The two terms travel together and get used interchangeably, which is a shame, because they answer different questions, and a model that runs one while believing it ran the other produces confidence with nothing underneath it. The distinction takes one paragraph and pays for itself on every deal after.
Sensitivity analysis moves one assumption at a time and watches the output respond. Its product is a ranking: which inputs is this deal most exposed to, per unit of movement. Rent growth up and down a point, exit cap up and down half, downtime longer and shorter, one at a time, everything else held still. What you learn is where the model's leverage lives, and the honest surprise is how often it lives somewhere other than where the debate has been. Teams argue for an hour about year-one rent bumps while the exit cap, moving the outcome five times as hard, sits unexamined. Sensitivity is how you find that out and reallocate the argument.
Stress testing moves several assumptions together, deliberately, in the direction the world actually moves them. Its product is a verdict: does the deal survive the plausible bad year. And the word together is the entire point, because adverse conditions arrive correlated. The environment that flattens rent growth also stretches downtime, widens the exit cap, and tightens the refinance, and a deal that survived each of those individually in the sensitivity pass can fail their arrival as a group. One-variable analysis structurally understates joint risk. That is not a flaw in sensitivity; it is the boundary of what the technique measures, and stress testing exists on the other side of the boundary.
The craft is running them in the right order, because each aims the other. Sensitivity first, to identify the three or four assumptions this specific deal actually pivots on. Then build the stress scenarios out of those, moved jointly, at plausible severities: the mild recession, the credit crunch, the slow-motion version where nothing breaks but everything drags. Stressing assumptions the sensitivity pass showed to be minor is theater. Stressing the levers, together, is the test.
And both must run against the real model to mean anything. Sensitivity on a blended-average pro forma can only tell you about the averages, while the deal's true exposures live in the lease schedule underneath them, which specific tenant, which expiration year, which suite. PropCalc™ runs the stress grid against the full per-tenant model, growth against exit against downtime, break points visible, so both questions get asked of the actual deal rather than a summary of it. Which assumptions matter, and does it survive. Two questions. Ask both, in that order.