delay, the easier it is to attribute the harm to another cause.

• Necessary counterbalances. Every relevant KPI should coexist with a measure that reveals its potential cost. Speed needs quality; sales need relevance and margin; savings need service impact; productivity needs rework, safety and sustainable workload.

• Right to explain the exception. No operational metric should prevent professional judgement from justifying a deviation. Exceptions should not become a permanent excuse, but neither should they become an automatic infringement. A mature hotel learns from them before penalising them.

To support diagnosis, each indicator can be rated from one to five across four dimensions: pressure intensity, ease of manipulation, cross-impact and delay of harm. The strength of its counterbalances is then assessed. I do not recommend turning the total into another KPI; that would be an irony all too fitting for Hospitality. The score is useful for prioritising conversations and identifying which metrics need urgent review.

An indicator subject to high pressure, easy to manipulate, capable of harming several departments and whose consequences emerge late represents a considerable risk. Monthly labour cost, room productivity, sales conversion or food cost may fall into this category if used in isolation. The greater the associated incentive, the more robust the limits must be.

A particularly useful practice is to create tension pairs. Every efficiency KPI is accompanied by another that protects quality or the final outcome. In this way, the organisation explicitly recognises that improving one dimension can harm another:

• Completed rooms and rework. Productivity is analysed alongside failed inspections, subsequent incidents, reopened rooms and differences in complexity. The goal stops being to clean more and becomes delivering correctly with reasonable use of resources.

• Check-in time and first-time resolution. A brief arrival is positive only if it does not force the guest to return to clarify information, complete formalities or request what could have been anticipated. Speed is balanced against the customer’s total effort.

• Food cost and contribution. The cost percentage must coexist with margin per dish, waste, volume, satisfaction and alignment with the proposition. A product with a higher cost percentage may generate greater contribution and strengthen positioning.

• Upselling and acceptance quality. In addition to revenue, it is advisable to monitor cancellations, refunds, complaints, margin and fit with the segment. Not every accepted sale represents a sound recommendation.

• ADR and net contribution. The rate must be read alongside acquisition costs, inclusions, cancellations, operational consumption and demand displacement. The visible price alone does not describe how much economic value a stay contributes.

• Closed incidents and recurrence. Maintenance needs to know how many problems reappear, how long the asset remains affected and what impact the fault has on rooms or services. Closing quickly loses merit if the repair fails again.

• Labour cost and service capacity. Savings must be analysed alongside overtime, absenteeism, turnover, vacancies, complaints, lost sales and workload per person. Sustainable efficiency requires the capacity to absorb reasonable operational variation.

Tension pairs are not intended for one figure to cancel the other out. Their purpose is to prevent us from interpreting an improvement without asking what has happened around it. In many hotel strategies, the genuinely useful debate emerges when two indicators move in different directions. If productivity rises and rework rises too, we can no longer celebrate the first figure without studying the second.

In addition to paired KPIs, I recommend including protective limits. These are conditions that invalidate an apparently good result when a critical dimension has been breached. A productivity bonus might not be activated if the rework rate exceeds a certain level. A sales reward might be reduced if cancellations or complaints linked to the sale increase. The message is clear: no goal is worth achieving by harming what the organisation considers non-negotiable.

These limits should be few and understandable. If we add twenty conditions, no one will know what to prioritise and the system will end up looking like a cancellation clause drafted on a Friday afternoon. Hotel strategic planning needs hierarchy. Safety, compliance, data integrity, guest treatment and respect between teams are often good candidates to serve as non-negotiable boundaries.

Another principle is to reduce the weight of individual comparison when the outcome depends on a chain of work. The hotel guest experience rarely belongs to one person alone. A request may start at Front Desk, pass through Housekeeping and end in Maintenance. If we reward only the department that closes the interaction, we risk making invisible the person who provided information, prepared the resource or prevented the problem.

Shared KPIs can improve this situation, although they must also be designed carefully. An overly broad collective goal allows no one to feel direct responsibility. This is why I usually combine three levels: a measure of the hotel’s overall outcome, a measure of the interdepartmental process, and a specific contribution over which the team has genuine decision-making capacity. Responsibility is distributed without being diluted.

The ability to influence is a fundamental criterion. We should not evaluate a person on a figure that depends mainly on factors beyond their control. If Front Desk does not decide staffing, deposit policy, booking configuration or group arrival times, holding it responsible for all waiting time is unfair and unhelpful. The indicator can still exist, but its interpretation must distinguish between responsibility, influence and context.

It is also worth abandoning the obsession with the single data point. A monthly figure may improve because of chance, demand mix or volume changes. Before rewarding or correcting, I prefer to observe trends, dispersion and exceptions. Two hotels can have the same average waiting time, even though one maintains consistent times and the other combines excellent arrivals with forty-minute episodes. The average is identical; the experience is radically different.

Dispersion offers particularly valuable information because it reveals process stability. In Housekeeping, it is not enough to know average time per room; it is important to understand why certain areas or room types systematically deviate. In Food and Beverage, average cost may conceal profitable dishes and others that destroy margin. In hotel marketing, overall conversion may hide profound differences between segments, campaigns and devices.

Financial incentives require even greater caution. The greater the reward tied to a figure, the greater the energy devoted to influencing it. That energy can produce innovation and discipline, but also target negotiation, selection of easy cases, cost deferral and disputes over ownership of the result. Before linking pay to a KPI, I would propose at least these conditions:

• Do not pay for a single figure. The incentive must include results, quality and collaboration. A dominant metric simplifies administration, but amplifies the risk of creating narrow behaviour.

