Your Restaurant Website Is Not One Website

Watch the address bar

Open a restaurant chain's website and click Order Online. Watch the address bar.

On most chain sites it changes. Sometimes to a subdomain, order.brandname.com, which at least looks like it belongs. Sometimes to a completely different domain owned by an ordering platform. The page keeps the brand's logo and colors, so from the visitor's side almost nothing happened. From the measurement side, quite a lot happened.

What reads as one brand site is usually a federation of vendor platforms wearing a shared skin. That is the thing worth internalizing before any of the specific problems make sense. There is no single site to instrument, no single place where consent applies, and no single owner who can describe the whole thing.

The four-platform pattern

Across the restaurant sites we look at, roughly the same four functions keep turning up on roughly four different vendors: online ordering, the loyalty program, gift cards, and catering enquiries. Sometimes reservations makes five. Occasionally two of them share a vendor, which counts as unusually tidy.

This is a pattern we see, not a statistic we have measured across a defensible sample, and it is worth being clear about that distinction. But the shape is consistent enough to predict, and the reason it happens is structural rather than careless.

Each of those systems was bought at a different time, by a different part of the business, to solve a different problem. Ordering was an operations decision, probably urgent, probably made in a hurry. Loyalty came from marketing, later, and was chosen for its CRM features. Gift cards came from finance or retail and are frequently the oldest system on the list. Catering was bolted on when catering became a line of business. Nobody in that sequence was choosing a measurement architecture. They were each solving a real problem with the best available vendor at the time, and the sprawl is what those decisions add up to years later.

So this is normal, not negligence. It is also unusually expensive in three specific ways.

One: the handoff strips your campaign context

A visitor clicks a paid ad. The landing page receives campaign parameters in the URL, analytics starts a session, and the source is recorded correctly. So far so good.

Then they click Order Online and land on a different domain. Unless cross-domain measurement was configured deliberately, the campaign parameters do not travel, the analytics session ends, and a new one begins on the ordering platform with no idea where the visitor came from. The order completes over there. The revenue is recorded over there, in the vendor's system, on the vendor's terms.

Two things follow, and both are bad in quiet ways. Paid traffic starts reporting as direct or referral, because by the time the purchase happens the campaign context is gone. And the campaign that generated the order cannot see the order, so whatever bidding system is optimizing that campaign is learning from an incomplete picture of what worked.

Notice what this does to the internal argument. Direct traffic looks strong and paid looks weak, so the obvious read is that the brand is doing the work and the media budget is not earning its place. That read is available in every dashboard, it is well supported by the visible numbers, and it can be exactly backwards. We are not going to put a loss rate on it, because the size depends entirely on how each handoff is configured, and anyone quoting a universal figure for this is guessing. What we will say is that the mechanism is real and it points in one direction: it always makes paid look worse than it was.

If your conversion volume in the ad platform has never quite matched what the ordering system says it sold, this is usually the first place to look. It sits alongside the more general patterns in 5 Signs Your Conversion Tracking Is Broken and the reconciliation problem in Why GA4 and Google Ads Never Match.

Two: your consent banner stops at the boundary

The brand site has a consent banner. A visitor declines analytics and advertising cookies. Then they click through to order, and they are on a domain the banner has no authority over.

Consent stored on one domain cannot be read by another. That is not a configuration failure, it is how browsers work, and it means the vendor platform starts from nothing. It either has its own consent mechanism, which may or may not match the one the visitor just used, or it has none, in which case the visitor's decline stopped applying at the exact moment they started transacting. We covered the mechanics of this at length in Your Checkout Is on a Different Domain, and the same boundary problem shows up across regional domains in The Consent Gap on Your International Domains.

So the compliance posture the brand believes it has usually describes the brand site only. And the brand site is the marketing layer. The vendor domains are where the actual personal data lives: names, phone numbers, delivery addresses, order histories, payment steps, loyalty identifiers tied to real people. The privacy program covers the pages that collect the least and stops at the systems that collect the most.

