Policies

How we rank things

Last updated July 28, 2026

This page is the authoritative statement of how ranking works in Voyade. If anything anywhere else — a marketing page, a support answer, a post — says something different, this page is right and that is wrong. It is dated, it is versioned at the bottom, and it is updated before a change ships, not after.

It is also, deliberately, the page you land on from “how we make money”, because the honest answer to that question and the honest answer to “why am I seeing this” are the same document.

The short answer

As of 28 July 2026, nobody pays to rank higher in Voyade. Not a hotel, not a tour operator, not an affiliate partner, not a creator, not us. §6 states that precisely, with the date on it, and sets out what we would have to do before that could ever change.

Commission does not influence ranking, suggestions, or the order of booking offers. It is not only a policy we ask you to trust: the code that ranks and suggests does not read commission data. §6.2 explains exactly how far that guarantee goes and where it stops, because “enforced in the architecture” is a claim worth being precise about.

There is one thing that genuinely is commercial and we would rather say it here than bury it: which booking partners exist at all is a commercial decision. We can only show you an offer from a company we have a programme with. That shapes the set of offers you see, and it is not the same thing as shaping their order. §4.2.

There are four places ranking happens, and each is explained below.

1. How we make money

Two ways, and no third.

Commission on some bookings. When you book a hotel, a tour, a ticket, a car or a flight through a link in Voyade, the partner usually pays us a share of what they earn. You pay the same price; the commission comes out of the partner’s margin. This is most of our revenue and it is why the free tier can be the whole product.

An optional subscription. €5.99 a month or €39 a year for Premium. It upgrades exactly one account.

And nothing else. No advertising, anywhere in this product. No sponsored slots. No sale of personal data, ever. No placement fees, no listing fees, no inclusion fees.

Plenty of what we show earns us nothing at all: restaurants, most free activities, the direct booking links in §4.3, and any partner we have not signed. Those are not marked as earning, because there is nothing to disclose.

When a creator’s trip produces a booking, they take a share of what we received — 70% when they brought the traveller, 30% when our feed did. That is set out in the Creator terms §5 and it changes nothing about ranking; see §7.

2. The feed

The feed is a list of complete, cloneable trips. This section describes the ranking that is actually implemented today, which is deliberately less than the ranking we intend to have. Publishing the ambition instead of the behaviour would make this page untrue, which is the one thing it cannot be.

2.1 What determines the order today

Which templates are eligible at all. Only templates that are published, and only canonical ones. A near-duplicate of an existing template is attached to that template as a traveller take rather than appearing as its own card — that is the variety rule, and it is in Creator terms §6.

Then the filters you set, which narrow rather than reorder: region or destination (including everything beneath it in the destination tree), trip length, month, budget band, pace, group type, and vibe tags.

Then one ordering, which you choose:

Order	What it sorts on	Default?
Recent	Most recently published first	Yes
Most saved	The number of people who saved it	No
Most cloned	The number of people who cloned it	No

That is the whole of it. There is no blended score, no secret weighting, and no arithmetic — it is one column, descending, with the template’s own identifier breaking ties so the order is stable between two page loads.

The featured trip at the top of the signed-out front page is the most-cloned template matching your filters, with saves breaking the tie.

Signed in, the feed adds three rows above the grid, in this order:

Because you cloned… — built from the most recent trip you cloned, matched on shared vibe tags, then the same destination, then the same group type, ordered by clone count. Templates you have already cloned or saved are excluded, because showing you a trip you already took is not a recommendation.
From creators you follow — their templates, most recently published first.
Saved for later — the ones you saved, most recently saved first.

Browsing signed out gives you no personalisation at all: the same feed as everybody else, on the ordering you picked.

2.2 What is not yet in the ranking, and we are not pretending it is

We intend the feed to reward the quality of a template and what travellers actually did with it. Today it does not. None of the following influences the order:

How complete a template is, or whether it survived an editorial pass.
Whether its photos are approved, or its cost bands honest.
Whether the routing makes sense.
How many clones became real trips with a confirmed booking.
Whether opening hours and prices have gone stale.

When any of these is added to the code, it is added to §2.1 in the same change. Until then, a recent template outranks a better older one, and you should know that.

2.3 What is not in the ranking and never has been

Commission rates. Affiliate revenue. Whether a template contains bookable items at all. Whether its creator subscribes to Premium. Whether its creator has a large following. Any payment of any kind. There are no sponsored slots and no promoted templates.

3. Suggestions inside a trip

When the Idea Pool offers you shelves of things to do, or the AI stitches a day together, or Nearby answers “what is worth doing in the next three hours”, this is what it is reasoning from.

3.1 What goes into an AI suggestion

The gaps in your plan. A day with no dinner, an evening with nothing on it.
Your group’s votes and constraints — the budget band, the pace, dietary and mobility needs, and the private budget caps, which are used only to derive whether the plan fits and are never shown to anyone.
Real candidate places, retrieved before the model reasons about them. The model is given a list of places that exist, with their opening hours and price band from a live source. A suggestion the model returns that was not in that list is marked as unverified rather than presented as fact. We do not let it invent an opening time.
What is already in your pool, so it does not suggest what you have already had.

