Ring sizes.
Clients loved her when she decided they belonged.
That was important.
She was not rude indiscriminately.
She was selective.
Her definition of promising customer heavily influenced the entire floor.
Those selected received extraordinary service.
Those not selected received efficiency at best and humiliation at worst.
The store’s sales growth was real.
But it came partly from concentrating labor on high-probability buyers while treating uncertain customers as distractions.
Economically, one could understand the instinct.
Luxury salespeople have limited time.
High-ticket clients matter.
The problem was that human beings, not objective models, decided who looked likely.
And humans smuggled assumptions into probability.
Alicia asked Gabriel to quantify.
He resisted.
“We can’t infer protected characteristics reliably from names.”
“Don’t.”
“Use complaints where self-identified.”
“Use camera sample under counsel.”
“Use observable service steps without trying to label race where uncertain.”
“What steps?”
“Greeting time.”
“Beverage offer.”
“Time before case access.”
“Request for budget.”
“Request for ID.”
“Manager escalation.”
“Security involvement.”
“Transaction outcome.”
Gabriel nodded.
“That is more defensible.”
“Do it.”
They sampled three hundred interactions from six months of security footage.
Not facial recognition.
Human-coded under legal protocol.
Results:
Clients arriving by appointment received excellent consistency.
Walk-ins showed wide variation.
Customers wearing conspicuous luxury markers received faster case access.
Older customers were offered seating more often.
Younger customers were more frequently asked for budget before high-value items were shown.
Black walk-in customers in the sample waited longer on average for initial assistance than white walk-ins.
The sample was not large enough for sweeping causal conclusions.
But difference persisted even after controlling for store traffic in their preliminary analysis.
More telling was manager intervention.
Vanessa appeared far more often in interactions involving Black clients who asked to see items above one hundred thousand dollars than she did for comparable white clients.
Sometimes intervention improved service.
Sometimes it constrained.
The pattern needed explanation.
Alicia read report twice.
Then set it aside.
She felt no satisfaction.
Evidence rarely feels triumphant when it confirms something ugly.
On Friday, Vanessa appeared for interview with attorney.
Alicia did not attend.
She watched recording later.
Vanessa was defensive initially.
Then less.
She denied consciously discriminating by race.
No evidence contradicted consciousness directly.
She admitted she profiled by perceived purchase likelihood.
Her attorney objected to term profiled.
Vanessa said:
“That’s what I did.”
Alicia paused video.
Rewound.
Vanessa continued:
“In luxury sales, you have to decide where to spend time.”
“People who say otherwise are lying.”
HR asked:
“How did you decide?”
“Experience.”
“What signals?”
“Appointment.”
“Presentation.”
“Behavior.”
“Knowledge.”
“Confidence.”
“Who they came with.”
“Sometimes what they drove.”
“Sometimes whether another client referred.”
“Did race matter?”
Vanessa sat quietly for eleven seconds.
Then:
“I never told myself it did.”
HR waited.
Vanessa looked down.
“I can’t say it never did.”
Alicia stopped video again.
That answer mattered.
Not because confession solved.