On accountability, the instinct is to say the vendor's domain is the vendor's problem. It is worth testing that instinct against how a customer or a regulator would see it. The customer chose your brand, on your site, following your ad, and they have no idea a boundary was crossed. Contractual responsibility varies and is a question for your counsel. Practical responsibility does not really move.

Three: nobody owns the map

Ask a straightforward question inside a restaurant brand: which vendor platforms does a customer touch between the homepage and a completed order, and who owns each one? In our experience the answer takes days and comes back partial.

That is not a knock on anyone. It is the arithmetic of separate procurement over years. Operations owns the ordering relationship. Marketing owns loyalty. Finance may own gift cards. IT owns the DNS records that point subdomains at all of them, without necessarily knowing what each subdomain does. The people who signed the original contracts have often moved on. Each team can describe its own vendor accurately and nobody has been asked to describe the whole path.

The consequence: no single person can answer whether the customer journey is measured correctly or governed correctly, because no single person can see the journey. And a map assembled from contracts and procurement records is not good enough, because it tells you what was bought, not what is currently loading in a visitor's browser. Those diverge. Vendors get replaced, embeds get added, subdomains get repointed, and a tag someone installed for a campaign in 2023 is still there.

One journey, counted

Here is a composite journey, assembled from the pattern rather than from any one brand. A regular customer, doing ordinary things.

1. Homepage, from a paid ad. Brand domain. Consent banner appears. Campaign parameters intact. Everything working as designed.
2. Order Online. Ordering platform, different domain. New session, campaign context gone unless someone configured otherwise. Different consent surface, or none.
3. Sign in to apply loyalty points. Loyalty platform, third domain, often in an embedded frame or a redirect and back. This is where the customer identifier gets involved.
4. Pay with a gift card balance. Gift card processor, fourth domain, frequently the oldest system in the chain and the least likely to have been reviewed recently.
5. Order confirmation. Back on the ordering platform. The conversion, if one is recorded at all for the brand, is recorded here, three boundaries away from the ad that started it.
6. A week later, a catering enquiry. Fifth platform, fifth domain, its own form, its own data store, its own tracking, and no connection to any of the above.

Four or five distinct vendor domains in one ordinary customer relationship. Every boundary is a place where campaign context can drop, where consent state does not carry, and where the data lands in a system somebody else operates and updates on their own schedule.

And the brand may not learn when any of it changes. A vendor ships a release, a page structure moves, a tag stops matching, and nothing on your side raises a hand. The report just carries on looking normal.

What to actually do

None of this argues for consolidating vendors. That is a multi-year procurement fight with real operational costs, and the sprawl is usually the correct answer to the business problems that created it. The fix is governance over the federation you already have.

1. Name an owner for the map. One person accountable for knowing every platform in the customer journey and every domain it runs on. Not a committee. The map has no owner today, which is why it does not exist.
2. Build the map from the visitor's side. Walk the journey in a fresh browser and record every domain the browser actually contacts and every tag that actually loads. Do not assemble it from contracts. Contracts describe intent; the browser describes reality, and the gap between them is the interesting part.
3. Configure cross-domain measurement deliberately. For each handoff, decide whether the session should continue and make it do so. This is well-trodden technical work. It is skipped because nobody owned the handoff, not because it is hard.
4. Scope consent to cover the vendors. Work out what consent state each platform can accept, then make the brand's decision reach it. Where a platform cannot accept one, that is a finding to escalate rather than a detail to absorb.
5. Monitor the boundaries on a schedule. The vendor will change something on their release schedule without telling you. Scheduled checks on the handoff points are the only way you find out from a monitor rather than from a quarter of soft numbers.

Step two is the one that changes the conversation, because it produces a document that did not exist and that nobody can argue with. Most of the surprises in this work come from that walkthrough, and they usually arrive in the first hour.

The site your customer sees

Your customer thinks they spent the whole visit on your website. Your analytics thinks they left after ninety seconds. Your consent banner thinks it covered them. Your ad platform thinks the campaign did not work.

All four are describing the same visit, and only the customer is right about what happened.