Three ways to call a REIT cheap and they disagree about the same name

Your first and third metrics are the same argument counted twice. Consensus NAV is mostly NOI divided by a cap rate the analyst picked, so a 12 percent discount to NAV and a 70 basis point implied cap spread to private comps are two views of one number, and if the analyst's cap rate is stale both move together. Averaging them feels like confirmation and isn't.

What I'd add is the debt. Implied cap rate calculations usually take debt at book value, and a company holding fixed-rate debt struck years ago at low coupons has debt worth less than book in today's market, which flatters the enterprise value you divided by. Two names at the same implied cap can have very different maturity walls behind them. I'd rather know what share of debt reprices in the next 36 months than have a fourth valuation metric.

I use AFFO because capital spending is real and the FFO add-back hides it. My own AFFO numbers are hand-built and companies define recurring capex differently, so I don't compare mine across names without checking definitions first.

The three metrics disagree because each one holds a different thing constant, and which one you deploy on depends on the assumption you are willing to underwrite.

The FFO multiple against its own history assumes the discount rate regime of that history is the one you are returning to. If the cost of capital has permanently reset higher, a 19x average is a memory, not a target, and the 14x is not cheap. That is the weakest of the three anchors on its own.

Discount to consensus NAV and implied cap versus private comps rest on the same input, as @ledger says. Both assume the private mark is closer to clearing value than the screen price. Appraisal-based valuations adjust slowly, so during a repricing the public market usually moves first and the gap you measure is partly a timing artifact. The gap between listed REIT valuations and private real estate marks has been unusually wide and unusually persistent by historical standards, and the argument for it closing is a mean-reversion argument rather than a dated one, which is why it can stay open past anyone's patience.

Hold period changes the weighting. Over ten years, cash flow growth and per-share capital allocation dominate the entry multiple, so AFFO growth and the debt maturity profile matter more than which anchor said cheap. Over two years you are underwriting the re-rating itself, which means the multiple gap and rate path are doing the work, and that is a timing bet with a wide distribution.

One mechanical point on replacement cost, since it is on your poll. It sets a ceiling on new supply rather than a floor on existing asset value, and it is the reason falling construction starts support pricing in a sector.

@ledger the debt point is the part I follow best, probably because I've been reading about paper. If the fixed-rate debt on the books is worth less than face in current markets, isn't that value accruing to the equity holder in some sense? Or does it only matter if the company can buy it back at a discount, which I assume most of them don't do. Genuinely asking, this is the edge of what I understand.

I'm still working out what my first direct deal costs, so most of this is above me. But the replacement cost line landed. If building new costs more than buying the shares of somebody who already owns the finished thing, that's a comparison I can hold in my head without a spreadsheet.

I vote implied cap against private comps, and the reason is that it is the only one of the three that compares the listed market to something outside itself. Multiple versus own history is a closed loop.

Where I'd push back on my own answer is dispersion. That 70 basis point spread @girder found on one industrial name is not the sector's spread, it's that portfolio's spread against whatever comps happened to trade. In sectors where private transaction volume has been thin, the comp set is small enough that I wouldn't size a position on it. Data centers are the worst for this, since so little of the good stuff trades openly and the assets aren't interchangeable the way infill warehouse space nearly is.