The occupant paid the taxes two days before the redemption deadline and I watched a deal unwind in real time

This is one of the clearest illustrations of the gap between certificate rate and actual portfolio yield I have seen described in this forum, so thank you for putting the numbers to it.

For anyone reading who is new to the strategy: a tax lien certificate comes with a statutory interest rate, meaning a rate set by state law, say 18% per year. That rate applies to the money you have out on that specific certificate. What it does not tell you is what rate you will earn on your total capital across the full year, because capital that sits idle between an early redemption and the next auction earns nothing. The annualized portfolio return is a different calculation, and as this example shows, it can land well below the certificate rate.

The reinvestment problem your contact hit is structural in single-county, single-cycle approaches. When most of your certificates sit in one jurisdiction and one auction window, early redemptions cluster, and the over-the-counter inventory, meaning liens available outside the main auction, is often thin and picked over by the time you need it.

The thing worth adding: the model needs a holding assumption and a reinvestment assumption entered separately, before bidding, not after. If you cannot name a plausible home for the capital in months two through twelve, that gap is already part of your return, and it belongs in the projection.

Multi-county or multi-cycle positioning addresses some of this, though it adds its own procedural weight, and I would not say it solves the problem cleanly.

What does your contact's county selection process look like right now, and are they working across more than one auction cycle?

The reinvestment gap you are describing has a structural fix that most single-county buyers do not build in before the first auction: running certificates across multiple counties in the same state, staggered by auction date. If county A runs in March and county B runs in September, early redemptions from March positions have a natural home six months later, and the capital does not sit earning nothing while you wait for over-the-counter scraps. The tradeoff is due diligence overhead per county and the fact that statutory rates, redemption penalties, and the order of premium bidding can differ meaningfully even within the same state. Your contact's model assumed one reinvestment environment when it should have priced in the possibility of zero inventory on the day the money arrives.

The part that connects to a different structure here is self-directed IRA positioning in these certificates. When the capital comes back inside a tax-advantaged account, the reinvestment drag still exists, but it compounds quietly against you for a different reason: idle cash in an SDIRA earns nothing and does not get a tax deduction for the opportunity cost. Some custodians allow the account to hold short-duration notes between auction cycles as a bridge, which at least puts the returned capital to work at some rate while the next auction opens. The question worth asking before that first bid, in either structure, is whether you have identified the next county's auction date, confirmed the OTC inventory with the county treasurer's office directly, and know on day one of redemption exactly where the cash goes on day two.

The two-year window priced as if it will run two years is the part that gets people. I had something adjacent with a note funded through my self-directed Roth, custodian moved slow on the front end, borrower paid off fast on the back end, and the actual deployed period was maybe five months out of twelve, so the annualized yield I was telling myself looked embarrassing by December. Running eight positions in one county is basically betting the same redemption-timing guess eight times with correlated outcomes, which is concentration risk that does not show up in the rate. What I want to know before bidding in a setup like that is whether a secondary market for certificates exists in that state, because if you can sell at a small discount the day redemption risk spikes, you control your own exit instead of waiting on an owner who may or may not cooperate with your spreadsheet.