Lookalike audiences for hotels: a setup checklist using first-party guest data
You will be able to build a Meta lookalike audience that finds new guests likely to book direct, seeded from consented guest WiFi opt-ins joined to booking-engine data.
Seed hotel lookalike audiences from your best direct-booking guests, not everyone who stayed. Combine consented guest WiFi opt-ins with booking-engine data, then upload 1,000 to 50,000 contacts, as Meta recommends. Start at a 1% lookalike and exclude existing guests. Judge it on cost per direct booking against OTA commission, measured with a holdout test.
What can lookalike audiences do for your hotel?
A lookalike audience is a group of people Meta selects because they resemble a list you supply. That list is a custom audience, often called the seed. Meta compares the matched profiles in your seed with people in the country you choose. It then finds the closest matches across Facebook and Instagram.
For a hotel, one outcome matters: new guests who book direct. Every booking through your own booking engine avoids the commission an online travel agency (OTA) charges on that stay. It also gives you the guest's real contact details before arrival, not after checkout.
The seed decides everything. Meta does not know which guests you value. It copies the pattern in whatever list you give it. If you seed from everyone who stayed, you blend OTA bookers into that pattern. If you seed from guests who booked direct, returned and spent on site, you ask Meta to find more of them.
There is a practical problem too. OTAs often pass the hotel a relay email address rather than the guest's own. Booking.com, for example, gives partners an alias address for guest messages. Those addresses rarely match a Meta profile, so OTA-heavy seeds shrink at upload.
Where guest WiFi fits
Guest WiFi changes the data you hold. When a guest logs in on property and opts in, you capture their own email address with a recorded consent. That is first-party data, and it includes guests who originally booked through an OTA.
Tools such as Mailchimp, Klaviyo and HubSpot send campaigns to a list you bring. They do not build that list from the people who walk through your door. With Purple Engage, the list builds itself from guest WiFi logins. Purple runs at 80,000+ live venues and recorded 440 million logins in 2024 (Purple data). The Purple Engage Product overview covers the platform in full.
For the wider difference between a venue CRM and an email platform, read Hotel CRM vs email marketing tool: what a venue CRM adds.
What do you need before you build a hotel lookalike?
You need six things in place before the first upload. Missing any one of them either blocks the lookalike or makes the results impossible to read.
| Requirement | Where it comes from | Why it matters |
|---|---|---|
| Consent and notice that cover ad matching | Your WiFi opt-in wording and privacy notice | GDPR Articles 6 and 13 require a lawful basis and transparency |
| A seed of 1,000+ matched guests | Purple Engage profiles joined to booking records | Meta requires 100+ people from one country and recommends 1,000 to 50,000 |
| Booking channel per guest | PMS or booking-engine reservation export | Separates direct bookers from OTA bookers |
| Spend or stay value per guest | PMS folio totals | Allows a value-based lookalike |
| Meta Pixel and Conversions API on your booking engine | Your booking-engine provider | Reports direct bookings and revenue back to Meta |
| A baseline cost per direct booking | Last quarter's paid social and OTA reports | Gives you a number to beat |
Settle the consent question first
Consent to receive your emails is not the same as consent to match an email address on Meta. Under GDPR, you need a lawful basis for that processing (Article 6). Your privacy notice must also tell guests you do it (Article 13).
The UK Information Commissioner's Office (ICO) covers list-based audience tools in its direct marketing guidance. Many hotels rely on consent and name social advertising in the opt-in wording. Check your wording with your data protection officer before the first upload. Meta's Custom Audience terms also require you to hold the rights to use the data you upload.
GDPR Recital 32 rules out pre-ticked boxes as valid consent. Purple's conscious-choice opt-ins ask the guest to make an active choice at login. That gives you a consent record you can stand behind if a guest or regulator asks.
How do you build a lookalike from guest WiFi and booking data?
