You wrote the message six times. You softened it, then softened it again, and by the time you hit send it read like a hostage note written by someone who wanted to be liked: Hey! Amazing job as always. Just one tiny thing — the guest mentioned the shower door had some water spots? No big deal at all!! You're the best. And then the next turnover was somehow worse. Not dramatically. Just — flatter. The little things she used to do without being asked, the fan of towels, the note by the coffee maker, quietly stopped. And you couldn't shake the feeling that you had broken something you didn't know how to fix.

Here is the uncomfortable part. You probably did. Not because you gave feedback, and not because you gave it too harshly. Because of what your feedback made her think about.

Feedback is not automatically helpful, and the research is blunt about it

In 1996, Avraham Kluger and Angelo DeNisi published a meta-analysis in Psychological Bulletin that pulled together decades of feedback studies — hundreds of effect sizes across workplaces, labs, and classrooms. The average effect of feedback on performance was positive, which is the finding everybody quotes. The finding nobody quotes is the one that matters to you: in more than a third of the cases, feedback made performance worse. Not neutral. Worse than saying nothing at all.

Their explanation is called Feedback Intervention Theory, and once you see it you can't unsee it. Feedback works by directing attention. Attention is finite. Where the feedback points the person's attention determines everything that happens next.

If feedback points attention at the task — the shower door has hard-water spots; a quick pass with the squeegee handles it — the person's cognitive resources go toward solving the problem. Learning happens. Performance rises.

If feedback points attention at the self — are you happy with me, am I about to lose this job, does she think I'm sloppy — the resources go toward managing the threat instead of the task. The person is now doing two jobs: cleaning, and defending their sense of being a competent person. The cleaning gets the leftovers.

And here is the trap. The more you soften, hedge, and pad a piece of feedback, the more clearly you signal that something bigger is at stake. Nobody writes you're the best!! twice unless they're managing an emotional risk. Your cleaner reads the padding, not the shower door. The message she receives is: this is a conversation about whether she is good enough. You aimed at the task and hit the self.

Why the feedback sandwich makes this worse, not better

The advice most hosts absorb somewhere is to sandwich criticism between two slices of praise. It feels kind. It is, in practice, a machine for producing self-focused attention.

Two things go wrong. First, the person learns the shape. After the second sandwich, praise stops being praise and becomes the sound of a knife being unwrapped. You have now made your compliments unusable — every kind word you say for the rest of the relationship arrives with a question mark attached. Second, the sandwich buries the actionable content. The one sentence containing real information sits in the middle, wrapped in reassurance, and the listener walks away remembering the emotional temperature and not the squeegee.

There is a related finding worth knowing. Feedback that is vague — do a better job on bathrooms — is reliably worse than feedback that specifies the behavior, because vague feedback gives attention nowhere to go except inward. If I tell you the bathroom wasn't good enough, the only thing you can examine is yourself. If I tell you the grout line behind the toilet needs the brush, not the cloth, you have somewhere to put your hands.

What this means for a turnover

A cleaner is not an employee sitting through a quarterly review. She is a contractor with a phone, four other clients, and a strong sense of whether working for you is pleasant. She has enormous discretion over the parts of the job you can't measure — the drawer she opens or doesn't, the hair she notices on the bath mat, the text she sends you about the broken lamp or doesn't. That discretionary layer is exactly what collapses when someone's attention moves to self-protection. She'll still do the checklist. She'll stop doing the noticing.

So the goal of turnover feedback is not to be nice. It's narrower and more achievable than that: keep her attention on the surface, the object, the step — and off her standing with you.

A few mechanics follow directly.

Point at the thing, not the person. The shower door had water spots this turn is a fact about a door. You missed the shower door is a fact about her. Same information, different attentional target. Passive voice, usually a writing sin, is here a small act of mercy.

Close the loop the same day. Feedback delivered four turnovers later can't attach to a specific action, so it attaches to the person — it becomes character, not correction. Same-day, it's just the day's work.

Give feedback in the same channel and cadence as everything else. If your only messages are corrections, your name on her screen becomes a stressor. If you text her routinely — schedule, confirmation, thanks, correction — a note about the shower door is one item in a stream, not a summons.

Show, don't describe. A photo of the actual door with the actual spots removes almost all interpretive room. There is nothing to defend against in a photograph. It's just a door.

Fix the system when the system is at fault. If three cleaners have all missed the same shelf, it isn't three character flaws. It's a shelf that is invisible from the doorway. Feedback pointed at a person for a problem caused by a floorplan is the fastest way to lose a good cleaner.

Your next moves

  • Rewrite your last correction, today. Pull up the last message you sent a cleaner about a miss. Delete every softener (just, tiny, no big deal, exclamation points). Rewrite it as one factual sentence naming the object and the fix: Shower door had water spots — squeegee on the hook fixes it. Read both versions side by side. Notice which one is actually kinder.
  • Set a same-day rule and write it down. Any correction goes out within 24 hours of the turnover or it doesn't go out at all. Old feedback isn't feedback; it's a grievance.
  • Send one non-corrective message this week. Confirm a schedule, thank her for a fast turn, tell her a guest mentioned the bed. Not as a setup for anything. Just so your name is not only ever bad news.
  • Replace one vague standard with one observable one. Find a line in your checklist like bathroom spotless and turn it into something you could photograph: shower door dry and streak-free; grout line behind toilet brushed. Vague standards force self-evaluation. Observable ones invite task-evaluation.
  • Audit for the invisible shelf. List the three spots most often missed across all your cleaners. For each, ask whether it's a person problem or a sightline problem. Fix the sightline — move the item, add a light, add a photo to the checklist — before you say a word to anyone.

The quiet version of all this

The hosts who keep cleaners for years are rarely the ones with the best scripts. They're the ones whose feedback arrives so routinely, so specifically, and so close to the moment that it never has time to curdle into a verdict about a person. The correction is small because it's early. It's early because the loop is short.

That's the loop Stayput exists to shorten. Each property gets its own scheduled SMS to its own cleaner, a photo confirmation that closes the turn, and restock alerts that flag the low shampoo before a guest does — so when something's off, you're looking at a picture of a shower door two hours after checkout, not composing a difficult message four days later about whether someone is doing a good job. It's $19/month per property, and it mostly earns its keep by making the hardest conversation in this business a smaller one. If you'd like to see how it fits your turnovers, take a look at stayput.lumenlabs.works.