3.2 Nearby: the actual weights

Nearby is the one surface with real numeric weights, so here they are. A candidate place starts at zero and is adjusted:

Factor	Effect on the score
It is indoors and the weather is bad right now	+3.0
It is already in your group’s idea pool	+4.0
Its rating, relative to 4.0 out of 5	(rating − 4.0) × 0.4
It is outdoors and the weather is bad right now	−1.5
There are children in the group and it needs more than 45 minutes on foot	−2.0
Your group’s pace is “chill” and it is 45 minutes or more away	−1.0
It would be dark by the time you arrived	−4.0

Equal scores are broken by whichever is closer, then alphabetically. Before any of that, a candidate has to fit the time budget including the journey back, be open when you would arrive, not collide with something you have already booked, be within the group’s stated mobility, and not be something you have already done or scheduled for another day.

Commission is not in that table, and there is no row missing from it.

3.3 The free-things-rank-equally rule

This one is worth stating on its own, because it is the rule most likely to be quietly broken by a product that lives on commission.

A free thing and a bookable thing compete on identical terms. The clifftop walk, the free-entry Sunday at the museum, the swimming spot, the session pub with music at nine — these are ranked by how good they are for your group, exactly like a €45 guided tour. They carry the same settled badge, the same prominence, the same card.

Every suggestion is labelled: Free, or a price band, and Bookable / Reserve ahead / Just show up. So you can always see which is which.

The consequence is one we accept on purpose: the best evening in Galway is often a €0 session pub, and we will recommend it over a bookable pub crawl, and earn nothing. The canonical case is car hire, where our own guide tells you not to buy the excess insurance at the rental desk — advice that costs the rental company money and would cost us a commission if we had one there.

4. The order of booking offers

4.1 How offers are ordered

When a booking task shows you two or three options for a hotel, a tour or a car, they are ordered by a fit score computed for that task, then by an editorial position we set by hand, then by the order they were added. The fit score is meant to weigh:

Fit to the actual requirement — your dates, your party size, and how far it is from that day’s route or your anchor for the night. A cheaper hotel forty minutes the wrong way ranks below a dearer one on the route.
Price, in your display currency, as advertised by the partner when we fetched it.
Cancellation flexibility, because a plan that is still moving needs it.
How reliable the data is — a live price from a partner API outranks a cached estimate, and we mark which is which.

Not: the commission rate, the size of the partner, whether the partner has an API relationship with us, or any payment.

Where this is today. The fit score is modelled and the ordering reads it, but the pipeline that computes it is not live yet. Until it is, offers appear in the order we set by hand, which is editorial and is not influenced by what an offer earns. This paragraph goes when the scorer ships.

4.2 The one commercial fact: which partners exist

Ranking is not the only thing that can be bent by money, and we would rather name the other one than let you find it.

We can only show you an offer from a partner we have a commercial programme with. Whether a company is in that set at all is a business decision, and commission terms are part of why we sign one partner rather than another. It affects the set of offers, not their order, and it is the honest limit on “commission never influences what we show you”.

Two things we do about it. First, the direct option below. Second, the price you see is the partner’s own advertised price — being in our set does not let a partner charge you more, and we do not take a cut from you.

4.3 The direct option

Alongside the partner offers, we show a direct, unaffiliated way to book — the property’s or operator’s own site — wherever we can identify one. We earn nothing from it. Sometimes it is cheaper; sometimes the cancellation terms are better; sometimes it is simply the right thing to do.

It sits in the same list, in the same style. It is not a footnote and it is not hidden behind “more options”.

4.4 Disclosure

Every link that can earn us a commission carries the same line, every time, on every surface:

We may earn a commission on some links. It never affects what we show or the order we show it in.

If a creator’s template earns them a share, the template says so on its own page: “Maya earns a commission if you book through this trip.” That label is applied by us automatically — a creator cannot switch it off, and it is not their job to remember. See Creator terms §11.

5. Search, guides and community

Search ranks on text relevance, the same signals as the feed, and your filters. Nothing commercial.

Guides are ordered by relevance to what you are doing right now — the car booking task surfaces the excess-insurance guide, an Ireland trip surfaces the driving guide. Guides contain affiliate links and are disclosed as such. A guide’s position is never influenced by what its links earn, and the excess-insurance guide is the proof: it tells you to buy a third-party policy instead of the one at the desk, and we would publish it if it earned us nothing, which we say inside the guide itself.

Community answers rank on how many people found them helpful, whether one has been accepted by the person who asked, and whether the answerer actually went — an answer from somebody who was there in May 2026 carries that, visibly. Reputation cannot be bought.

6. Does anyone pay to rank higher?

This is the section the law actually asks about, and the section that would have to change first if the answer ever did.

It is written as a statement of the current position with a date on it, and not as a promise about all future time. That is deliberate. A company that promises “never” and later needs the revenue rewrites the promise and hopes nobody kept a copy. We would rather publish something we can keep accurate: what is true today, what we have not ruled out, and — the part that binds us — the process we are committing to if it ever changes.

