这位客人到底是谁的？—— 为什么独立酒店一直在丢失直客关系

By Dr. Tong Yin

The first two articles in this series looked inward — at how a hotel sets its prices, and at how its many systems do, or do not, work together. This final article looks outward, at the channels through which guests actually arrive. For most independent and mid-market hotels, that means the online travel agencies. The relationship with the OTAs is the most consequential, and the most uncomfortable, in the business: they deliver enormous demand, and they take a large and growing share of the value of every guest they send. The question is not whether to use them — most hotels cannot do without them — but how to stop the dependence from deepening year after year, and how to do so without starting a price war the small operator can only lose.

1. The Dependence Is Deepening, Not Easing

For most of the last decade the industry has urged hotels to "book direct." Most of them have quietly gone the other way. According to the Cloudbeds 2026 State of Independent Hotels Report, drawn from roughly 90 million bookings across tens of thousands of properties in 180 countries, the OTA share of independent hotel bookings rose to 63.4% in 2025, up from 61% the year before, with some markets approaching 80%1. This happened in a year when global RevPAR for independent properties fell 5.4% and ADR dropped 5.8%2 — that is, hotels leaned harder on their most expensive channel precisely as their own revenue was shrinking. Demand is concentrating in a few platforms controlled by a small number of companies, and newer AI-driven discovery tools appear to be funneling even more traffic toward those platforms rather than toward hotel websites.

This is not an even burden. As I argued in an earlier piece, the structural divide in this industry runs along the line of who has scale3. Large branded portfolios negotiate lower rates, fund loyalty programs, and command direct demand that independents simply do not have. The independents — the backbone of the sector — are the ones most dependent on the OTAs and least able to push back. The dependence is structural, and it is getting deeper.

Figure 1. Independent hotels are leaning harder on the channel that costs them the most.

2. The True Cost Is More Than the Commission

The headline commission is only the visible part of the cost. Most major OTAs charge between 15% and 30% of the booking value4, and for an independent hotel the OTA commission line is often the single largest controllable cost in the P&L. But the headline rate understates the damage. In some markets, commission is applied to a tax-inclusive booking value, which can push the effective cost toward 36% on a standard booking5. On top of that sit promotional-participation fees, payment costs, and a cancellation rate that runs far higher on OTA bookings — 21.8% in 2025, against 10.6% for direct bookings from the same properties6. Each cancelled OTA reservation is a room that was held off the market, then released too late to resell.

Figure 2. A direct booking costs a fraction of an OTA booking — once the full cost of each is counted.

Once a hotel has a functioning website and booking engine, a direct booking costs roughly 4 to 5% of the booking value, against the 15 to 30% of an OTA reservation7. Direct bookings also tend to be more profitable per reservation — industry analyses put the gap at around 9 to 10% before ancillary spend8 — and the acquisition cost of a direct guest falls over time as the relationship matures, while the cost of an OTA guest stays the same on every repeat stay. A 100-room hotel at a $150 average rate with a 55% OTA share can pay on the order of $450,000 a year in commissions alone9. Even a modest shift of share from indirect to direct can move real money to the bottom line.

3. The Honest Caveat — and a Tactical Way Forward

Here I want to be careful, because the "book direct" message is often oversold. Not every direct booking is a saved commission. The OTAs perform real work — demand generation, global reach, a trusted booking experience — and shifting volume to direct channels carries its own costs: marketing, technology, staff time, and the risk of lower occupancy if the demand simply does not materialize. At least one careful study concluded that, once acquisition and service costs are modelled honestly, the net saving from moving bookings to direct can be marginal, or even negative, for some hotels10. The right answer is therefore not to wage war on the OTAs, nor to chase direct bookings at any cost. It is a deliberate channel mix — keeping the OTAs for their reach while steadily lowering the cost of the demand a hotel can capture on its own11.

A tactical layer, not another loyalty scheme

This is where the third model in our system, internally branded NOVA, fits. I describe it here as a worked example rather than a product pitch. The idea is narrow on purpose: a lightweight tactical layer that helps a small hotel reduce its OTA dependence by acting on the things it can actually control — the timing and targeting of direct offers, the recapture of guests an OTA delivered once, and the small adjustments that nudge the next booking toward the hotel’s own channel. It does not try to out-spend Booking.com on marketing, and it does not pretend an independent can build a loyalty program to rival a global chain. It works at the margin, where the margin is, and it stays anchored to the same pricing discipline described in the first article — the OTA price is a reference point, not a target to undercut into a race to the bottom.

