A mortgage recast looks like a simple arithmetic problem, which is exactly why the calculators for it tend to be wrong.
Here is the mechanic. You owe a balance. You pay a lump sum toward principal. The servicer re-amortizes the remaining balance over the remaining term at the same rate. Your payment drops. Your rate does not change, your payoff date does not change, and no new loan is created — which is what separates it from a refinance.
Balance − lump sum, spread over months left. Two lines of arithmetic.
The interesting decisions are all in what you choose to put on the screen.
The fee is not the cost
The servicer charges something — typically $150 to $500, occasionally nothing, almost never more than $1,000. That range is one to two orders of magnitude below refinance closing costs, and it is the reason the product exists at all.
But modelling the fee as "the cost" produces a break-even number that is too optimistic, because it ignores the second cost: the money you sent sits in the servicer's process, not in your mortgage. Interest accrues on the old balance while the new payment is being set up.
On a $50,000 lump sum at 6.5%, that gap runs about $8.90 a day. Four weeks of processing is roughly $250 — comparable to the fee itself, and invisible if your model only counts the fee.
That single line changes the answer for anyone whose break-even is measured in months rather than years.
Break-even is a range, not a date
The temptation is to print "pays for itself in 1 month" and be done. The problem is that the three inputs to that sentence are not equally knowable:
- the monthly saving is arithmetic, and exact;
- the fee is a published number that varies by servicer and sometimes by promotion;
- the processing time is not published, varies by servicer, and moves with their backlog.
So the honest output is a range: fee divided by saving gives the best case, (fee + interest during processing) divided by saving gives the likely case, and the gap between them is the part you cannot control.
That is less satisfying to look at than a single date. It is also the only version that stays true when the reader calls a different servicer.
Where precision becomes dishonest
Financial calculators have a failure mode that software tools usually avoid: they can be precise about things that are not knowable, and the precision reads as authority.
Two examples from this problem space.
Publishing a fee table per servicer is a maintenance trap. Servicer policies change without notice, and a table that was right in March is confidently wrong in October. The rule I use: cite the servicer's own document, date the row, and show nothing for servicers where the document could not be found — rather than interpolating from a neighbouring servicer.
An empty cell is a feature. If a servicer does not publish whether they allow recasts at all, the row stays blank. A blank invites the reader to call; a guess invites them to plan around a number that was invented.
The part most calculators skip: eligibility
The arithmetic is the easy half. The half that decides whether the number means anything is eligibility, and it is not smooth:
- conventional loans backed by Fannie Mae or Freddie Mac generally can recast;
- FHA, VA and USDA loans generally cannot, because their servicing rules have no voluntary re-amortization path;
- jumbo and portfolio loans are at the lender's discretion;
- most servicers require the loan to be current, with some payment history;
- there is usually a minimum lump sum, commonly $5,000–$10,000 or a percentage of the balance.
A calculator that runs the arithmetic on an FHA loan is not giving an estimate. It is giving a wrong answer to a question the user cannot ask.
What this changes about the tool
The output that survived all of the above is deliberately modest: a new payment, a monthly saving, a fee, an interest-during-processing line, and a break-even range — plus a clear statement of which loans can and cannot use it.
No annual-savings projection over thirty years, because the payoff date does not move and the projection would flatter the decision. No "you will save $X" headline, because the honest version of that sentence depends on inputs the tool cannot see.
A calculator's job is to make the unknown visible, not to fill it in.
The worked version of this — with the fee, the processing-interest line and the eligibility rules on the same page — is at recastdesk.net.
Top comments (0)