Work through these eight steps in order. Steps 1 to 4 happen in your own data. Steps 5 to 8 happen in Meta Ads Manager, where Meta's Business Help Centre documents the current screens.
1. Capture consented emails at every login
Your seed can only be as large as your opt-in list. Make the opt-in clear and say what guests will receive. If you rely on consent for ad matching, include a plain line about social advertising.
2. Join WiFi profiles to booking channel
Export reservations from your PMS with email, booking channel, stay dates and folio total. Match them to Purple Engage profiles on email address. Flag every guest as direct, OTA, or group and corporate.
Guests who booked through an OTA but opted in on WiFi are valuable, just not as seed. Move them into a direct-rebooking email programme in Purple Engage instead. Their next stay should not carry a commission.
3. Choose the seed
The table compares the seed options most hotels have. Start with the narrowest seed that still clears 1,000 matched people.
| Seed option | What it teaches Meta | Main risk | Use it when |
|---|---|---|---|
| All WiFi opt-ins | Anyone who visited, including restaurant and bar guests | Mixes non-staying visitors with staying guests | Your F&B trade is a main revenue line |
| All direct bookers | People who chose your own website | Includes one-off bookers | Default starting seed |
| Repeat direct bookers (2+ stays) | Loyal guests who return without an OTA | Can fall below 1,000 matched | You hold 1,000+ such guests |
| High-value direct bookers with a value column | Guests with the highest folio spend | Group master bills skew the values | You want spend, not only bookings |
| OTA bookers | People who book through intermediaries | Relay addresses fail to match | Never as seed; use as an exclusion |
4. Add a value column
Meta supports value-based lookalikes, where each person in the seed carries a number. Use total folio spend over the last 24 months. Strip out group master accounts and comped stays first, or one conference bill will distort the pattern.
5. Upload the seed as a custom audience
Upload email plus phone number, first name, last name and country, where you hold them with consent. Meta matches on every identifier you provide, so more fields give it more chances to match. Ads Manager hashes the identifiers before they leave your browser.
Check the matched size after upload. If it falls under 1,000, widen the seed from repeat bookers to all direct bookers. Do not pad it with OTA records.
6. Create the lookalike
Choose the country your guests come from. The source must contain at least 100 people from a single country. Start with a 1% lookalike, the closest match Meta offers.
Meta lets you go up to 10%, and larger percentages trade similarity for reach. Build a 1% to 3% band as a separate ad set later, once the 1% has a baseline. If you draw guests from several countries, create one lookalike per main market so each can be read on its own.
7. Exclude the people you already know
Exclude a custom audience of all past guests and recent website bookers. You are paying to reach strangers. Your existing guests belong in Purple Engage email campaigns, where reaching them does not cost you an impression.
8. Refresh on a schedule
A list-based custom audience does not update itself. Replace the seed monthly with a fresh export. Each month, new direct bookers who logged in to guest WiFi join the seed, so the pattern keeps improving.
What should you run against a hotel lookalike, and when?
A lookalike is a cold audience. These people have never stayed with you. Lead with the property and the reason to book direct, not a discount that trains them to wait for one.
Direct-booking perks you control work well because they are specific: late checkout, breakfast included, a room upgrade on arrival. Use creative that shows the rooms, the food and the location your repeat guests already rate.
Match the seed to the gap you need to fill
| Your situation | Seed | Lookalike size | Lead message |
|---|---|---|---|
| Independent hotel, under 1,000 repeat direct bookers | All direct bookers | 1% | The property plus one direct-booking perk |
| Established hotel, 1,000+ repeat direct bookers | Repeat direct bookers | 1%, then 1% to 3% | Reasons guests return: rooms, food, location |
| Hotel with a strong spa or restaurant | High-value direct bookers with a value column | 1% | Packages that include on-site spend |
| City hotel with weekday corporate demand | Weekend leisure direct bookers only | 1% | Weekend breaks |
| Guests from two or more countries | One seed per main country | 1% per country | Creative localised to each market |
Time it from your own booking window
Use your PMS pickup report, not a generic calendar. Find the dates where occupancy lags last year. Then launch ahead of them by your typical booking lead time, which the same PMS data shows. A beach hotel with a 60-day leisure window and a city hotel with a 10-day window need different launch dates for the same gap.