Why this only works as part of one operating loop

A tactical layer like this is only as good as the data beneath it. Winning back the direct relationship requires knowing who the guest is, what they paid, and when they are likely to return — which is exactly the first-party guest data the OTAs withhold, and exactly what a fragmented technology stack fails to assemble. This is why the three articles in this series belong together: pricing, the integrated operating loop, and distribution are not three separate projects but three faces of the same problem. A hotel that cannot see its own guest cannot win that guest back, however clever the campaign.

What the testing showed

Like the pricing engine and the operating loop in the first two articles, this distribution layer was not taken on faith. It was put through the same validation program — the same 76-hotel test universe in Macau, the same three independently built systems, and the same deliberately adversarial scenarios — so that its behavior could be checked across architectures rather than self-reported. Of the three models in the system, this one came through the cleanest. On a dual scorecard spanning normal and extreme conditions it scored 94.8 out of 100, the highest of all nine models tested, and its recommendation anomaly rate sat at essentially zero — the most stable of the three engines. Like the others it ran with a 0% failure rate across all three systems, and its machine-learning layer fired on every record it was given, 23,408 of 23,408. The reason it is the steadiest is also the reason it is the most modest: it works narrowly, on the targeting and timing of direct offers and the recapture of guests an OTA delivered once, where the signal is clean and the action is well defined.

Figure 3. The distribution layer under cross-validation: the highest score of all nine models and near-zero anomalies (76 hotels, Macau).

A necessary caveat, the same one I have applied throughout: these are results from rigorous testing and simulation, not yet from a year of live deployment. They are strong enough to justify a careful, gradual rollout against real bookings at a small number of hotels — which is the only way to confirm in the field what the testing indicates — but I would not have a reader take them for more than they are.

4. Fair Terms, and the Limits of Technology

As with the other tools in this series, the terms matter as much as the technology. A genuinely fair approach for a struggling small hotel asks for almost nothing upfront: no binding contract, no bank or credit-card details, and no integration with the hotel’s internal systems at the outset. The operator can simply try it and, if unconvinced, stop at any time. If, after a trial, the hotel finds it useful, the pricing of the tool itself is aligned with the hotel’s success: a modest, affordable subscription, with the primary revenue coming from a share of the additional profit the system helps generate. Risk is shared, not transferred. If the product does not help the hotel earn more, it earns nothing. The vendor only wins when the hotel wins — a principle that matters most of all in distribution, where it is so easy to sell a hotel an expensive promise that never reaches its bottom line.

And the same caution I have raised twice already applies here too: no software, however well-targeted, can rescue a struggling hotel on its own12. Technology has wings, but it needs roots. A tactical distribution layer can shift a few points of channel mix and recover some margin; it cannot fix a weak product, a confused market position, a demoralized team, or a flawed strategy. Where a hotel needs more, comprehensive management diagnosis and consulting — grounded in research and matched to the specific business — has to work alongside the technology to address the problem at its root. The tools are one part of the answer, never the whole of it.

It is also why InsightBridge Global LLC, the consulting firm I founded, is deliberately not run to maximize profit. Its purpose is to use high-quality research — expressed through both consulting and these lightweight tools — to help the industry through its present difficulties and toward its future. Helping the industry develop, rather than extracting the maximum return, is the firm’s defining objective, and the most fundamental difference between it and a conventional commercial vendor.

That same commitment extends beyond individual hotels. InsightBridge Global LLC is prepared to offer pro bono strategic planning, research, and advisory services to major tourism-destination countries and cities — putting its promise not to pursue profit maximization into practice, and placing high-quality research at the service of the public interest where it can do the most good for the wider industry and the communities that depend on it.

That commitment now extends into longer-horizon research. We have established InsightBridge Global Lab, which will collaborate with Auburn University and other strong universities in the Atlanta region to study the artificial intelligence and AI-driven robotics that will shape the future of the hotel and tourism industry. As technology, shifting markets, and geopolitics reshape the landscape, the Lab’s aim is to provide the rigorous research and leading-edge technical support the industry will need to meet those challenges — and to navigate the transition steadily, and on its own terms.

This concludes the three-part series. Taken together, the articles argue one thing: that the hotels most in need of better pricing, a connected operation, and a fairer distribution mix are precisely the ones the current technology market has served least well — and that the way to serve them is lighter, cheaper, and more honest than the industry has so far offered.

Dr. Tong Yin holds a PhD in hospitality management and has 25 years of experience in the industry. He is the founder of InsightBridge Global LLC.

