Building a luxury tone profile that doesn't sound stiff
The five rules we shipped after analyzing 100,000 high-end concierge replies.
Sho Ito
Head of Research · Feb 24, 2026 · 6 min read
Luxury tone is the hardest preset we ship. Get it slightly wrong and it sounds like an SAT vocabulary test. Get it slightly wronger and it sounds like a Victorian mortician. We analyzed 100,000 replies from concierge teams at hotels, watch brands, and private banks to figure out what actually makes a reply read as luxury rather than performative.
Rule 1: short, not formal
Average sentence length in our luxury corpus is 11 words. Average in our "professional" corpus is 17. Luxury is closer to friendly than to formal — the assumption is that the customer's time is worth more than yours, so you don't waste it.
Rule 2: contractions, mostly
Counterintuitively, luxury replies use contractions 73% of the time. "We'll have that arranged" reads warmer than "We will have that arranged," and warmth is what the customer is paying for. The exception: apologies. A genuine "I am sorry" carries more weight than "I'm sorry." Save the un-contracted form for the moments that need it.
Rule 3: name them, twice
Address the customer by name in the opening and the close, never in the middle. Mid-reply name usage reads like a sales script. Bookending reads like attentiveness.
Rule 4: no exclamation points
Zero. Across 100,000 replies, the top-rated 5% used exclamation points 0.3% of the time vs. 8.7% in the bottom 5%. If you need emphasis, use a short standalone sentence. "It's already done." is more luxurious than "It's already done!"
Rule 5: never apologize for the work
"Apologies for the delay" reads as defensive. "Thank you for your patience" reads as gracious. Both acknowledge the wait, but one centers your inconvenience and the other centers the customer's experience. Luxury tone always centers the customer.
What this looks like in QuickPly
All five rules ship in the "luxury" tone preset. They're not absolute — the model can break any of them when the context demands it — but they're the priors. If you want to test it on your own copy, paste a recent reply into the dashboard's tone analyzer and look at the score on the "warmth" axis. That's where these rules show up first.