# Understanding gender by name segmentation

How we match shopper first names to gender, and how it helps personalize your PDQ setup

May 28, 2025

## How we segment by gender

1. **We take the shopper’s first name**  
   This comes directly from the checkout field, no extra input needed.

2. **We clean the name before matching**  
   Names are lower‑cased, stripped of accents, and transliterated (so “Élise” becomes “elise”).

3. **We match it to our reference table**  
   Names are checked against a public dataset of 200,000+ first names. If we find a confident match, we assign:
   - **Male**  
   - **Female**  
   - If no match → **Unknown**

* * *

## A quick example

A shopper named “Noa” checks out.

→ We clean the name to `noa`
→ Match it to the dataset
→ Assign **Female** (based on majority labeling in the data)

* * *

## What you can expect

- On average, **97–98% of names are recognized**
- Gender-neutral names (e.g., “Alex”, “Jamie”) may be misclassified
- If a name is too rare (<10 matches in the dataset), it defaults to **Unknown**

* * *

## Continuous improvements

Every night, we:

- Check for names that showed up as “Unknown”
- Flag high-frequency names
- Add them (with the correct gender) to our reference table. That means your data keeps getting more accurate, automatically.

How accurate is the gender data? In our tests, it’s **~95–98% accurate**. Most mistakes are due to gender-neutral names.

* * *

That’s it! Your PDQ gender segmentation is working behind the scenes — helping you personalize more intelligently, with zero setup required.