6.1 The current position

As of 28 July 2026, no direct or indirect remuneration of any kind influences ranking anywhere in Voyade — not in the feed, not in suggestions, not in Nearby, not in the ordering of booking offers, not in search, not in guides, not in community.

Specifically, none of the following exists in this product:

Sponsored or promoted placements.
Paid feed slots.
Higher commission rates causing an offer to appear higher.
Partners paying for position, inclusion, a badge or a label.
Creators paying for feed placement.
Any commercial factor of any kind in the code that ranks or suggests.

6.2 How far the architectural guarantee goes

We say the recommender “cannot see” commission data, and that deserves to be exact.

What is true. Commission rates live on the partner and conversion records. The code that orders the feed, the code that builds AI suggestions and the code that scores Nearby candidates do not reference those records at all — not the rate, not the earnings ledger, not a conversion. There is no path from a ranking query to a commission figure.

That is not left to good intentions. A test reads those files on every build and fails if any of them so much as names a commission table, a commission rate or the earnings ledger. It is deliberately blunt: it catches the innocent version of this mistake — an association added to make a page faster — which is the version that actually happens.

What is not true, and we will not pretend it is. This is a single application over a single database. The separation is a boundary we maintain and test, not a wall the database itself enforces — nobody has taken the commission tables away from the process that ranks. If the boundary were removed, the code would still run; it is a test that would fail, not a permission.

What that means for you. It is stronger than a policy and weaker than a law of physics, and the honest description is: an invariant we test on every build.

6.3 What we have not ruled out, and what binds us if it changes

We have not promised never to weight a commercial factor. A future Voyade might, for instance, place a clearly-marked sponsored result in a search, or take part in a partner programme that pays for a position.

If we ever do, all of the following bind us, and they are the commitments this section exists to make:

This section is rewritten first. Before the change ships, not after. In plain language, saying exactly what is paid for, by whom, how it affects position, and how you can tell.
It is labelled where you see it, at the result itself, in a word that means what it says. Not “featured”, not “enhanced visibility”, not “our picks”.
The change is dated in the change log at the bottom of this page, and existing users are told in the app.
The rules in §3.3 and §4.3 survive it. Free things still rank equally with bookable ones; the direct, unaffiliated booking option is still shown in the same list, in the same style.
The guide standard survives it. If a recommendation would be different without the money, it does not go up. That is what the excess-insurance guide is for.

Until then, §6.1 is the answer, and it is dated so you can tell how fresh it is.

7. Do we treat our own things better than other people’s?

The honest answer is: we publish into the same feed, on the same terms, and here is where the differences actually are.

Voyade’s own editorial templates and creators’ templates sit in one feed, ranked by the same rules in §2. An editorial template gets no boost for being ours. A creator’s template gets no penalty for being theirs.

What is different:

We wrote the seed catalogue. At launch most templates are ours, simply because we made them before anyone else could. That is a starting condition, not a weighting.
Editorial templates carry a different attribution line — “Voyade editorial” rather than a creator’s handle — and they are not attached to a creator’s earnings.
We can edit any template to correct a stale fact: an opening time, a closed restaurant. Those edits are logged. We do not use them to change a template’s ranking or its substance.
We decide which booking partners are in the set at all. See §4.2.

What is not different: the ordering, the eligibility rules, the canonical-template rule, and the appeal route. A creator who thinks their template has been ranked or suppressed unfairly appeals to a human — Creator terms §6.

8. Personalisation, and turning it down

Feed ranking is personalised for signed-in travellers. Suggestions inside a trip are personalised to the group.

To see the feed without personalisation, browse signed out. That is a complete switch, not a softening: signed out there is no personalisation at all.
To see the reasoning behind an AI suggestion, open the proposal — every change carries a written rationale, and that is enforced in the data model, not a nicety.
To stop a signal feeding your feed, delete the thing that produced it: unsave a template, delete a trip.

Product note. A “turn off personalisation” switch in Settings does not exist yet, and this page will say so until it does. It is on the list.

9. How to challenge this page

If you think the product is not behaving the way this page describes — an offer order that looks commercial, a guide that reads like an advert, a template that seems promoted — tell us at [email protected] and say what you saw. We will look, and we will answer you within 14 days. If we were wrong, we will fix it and note it in the change log below.

If you are a creator and you believe your template has been ranked or suppressed unfairly, the appeal route is in Creator terms §6 and it goes to a human who was not part of the original decision.

Change log

Date	What changed
2026-07-28	First draft. Not yet in force. No commercial weighting anywhere in the product. §2 rewritten to describe the ranking that is implemented rather than the ranking intended, with the unimplemented signals listed in §2.2. §4.2 added, disclosing that partner selection is a commercial decision. §6 restructured as a dated current position plus a binding change protocol. §7 added: differentiated treatment.

Date

What changed

2026-07-28

First draft. Not yet in force. No commercial weighting anywhere in the product. §2 rewritten to describe the ranking that is implemented rather than the ranking intended, with the unimplemented signals listed in §2.2. §4.2 added, disclosing that partner selection is a commercial decision. §6 restructured as a dated current position plus a binding change protocol. §7 added: differentiated treatment.