By Dr. Tong Yin

The first two articles in this series looked inward — at how a hotel sets its prices, and at how its many systems do, or do not, work together. This final article looks outward, at the channels through which guests actually arrive. For most independent and mid-market hotels, that means the online travel agencies. The relationship with the OTAs is the most consequential, and the most uncomfortable, in the business: they deliver enormous demand, and they take a large and growing share of the value of every guest they send. The question is not whether to use them — most hotels cannot do without them — but how to stop the dependence from deepening year after year, and how to do so without starting a price war the small operator can only lose.

1. The Dependence Is Deepening, Not Easing

For most of the last decade the industry has urged hotels to "book direct." Most of them have quietly gone the other way. According to the Cloudbeds 2026 State of Independent Hotels Report, drawn from roughly 90 million bookings across tens of thousands of properties in 180 countries, the OTA share of independent hotel bookings rose to 63.4% in 2025, up from 61% the year before, with some markets approaching 80%1. This happened in a year when global RevPAR for independent properties fell 5.4% and ADR dropped 5.8%2 — that is, hotels leaned harder on their most expensive channel precisely as their own revenue was shrinking. Demand is concentrating in a few platforms controlled by a small number of companies, and newer AI-driven discovery tools appear to be funneling even more traffic toward those platforms rather than toward hotel websites.

This is not an even burden. As I argued in an earlier piece, the structural divide in this industry runs along the line of who has scale3. Large branded portfolios negotiate lower rates, fund loyalty programs, and command direct demand that independents simply do not have. The independents — the backbone of the sector — are the ones most dependent on the OTAs and least able to push back. The dependence is structural, and it is getting deeper.

Figure 1. Independent hotels are leaning harder on the channel that costs them the most.

2. The True Cost Is More Than the Commission

The headline commission is only the visible part of the cost. Most major OTAs charge between 15% and 30% of the booking value4, and for an independent hotel the OTA commission line is often the single largest controllable cost in the P&L. But the headline rate understates the damage. In some markets, commission is applied to a tax-inclusive booking value, which can push the effective cost toward 36% on a standard booking5. On top of that sit promotional-participation fees, payment costs, and a cancellation rate that runs far higher on OTA bookings — 21.8% in 2025, against 10.6% for direct bookings from the same properties6. Each cancelled OTA reservation is a room that was held off the market, then released too late to resell.

Figure 2. A direct booking costs a fraction of an OTA booking — once the full cost of each is counted.

Once a hotel has a functioning website and booking engine, a direct booking costs roughly 4 to 5% of the booking value, against the 15 to 30% of an OTA reservation7. Direct bookings also tend to be more profitable per reservation — industry analyses put the gap at around 9 to 10% before ancillary spend8 — and the acquisition cost of a direct guest falls over time as the relationship matures, while the cost of an OTA guest stays the same on every repeat stay. A 100-room hotel at a $150 average rate with a 55% OTA share can pay on the order of $450,000 a year in commissions alone9. Even a modest shift of share from indirect to direct can move real money to the bottom line.

3. The Honest Caveat — and a Tactical Way Forward

Here I want to be careful, because the "book direct" message is often oversold. Not every direct booking is a saved commission. The OTAs perform real work — demand generation, global reach, a trusted booking experience — and shifting volume to direct channels carries its own costs: marketing, technology, staff time, and the risk of lower occupancy if the demand simply does not materialize. At least one careful study concluded that, once acquisition and service costs are modelled honestly, the net saving from moving bookings to direct can be marginal, or even negative, for some hotels10. The right answer is therefore not to wage war on the OTAs, nor to chase direct bookings at any cost. It is a deliberate channel mix — keeping the OTAs for their reach while steadily lowering the cost of the demand a hotel can capture on its own11.

A tactical layer, not another loyalty scheme

This is where the third model in our system, internally branded NOVA, fits. I describe it here as a worked example rather than a product pitch. The idea is narrow on purpose: a lightweight tactical layer that helps a small hotel reduce its OTA dependence by acting on the things it can actually control — the timing and targeting of direct offers, the recapture of guests an OTA delivered once, and the small adjustments that nudge the next booking toward the hotel’s own channel. It does not try to out-spend Booking.com on marketing, and it does not pretend an independent can build a loyalty program to rival a global chain. It works at the margin, where the margin is, and it stays anchored to the same pricing discipline described in the first article — the OTA price is a reference point, not a target to undercut into a race to the bottom.

