Bias Cascade

How small biases at each stage compound into large disparities

A 5% bias doesn't sound like much. But when that 5% appears at every stage of a multi-stage process, the effects compound dramatically.

Adjust the bias at each stage of this hiring funnel and watch how Group A and Group B diverge.

Starting Population

500
500

Both groups are equally qualified. Only the bias differs.

Bias at Each Stage

Positive = favors Group A, Negative = favors Group B

+5%
+5%
+5%
+5%

The Funnel

Group A
Group B
Started Equal?
Yes
Group A Hired
0
0%
Group B Hired
0
0%

Compound Effect

Individual stage biases:

Combined effect: disparity in outcomes

The math of compounding bias

Why this matters

Each evaluator might genuinely believe they're being fair. A 5% bias is nearly imperceptible at the individual level—it's the difference between a "maybe" becoming a "yes" or a "no" in ambiguous cases.

But systems have many stages. And biases compound multiplicatively, not additively. Four stages of 5% bias each don't create 20% disparity—they create much more.

This is why looking at outcomes matters. If your final results show large disparities despite equal starting populations, the bias is hiding somewhere in your process—even if no single stage looks problematic.