Your returning guest looks new to every system you own
The same guest booked through an OTA, paid at the bar under a room number, joined the wifi under a personal address and reviewed under a nickname. Four systems hold four different people, and none of them is the guest.
Your returning guest booked through an online travel agent in March, paid at the bar in July under a room number, joined the wifi under a personal address and left a review under a nickname. Four systems hold four different people. None of those four is the guest.
Personalisation projects usually solve the wrong half
The pitch is always a platform that unifies guest data. The unspoken assumption underneath it is that the data was captured with something to unify it by. Usually it was not. The travel agent masked the email, the point of sale recorded a room number rather than a person, the wifi portal never wrote anything back, and the property management system created a fresh profile at every booking because the name was spelled slightly differently.
The key is a decision, not a technology
Before any tool, somebody has to answer one question: what identifies a guest here. Email is the obvious candidate and the agent channel breaks it. Phone survives more of the journey. A loyalty number works if enrolment happens at a moment the guest has a reason to enrol, which is almost never at check-in, when what they want is their key.
Whatever the answer turns out to be, it has to be captured at every touchpoint that matters and it has to be the same field every time. That is unglamorous work, and it is the whole project.
What becomes measurable once the key exists
- What a returning guest is actually worth against a first-time guest, net of channel cost.
- Which stay-level experiences correlate with a return, and which correlate with nothing at all.
- How much of your new demand is a previous guest you failed to recognise.
That third number is the one that moves budgets. Most properties assume a repeat rate from the loyalty programme. The real rate includes everyone who came back without ever being counted.
Start smaller than the vendor suggests
Pick one key. Capture it at two touchpoints rather than eight. Prove you can match a stay to a previous stay across a single month of arrivals, by hand if that is what it takes. If it fails at two touchpoints it will fail at eight, and you will have paid for a platform to store data that never joined.