You asked whether I can cull the data and get it working. Rather than answer that, here is your file with the criteria applied to all 544 properties. The model is sound. The data underneath it is not ready, and the gap has a specific shape that will show up in revenue if it ships as it stands.
The earlier analysis was built on the assumption that ID commits to inventory and carries the risk of unsold rooms. That was wrong, and it changes the shape of the problem in your favour.
You hold direct relationships with roughly 350 hotels and airlines. Real time availability and dynamic pricing pass into your system at agreed net rates. You are not exposed to unsold inventory.
But you are not indifferent to where a booking lands. Discounts vary by property, overrides pay against production goals, and some partners carry non-financial value. That makes this a steering problem, not a distribution problem.
Steering is the better problem. It is a ranking question, and ranking is the part of this that software genuinely does better than people. The rest of this page is about whether the data can currently support that ranking.
Four of the eight criteria are populated on nearly every property. Three are not. The three that are not include the one the entire programme exists to capture.
This is the structural problem, and it follows directly from the first finding.
No property can reach Glo on the well-populated data alone. Every Glo designation currently depends on the three columns that are mostly blank.
A hotel with a genuinely strong override and a blank cell scores identically to a hotel with no override at all. The score cannot tell the difference between a bad partner and an unrecorded one.
If the ceiling argument holds, we should see it in the outcome. We do.
| Group | Properties | Avg score | Reaches Glo |
|---|---|---|---|
| Override field filled in | 37 | 10.1 | 51.4% |
| Override field blank | 507 | 6.2 | 10.1% |
Having the override cell completed makes a property 5.1 times more likely to qualify as Glo, and lifts its average score by 3.9 points out of 18.
The scorecard is not currently ranking partner quality. It is substantially ranking which rows someone finished filling in. That is a data-collection artifact being read as a commercial signal, and once it drives incentives it becomes self-reinforcing.
The artifact is not evenly distributed. It has a geography, and this is the finding with the clearest commercial consequence. The USA is set aside throughout: with an outbound client base, thin domestic coverage is expected and is not the issue.
| Region | Properties | Glo (12+) | Share of region |
|---|---|---|---|
| Mexico | 130 | 35 | 26.9% |
| Caribbean | 172 | 32 | 18.6% |
| Central America | 12 | 1 | 8.3% |
| South East Asia | 46 | 1 | 2.2% |
| Indian Ocean | 60 | 1 | 1.7% |
| USA (expected, outbound client base) | 30 | 0 | 0.0% |
| Europe | 55 | 0 | 0.0% |
| Pacific Ocean | 25 | 0 | 0.0% |
| Middle East | 14 | 0 | 0.0% |
Shipped as it stands, Glo would direct your advisor network toward Mexico and the Caribbean, and pay incentives accordingly. Not because those are your strongest partners, but because that is where the data collection reached.
314 properties across Mexico, the Caribbean and Central America produce 68 Glo designations. The remaining 200 outbound properties, across Europe, the Indian Ocean, South East Asia, the Pacific and the Middle East, produce two between them. Every one of those is a destination an American client base travels to, so where your travelers are going does not explain the gap.
An advisor selling Amalfi, the Maldives, Phuket or Dubai would receive no Glo guidance at all and would earn less on equivalent effort. Over a season that is a real distortion in both where volume goes and who on your team is rewarded for it.
Every other region mixes strong and weak data, so it can always be argued that the scores reflect real partner quality. Europe cannot be argued with. The commercial fields are complete and excellent, the relationship fields are empty for every single property, and the result is a categorical zero.
Twenty of the 55 earn full marks on commission, which is the heaviest weighted criterion in the model. That includes the Baglioni properties in Italy, Andronis in Greece, and the Myconian group in Mykonos at 26.88%, which sits above your own top band of 25%.
These are premium partners paying at or above your best rate. Not one of them can be designated.
| Highest-scoring European properties | Country | Score | Commission | Override |
|---|---|---|---|---|
| Baglioni Hotel Luna | Italy | 11 | 4 / 4 | blank |
| Baglioni Hotel Regina | Italy | 11 | 4 / 4 | blank |
| Santo Pure OIA Suites and Villas | Greece | 11 | 4 / 4 | blank |
| Casa Baglioni Milan | Italy | 10 | 4 / 4 | blank |
| Palazzo Firenze by Baglioni Hotels | Italy | 10 | 4 / 4 | blank |
| Andronis Arcadia | Greece | 9 | 4 / 4 | blank |
The top three score perfectly on every field that was filled in. Full commission, full amenities, full connectivity, full payment terms. They land on 11 and stop, one point short, because three columns nobody completed hold five points they have no way to earn.
The mean European score is 4.4 out of 18. No European property has ever been eligible for Glo, and none can become eligible by improving as a partner. The only thing that would change it is filling in the sheet.
Separate from completeness, there is a modelling question worth settling before the designation is finalised.
Commission, amenities, payment terms and connectivity are constant per booking. An override is not. It pays against a production goal, which makes it a threshold.
A property sitting just below its tier is worth far more than its override percentage suggests, because the marginal booking may trigger the entire payment. A property that will not reach its tier this year is worth only its discount. These are the same two points on the current scorecard.
Nobody can hold that across 350 properties in real time while serving an advisor on the phone. A system can, and it can surface it at the moment of booking rather than in a review afterwards. That is the single highest-value thing this dataset could be made to do, and it is not what it currently does.
In this order, because each one makes the next worth doing.
Override, marketing and res support across all 544. Until these are populated, no ranking built on this data can be trusted, and no incentive attached to it is fair.
The sheet cannot currently distinguish "no override" from "not yet recorded". Those are different facts and they should not score the same. Unknown should be visible, not silently worth nought.
Score proximity to the production goal, not the headline rate. This requires booking and production data joined to the properties list, which is the question I would want answered on the call.
At 12 you designate 75 properties. At 13 you designate 35. That is a material difference in how concentrated your steering becomes, and it should be decided against completed data rather than before it.
Every property on the Base Data tab was scored using the rules exactly as written on the Criteria tab and the header rows of ID GLO Score. Commission bands, amenity tiers, override bands, marketing bands, blackout penalties, res support, connect and payment were transcribed, not interpreted.
No property was added, removed or altered. No figure on this page is estimated. The script that produces every number here runs in a few seconds and is yours, so your team can re-run it against a corrected sheet and watch the distribution change.
Where I was uncertain, I have said so rather than smoothed it over. The Marketing column shows both a percentage ladder and an Amount ladder on the Criteria tab. The data is Amounts, since 193 of the 194 numeric entries are at or above 1,000, so the Amount ladder is the one applied here, with Pending scored at 1. A small number of commission values sit above 25% and are scored at the top band, and "Partial" commission entries are scored at 2 per the criteria note. If any of those readings is wrong the numbers move, and I would rather you correct me than have me guess quietly.
Five things I cannot know from the file, each of which would change the conclusions.
One. Are the sparse fields genuinely uncollected, or held somewhere the spreadsheet does not reach?
Two. Is the regional skew an artifact of collection order, or does it reflect where your override agreements actually are?
Three. What system did you move to a year ago, and what will it let a ranking connect to?
Four. Who assigns the points, and how often are they re-scored?
Five. Can booking and production data be joined to this list? Without it, override proximity cannot be computed at all.
The model is right. It is measuring the correct eight things, and the override insight inside it is the most valuable idea here. The issue is only that the data cannot yet carry it, and that is fixable in a way the design would not have been. RIO COSTA · UPRISE