What 45 days of real booking data reveal about connectivity infrastructure
Less than 1% versus 12%. That’s the gap we see between the best and worst suppliers on our platform when it comes to one specific failure: a booking that gets rejected because the availability was no longer real. At Globick, connectivity is our whole business, so this is the kind of number we pay close attention to — and the kind worth explaining. Availability errors metric gets surprisingly little attention, despite shaping revenue, customer experience and operational load.
The industry loves to celebrate how many products are connected, how many suppliers are live, how many destinations can be sold. But what actually matters is simpler:
Can the booking be completed?
Over the last 45 days we analyzed booking performance across the Globick platform to understand one of the industry’s most common and costly problems: availability-related booking failures. The results are a useful benchmark, and they point at a topic that deserves far more attention: the quality of connectivity infrastructure.
As always, supplier names and commercial details are anonymized. The goal isn’t to criticize anyone. It’s to share real operational data that can help the whole industry improve.
The invisible problem behind failed bookings
Every connectivity professional knows this scenario. A customer searches for an activity. Availability appears to exist. They go to checkout. The booking is sent to the supplier. The supplier rejects it, because the inventory is already gone.
To the customer, that feels like a broken promise. To the distributor, it’s lost revenue. To the supplier, it’s an avoidable support ticket and a dent in trust.
This is what we call an availability error. Unlike a technical outage or a configuration bug, it doesn’t mean the booking system is broken. It means the availability shown to the customer no longer matched reality by the time they tried to book.
The system worked. The information was stale.
The benchmark: how often do availability errors happen?
Across the period we analyzed, availability errors hit roughly 1 in 22 booking attempts — a platform-wide rate of 4.63%.
That number alone is worth pausing on. One in twenty-two is not a rounding error. At scale, it’s thousands of travelers who thought they had booked something that wasn’t really there.
But the platform average hides the real story. The moment we split the data by how the inventory reaches us, the picture changes completely.
Not all connectivity models behave the same
At a high level, the industry runs on two models.
Direct contracting. The product is contracted directly with the operator who runs the activity. Whoever sells it is sourcing availability from the party that actually owns and controls the inventory, with no commercial layer in between.
Intermediated contracting. The product is sourced through one or more commercial intermediaries — aggregators, resellers, bedbanks — who sit between the operator and the seller. These intermediaries do a lot of valuable work: product aggregation, commercial contracting, supplier onboarding, customer service, content curation, distribution reach.
These services create enormous value. Honestly, much of the tours and activities ecosystem simply could not function without them — they are what let a small distributor sell a global catalogue, and what let a local operator reach markets it would never touch alone. The industry needs them.
But every extra layer is one more place where inventory information has to stay in sync. And that’s where it gets interesting.
The difference is significant
Here is the availability error rate for each anonymized supplier, colored by model.
The pattern is not subtle. Intermediated suppliers run availability error rates from about 1% up to roughly 12%. Direct connections mostly sit under 2%.
Rolled up, the gap is stark.
Table 1 — Availability error rate by connectivity model
Availability failures were roughly 5.6 times higher when an intermediary layer sat between the distributor and the original inventory source.
Before anyone jumps to conclusions: this is not an argument against intermediaries. They solve real business problems. They accelerate distribution, they simplify contracting, they help suppliers reach markets they’d never access alone. The data isn’t saying intermediaries are worse. It’s saying something more precise: every additional commercial step the inventory passes through is another point where its availability can drift out of sync, and that demands a great connectivity infrastructure..
First, a quick word on caching
To understand why extra layers cause trouble, you need one simple concept: caching.
A cache is just a saved copy of information so you don’t have to ask for it every single time. When a customer searches for an activity, the platform doesn’t call the supplier live for every result — that would be slow and would bury the supplier under millions of requests. Instead it keeps a recent copy of «what’s available» and shows that.
The catch is obvious once you say it out loud: a saved copy is only as good as the last time you updated it. The longer ago you refreshed, the more likely reality has moved on — a slot sold out, a price changed, a session got paused. That gap between the saved copy and reality is exactly where availability errors come from.