Close the loop on arrival
The lookalike does its job when a stranger books direct. Purple's job starts when that guest arrives and logs in to guest WiFi. Their opt-in enters Purple Engage, which sends post-stay and return-stay emails. Next month, that guest joins the seed. Each cycle gives Meta a larger and more accurate pattern of the guests you want.
How do you know your lookalike audiences are working?
Clicks and cheap impressions tell you Meta found people. They do not tell you those people booked. Judge the lookalike on the six measures below, in this order.
| Metric | How to read it | Warning sign |
|---|---|---|
| Cost per direct booking | Ad spend divided by booking-engine purchases attributed to the ad set | Higher than the OTA commission on an equivalent booking |
| Direct booking revenue per £1 spent | Booking value reported through the Pixel and Conversions API | Falling while cost per thousand impressions also falls |
| New-guest share | Bookers' emails matched against Purple Engage and PMS history | Most bookers are already in your database |
| Incremental bookings | A/B test or holdout against a broad audience | No difference between lookalike and broad |
| Matched seed size | Custom audience size after each refresh | Under 1,000, or shrinking between refreshes |
| Frequency | Average number of times each person saw the ad | Rising every week while bookings stay flat |
Test against a control
Meta's A/B testing feature splits delivery so the two groups do not overlap. Run the lookalike against a broad audience with the same creative, budget and dates. For larger budgets, Meta also offers Conversion Lift studies, which hold back a group that sees no ads at all.
Check one setting before you read any result. On some campaign objectives, Meta can deliver beyond your lookalike through Advantage+ audience or Advantage lookalike. If expansion is on, you are partly measuring Meta's own targeting. Note the setting in your test log so the comparison stays honest.
Set the break-even before you launch
The question is simple: does a direct booking from the lookalike cost less than the commission you would pay an OTA? Take your average direct booking value and multiply it by your OTA contract commission rate. That figure is the most you should pay per direct booking from paid social.
Worked scenario: a 200-room coastal hotel
Figures in both scenarios are illustrative, built to show the arithmetic. Use your own numbers.
Situation. A 200-room coastal hotel took most summer bookings through OTAs. Its first lookalike used a PMS export of 18,000 past guests. Fewer than 6,000 matched on Meta, because relay addresses failed. The 1% lookalike delivered cheap reach but few direct bookings.
What they did. The marketing team rebuilt the seed from 4,100 direct bookers with a WiFi opt-in in Purple Engage. They added a 24-month folio value column and created a 1% UK lookalike. They excluded all past guests and ran an eight-week A/B test against a broad audience.
Outcome. The lookalike cost £38 per direct booking against £61 for broad targeting. Of the lookalike bookers, 81% were new to the hotel. The average direct booking was £420. At an 18% OTA contract rate, the same booking through an OTA would cost £75.60 in commission. The lookalike cleared break-even by £37.60 per booking.
What goes wrong with hotel lookalikes, and how do you avoid it?
Most failures trace back to the seed, the setup or the way results are read. The second scenario shows the most common one.
Worked scenario: a 120-room city hotel with events space
Situation. A 120-room city-centre hotel wanted more weekend leisure guests. Its seed held every direct booker, including conference delegates booked directly by their employers. The lookalike delivered people who looked like weekday business travellers. Weekend break ads converted poorly.
What they did. The team split the seed by stay pattern. Guests with Friday or Saturday arrivals and personal email addresses formed a leisure seed. Corporate and delegate stays became a separate seed, held back for a weekday campaign. They rebuilt the 1% lookalike from the leisure seed alone.
