What rate do you put in the refinance line when you're making the offer?

Before you pick a number, look at what the exit lender's commitment actually promises. Most of what I read commits to a program and an LTV cap subject to appraisal and coverage at the rate in effect at lock, which means the rate line in your model is not a forecast of the market, it's a forecast of which side of their coverage test you land on. Some programs also step the max LTV down as coverage falls, so you don't get a clean 75 at 1.05, you get 70 or 65. Ask for the coverage-to-LTV grid in writing. That grid tells you exactly how much rate movement you can absorb before the cash-out shrinks, and then you're stress testing a documented threshold instead of guessing.

Fathom's grid is the answer to your question, because it converts the rate cell into a break-even. You don't need to predict 8.25 versus 9, you need to know the rate at which the deal stops giving cash back and then decide whether you'd still own it there. If the answer is yes, plug today's rate and note the break-even next to it. If the answer is no, you're not underwriting a rate, you're underwriting a bet on rates, and the client should be told that in those words.

I plugged today's rate on a small deal and closed the refinance eleven months later at a rate that was nowhere near it, so...

Sitting on the lending side of this, the thing I'd flag is that coverage gets tested on the appraiser's market rent and the lender's expense assumptions, not on your actual rent roll or your actual expenses. I've seen underwriters haircut rent to appraised market and add a management factor even when the borrower self-manages, and a deal that showed 1.11 on the borrower's numbers came in under 1.05 on ours. If your cushion is only on rate, you've left the other half uncovered. Confirm which rent figure and which expense load the program uses before you set the cell.

As someone still building a list rather than buying, the third approach is the only one I can actually execute, because I have no way to judge a rate forecast. If a deal only works with cash back, I'd rather find that out before I own it. It does mean I've passed on things that probably worked. I've made peace with that.

Sextant, the middle option isn't unprincipled if you tie the cushion to a documented threshold rather than a round number. Take fathom's grid, find the rate where max LTV steps down, and use that as your test rate. Now the 75 or 150 basis points isn't a vibe, it's the distance to the cliff. What I'd add is that the cushion belongs on rent too. Tender's point about appraised market rent versus your rent roll has cost me more dollars than rate movement ever has, and I keep two rows for it now.