So every layer between the supplier and the customer typically keeps its own cache, refreshed on its own schedule. More layers means more copies, each potentially a little out of date. The problem usually isn’t that anyone is doing a bad job — it’s that keeping many copies in sync is genuinely hard.
Which brings us to the trade-off at the heart of every connectivity platform.
The cache paradox
Caching isn’t optional — it’s the only way to keep search fast and keep suppliers from being buried under requests. But it forces a three-way trade-off that no platform fully escapes.
Table 2 — The cache triangle
The catch is that pushing on one corner usually pulls on the others. More caching means faster searches and lower cost — but a higher risk of stale inventory. Less caching means better accuracy — but more API calls and higher operating cost.
Connectivity architecture is, at heart, the art of balancing these three forces. Nobody gets all three for free.
Four ways to cache availability
Most availability caches can be described along two dimensions: what you cache, and how often you refresh it. That gives four common approaches.
Dates only + fixed refresh is the simplest. You store available dates and refresh on a fixed clock — every hour, every six hours, every day. Easy to build, cheap to run, but a thinner customer experience and more live supplier requests.
Dates only + progressive refresh keeps the cache lightweight but makes the refresh dynamic: a slot tomorrow gets checked every few minutes, the next week every hour, anything beyond a month once a day. It puts resources where volatility actually is, and it’s far more efficient than a flat clock.
Dates + sessions + prices + fixed refresh makes the choice of refresh frequency unforgiving. Refresh too rarely and all that detailed information goes stale fast, since sessions and prices change more often than dates do. Refresh too often and you flood the supplier with requests, losing the very efficiency the cache was meant to provide — at the extreme, a high-frequency refresh costs almost as much as having no cache at all. With a fixed clock there’s no good answer: the same frequency is always too slow for near dates and wasteful for far ones.
Dates + sessions + prices + progressive refresh is the most sophisticated. You store a lot of commercial information and refresh it intelligently based on how volatile each date is. More engineering, more monitoring — but in our experience the best balance of performance and accuracy by a wide margin.
The last line of defense
No matter how advanced your cache gets, one principle should stay non-negotiable: validate availability in real time, immediately before creating the booking.
A cache is a prediction. A booking is a commitment. Those are not the same thing.
Even the best cache can’t perfectly predict the moment inventory changes. Bookings land. Capacity shifts. Operators pause products or adjust manually. Real-time validation right before the booking is the safeguard that protects customers, suppliers, distributors and support teams from a failure that was entirely preventable. Skip it, and you’re betting customer experience against cache accuracy — rarely a good trade. It’s exactly how a supplier that could sit at 1% ends up closer to 12%.
What the industry should measure
We’ve gotten very good at measuring growth: connected suppliers, connected products, API calls, revenue, transactions. We pay far less attention to the metrics that decide whether a booking actually works:
- Booking success rate
- Availability error rate
- Inventory synchronization latency
- Availability freshness
These don’t look as impressive on a slide as product counts. But a platform with fewer suppliers and higher booking reliability can deliver more real value than a giant catalogue that can’t be trusted at checkout.
What we learned
The takeaway isn’t that direct connectivity is superior, or that intermediaries should be avoided. Both are essential parts of a healthy distribution ecosystem, and intermediaries in particular create value that the industry genuinely could not replicate without them.
The real lesson is simpler: what separates a 1% supplier from a 12% one is investment in connectivity infrastructure. Cache architecture matters. Refresh strategy matters. Real-time validation matters most of all. None of these come for free — they take engineering time, monitoring, and a deliberate commitment to treat reliability as a feature, not an afterthought.
That investment is rarely visible from the outside, and it’s easy to underestimate. We dug into exactly how much it costs to build and maintain connectivity properly in a rare, data-driven look at the true cost of API integrations — worth a read if you want to understand what’s really behind a reliable booking flow.
The industry’s next competitive advantage probably won’t come from connecting more products. It’ll come from making sure the products already connected can actually be booked.
Because connectivity isn’t measured by how many activities appear on a screen. It’s measured by how many bookings reach the supplier and stick. And increasingly, reliability is the most important feature of all.
This benchmark reflects aggregated, anonymized data observed across the Globick platform over a 45-day period. We share it because the tours and activities industry has too little public, real-world data on connectivity performance — and you can’t improve what you can’t see.