Outcome. Over one quarter, weekend direct bookings from paid social rose from 22 to 49 a month. Cost per direct booking fell from £72 to £44. The weekday seed later ran its own campaign without diluting the weekend one.
Venues with ticketed events face the same blended-seed problem with ticket buyers. Stadium fan marketing with WiFi data: a match-day campaign checklist covers splitting fan lists by intent.
The eight failure modes to check
- Seeding from OTA bookers. Relay addresses fail to match, and the pattern favours intermediaries. Use OTA bookers as an exclusion, never as seed.
- A seed that is too small. Below 1,000 matched people, the pattern is thin. Widen from repeat bookers to all direct bookers before you consider anything else.
- A blended seed. Mixing corporate, leisure and F&B-only visitors teaches Meta a muddled pattern. Split by stay type.
- A consent gap. Uploading opt-ins that never covered ad matching breaks GDPR transparency rules. Fix the opt-in wording and privacy notice first.
- Judging on clicks. Low cost per click can mask zero bookings. Read cost per direct booking and new-guest share.
- A stale seed. List-based audiences do not refresh themselves. Replace the seed monthly.
- No exclusions. Without a past-guest exclusion, you pay to reach people Purple Engage can email at no media cost.
- Unchecked audience expansion. If Advantage+ audience or Advantage lookalike is on, record it, or your test compares two things at once.
Frequently asked questions
Is it GDPR compliant to upload guest WiFi emails to Meta for lookalike audiences?
Yes, provided you have a lawful basis and your privacy notice says you do it. GDPR Article 6 requires the lawful basis and Article 13 requires transparency. The UK ICO's direct marketing guidance covers list-based audience tools. Many hotels rely on consent and name social advertising in the WiFi opt-in wording. Purple's conscious-choice opt-ins record an active choice at login. Confirm your wording with your data protection officer before the first upload.
Does Purple Engage replace Mailchimp or Klaviyo for hotel email?
Yes, for most hotels it can, because Purple Engage both builds the list and sends the campaigns. Mailchimp, Klaviyo and HubSpot send to a list you bring. They do not capture guests as they log in on property. Purple Engage builds the list from guest WiFi logins as consented first-party data, including guests who booked through an OTA. You then email them and use the same profiles to seed Meta lookalikes.
Do I need new WiFi hardware to capture guest data for lookalikes?
No, Purple is hardware-agnostic and runs as a cloud overlay on the access points you already own. Supported hardware includes Cisco Meraki, HPE Aruba, Ruckus, Juniper Mist, Ubiquiti UniFi, Cambium, Extreme and Fortinet. Guests log in through Purple's guest WiFi journey on your existing network, and their opt-ins flow into Purple Engage. You do not need a rip-and-replace project to start building a seed.
How many guests do I need for a hotel lookalike audience?
You need at least 100 matched people from a single country, but aim for 1,000 or more. Meta sets the 100 minimum and recommends a source of 1,000 to 50,000 people. Count matched profiles after upload, not rows in your file, because some records will not match. If your repeat direct bookers fall short, use all direct bookers until your WiFi opt-ins grow.
Can I include OTA guests in my lookalike seed?
No, keep OTA bookers out of the seed and use them as an exclusion instead. OTAs often pass relay email addresses, which rarely match a Meta profile, so they shrink your seed. They also teach Meta the pattern of people who book through intermediaries. OTA guests who opt in on guest WiFi belong in a direct-rebooking email programme in Purple Engage, where their next stay avoids commission.
How long does it take to see whether a lookalike is working?
Plan for a test of six to eight weeks before you judge a hotel lookalike. Bookings arrive more slowly than clicks, and hotel booking windows can run for weeks. Run a Meta A/B test against a broad audience with identical creative and budget. Compare cost per direct booking, new-guest share and booking revenue. Set your break-even against OTA commission before launch, so the result is a yes or no.