• Avoid artificial cliffs. Thresholds that separate receiving the entire bonus from receiving nothing encourage extreme decisions near closing. A reasonable progression reduces the temptation to bring forward revenue, postpone costs or reclassify cases.

• Reserve a portion for professional judgement. Qualitative assessment makes it possible to recognise contributions that a figure does not capture, such as helping another department, training someone or protecting the relationship with a guest. This discretion requires transparency so that it does not become favouritism.

• Include deferred outcomes. Some decisions appear good today and reveal their cost later. Retaining part of the incentive until quality, recurrence or sustainability can be checked reduces the preference for immediate gains.

• Invalidate achievements obtained against the limits. A sales improvement achieved through misleading promises, or productivity achieved by failing to meet standards, should not generate a reward. The result and the method are part of the same assessment.

• Review the system with those who live it. Frontline colleagues often detect possible shortcuts, unfair comparisons and invisible work before anyone else. Listening to them before launch is operational design, not a concession.

The monthly conversation also needs to change. Many meetings are limited to asking who met the target, who deviated and what action will restore it. That structure encourages people to justify the past and promise corrections. A more mature meeting asks what behaviour produced the figure, what cost does not appear in the report, and which department is absorbing the effort.

There are questions I use to shift the debate from defending the number to understanding the system: what had to get worse for this KPI to improve, which cases were left out, what work was transferred, which decision would we repeat if the indicator did not exist, what would someone interested only in protecting the figure do, and what is the guest seeing that our dashboard does not show? None requires a sophisticated tool. They require curiosity and a willingness to hear uncomfortable answers.

A good dashboard should include at least four families of measures. First, business outcomes, such as contribution, revenue and profitability. Second, guest outcomes, including effort, recovery and consistency. Third, operational and human health, covering capacity, quality, safety, turnover and rework. Fourth, learning, observing whether incidents recur and whether improvements are sustained. No family alone can represent the health of the hotel.

This requires accepting a tension that can sometimes make owners and teams uncomfortable: the best indicator system does not always provide a quick answer. At times, a decision increases costs in one area in order to reduce them in another, or worsens a KPI for several weeks before improving the complete process. Leadership in the hotel sector also means protecting those decisions from premature interpretation.

To begin the review without paralysing operations, I propose a thirty-day cycle:

1. Take inventory of existing KPIs. During the first week, gather indicators, targets, owners, frequency, data sources and associated incentives. It is often revealing to discover how many metrics exist and how many people do not know which specific decision they are used for.

2. Listen to real behaviours. During the second week, hold brief conversations with those who work under each indicator. I would not ask whether they like the KPI, but what they do when they are close to missing it, which decisions they avoid and what work remains outside the record.

3. Apply the harm matrix. During the third week, identify defensive behaviours, possible manipulation, cross-impacts and delayed consequences. Then select the three indicators with the greatest risk, without attempting to redesign the entire system at once.

4. Create counterbalances and limits. Each priority KPI receives a quality measure, a protective limit and a shared definition of exceptions. Also review who should share responsibility and which part of the outcome the team does not directly control.

5. Test before rewarding. During the fourth week, use the new design in learning mode, without immediate pay consequences. The objective is to discover new distortions before turning them into established behaviour.

I recommend documenting each indicator on a short factsheet. It should include purpose, formula, frequency, owner, decisions it enables, desired behaviour, possible wrong behaviour, counterbalances, exclusions and review date. If we cannot explain a KPI on one page, we probably cannot expect the team to interpret it consistently either.

The review date matters because a useful metric can cease to be useful. Changes in segmentation, processes, value proposition or operating structure alter the meaning of the data. An indicator created to solve a particular problem may survive for years after that problem has disappeared. It then consumes attention and continues shaping behaviour without supporting a relevant decision.

Removing KPIs is also part of good hotel management. Every indicator carries a cost of capture, interpretation and conversation. More importantly, it occupies mental space. When everything is a priority, the team ends up attending to what creates the most pressure, not necessarily what creates the most value. An excellent dashboard is not the one containing the most information, but the one that protects the decisions that truly matter.

The purpose of measurement should be to learn and make better decisions, not to prove that everything was under control. If the team feels that every deviation will be used to find someone to blame, it will learn to present safe figures. If it perceives that information serves to understand causes and improve processes, the quality of what it communicates will increase. Data integrity depends as much on culture as on formula.

My practical advice is that next Monday you choose just one relevant KPI and ask three frontline colleagues what they do differently because they know that figure is being watched. Do not ask what they should do according to the procedure. Ask what really happens on a high-occupancy day, with absences, early arrivals and simultaneous incidents. That conversation will probably reveal more about the indicator than several pages of reporting.

Then review which dimension could deteriorate while that KPI improves and add a simple counterbalance. If you measure speed, monitor repetition or rework. If you measure savings, monitor capacity and quality. If you measure sales, monitor margin and trust. If you measure satisfaction, also consider which comments, guests or moments are being left out. You do not need to rebuild the entire dashboard to begin reducing KPI-Induced Harm.

Numbers are indispensable, but they are not innocent. They express priorities, distribute attention and teach the team what the hotel understands by success. That is why they deserve the same care we devote to service, positioning or the budget. In the end, we will get more of what we reward, even if it is not exactly what we intended. Our responsibility is to ensure that improving a figure also helps improve the hotel.