Why this only works as part of one operating loop

A tactical layer like this is only as good as the data beneath it. Winning back the direct relationship requires knowing who the guest is, what they paid, and when they are likely to return — which is exactly the first-party guest data the OTAs withhold, and exactly what a fragmented technology stack fails to assemble. This is why the three articles in this series belong together: pricing, the integrated operating loop, and distribution are not three separate projects but three faces of the same problem. A hotel that cannot see its own guest cannot win that guest back, however clever the campaign.

What the testing showed

Like the pricing engine and the operating loop in the first two articles, this distribution layer was not taken on faith. It was put through the same validation program — the same 76-hotel test universe in Macau, the same three independently built systems, and the same deliberately adversarial scenarios — so that its behavior could be checked across architectures rather than self-reported. Of the three models in the system, this one came through the cleanest. On a dual scorecard spanning normal and extreme conditions it scored 94.8 out of 100, the highest of all nine models tested, and its recommendation anomaly rate sat at essentially zero — the most stable of the three engines. Like the others it ran with a 0% failure rate across all three systems, and its machine-learning layer fired on every record it was given, 23,408 of 23,408. The reason it is the steadiest is also the reason it is the most modest: it works narrowly, on the targeting and timing of direct offers and the recapture of guests an OTA delivered once, where the signal is clean and the action is well defined.

Figure 3. The distribution layer under cross-validation: the highest score of all nine models and near-zero anomalies (76 hotels, Macau).

A necessary caveat, the same one I have applied throughout: these are results from rigorous testing and simulation, not yet from a year of live deployment. They are strong enough to justify a careful, gradual rollout against real bookings at a small number of hotels — which is the only way to confirm in the field what the testing indicates — but I would not have a reader take them for more than they are.

4. Fair Terms, and the Limits of Technology

As with the other tools in this series, the terms matter as much as the technology. A genuinely fair approach for a struggling small hotel asks for almost nothing upfront: no binding contract, no bank or credit-card details, and no integration with the hotel’s internal systems at the outset. The operator can simply try it and, if unconvinced, stop at any time. If, after a trial, the hotel finds it useful, the pricing of the tool itself is aligned with the hotel’s success: a modest, affordable subscription, with the primary revenue coming from a share of the additional profit the system helps generate. Risk is shared, not transferred. If the product does not help the hotel earn more, it earns nothing. The vendor only wins when the hotel wins — a principle that matters most of all in distribution, where it is so easy to sell a hotel an expensive promise that never reaches its bottom line.

And the same caution I have raised twice already applies here too: no software, however well-targeted, can rescue a struggling hotel on its own12. Technology has wings, but it needs roots. A tactical distribution layer can shift a few points of channel mix and recover some margin; it cannot fix a weak product, a confused market position, a demoralized team, or a flawed strategy. Where a hotel needs more, comprehensive management diagnosis and consulting — grounded in research and matched to the specific business — has to work alongside the technology to address the problem at its root. The tools are one part of the answer, never the whole of it.

It is also why InsightBridge Global LLC, the consulting firm I founded, is deliberately not run to maximize profit. Its purpose is to use high-quality research — expressed through both consulting and these lightweight tools — to help the industry through its present difficulties and toward its future. Helping the industry develop, rather than extracting the maximum return, is the firm’s defining objective, and the most fundamental difference between it and a conventional commercial vendor.

That same commitment extends beyond individual hotels. InsightBridge Global LLC is prepared to offer pro bono strategic planning, research, and advisory services to major tourism-destination countries and cities — putting its promise not to pursue profit maximization into practice, and placing high-quality research at the service of the public interest where it can do the most good for the wider industry and the communities that depend on it.

That commitment now extends into longer-horizon research. We have established InsightBridge Global Lab, which will collaborate with Auburn University and other strong universities in the Atlanta region to study the artificial intelligence and AI-driven robotics that will shape the future of the hotel and tourism industry. As technology, shifting markets, and geopolitics reshape the landscape, the Lab’s aim is to provide the rigorous research and leading-edge technical support the industry will need to meet those challenges — and to navigate the transition steadily, and on its own terms.

This concludes the three-part series. Taken together, the articles argue one thing: that the hotels most in need of better pricing, a connected operation, and a fairer distribution mix are precisely the ones the current technology market has served least well — and that the way to serve them is lighter, cheaper, and more honest than the industry has so far offered.

Dr. Tong Yin holds a PhD in hospitality management and has 25 years of experience in the industry. He is the founder of InsightBridge Global LLC.