The counterfactual toggle
Letting the user ask 'what if that input changed?'
The model gives you a confident answer and a small pile of assumptions that underlie it. What happens to the answer if one of those assumptions flips? Most products can't tell you. The user has to re-prompt, re-ask, re-read. A counterfactual toggle does it in one tap.
The pattern is interactive sensitivity analysis. Every assumption shows up as a chip. Flipping the chip recomputes the answer inline. The user can lock the assumptions they trust and focus on the ones they don't.
"Confidence without sensitivity is performance. An answer that can't tell you how it would change isn't an answer, it's a press release."
Assumptions as chips. Deltas as live updates.
Each assumption lives in a left panel as a card: baseline value on one side, alternative on the other, with the expected percent delta if flipped. A lock icon freezes the assumption. The right panel shows the live answer, updating as the user toggles. Totals summarize the compound delta, direction and magnitude.
All assumptions at baseline. Flip any chip on the left to see the projection change.
Flip an assumption; watch the answer update. Lock the ones you trust so you can focus on the ones you don't.
Sensitivity is how experts read answers.
The difference between a number and a useful number is knowing what it's sensitive to. A finance team that pulls a model's projection wants to know: if churn rises, does this fall apart? If hiring slips, does this still hold? A counterfactual toggle exposes that sensitivity as an interaction, so the user develops intuition instead of just reading a conclusion.
Toggles that earn their place.
- Name the assumptions. Don't hide them inside the answer. Break them out so they can be flipped one at a time.
- Show deltas before flipping. The user should know what a flip would cost before they commit. This encourages real exploration, not random toggling.
- Support locks. Trust in some assumptions; probe the others. A lock icon focuses the analysis without hiding the alternative.
Toggles with fake math.
The dangerous failure mode is a toggle whose delta is an educated guess. The user flips churn to 'bear case' and sees -8.6%, but the -8.6% is a heuristic, not a recomputation. The user now has a false sense of sensitivity, which is worse than having none.
What this pattern gets wrong when it gets wrong.
- Confidence theater
- Language or typography that performs certainty beyond what the model actually has.
- Unverified claim
- A figure or fact presented without provenance, in a place where the reader will treat it as cited.
Three shipping variants worth copying.
- Inline chips for each assumption with click-to-flip
- Delta preview fades in as the toggle moves
- A 'lock this assumption' option for the curious-but-busy