What an outcome band means, and why we hold memos instead of guessing

Two Indiana data-center sites, 80 miles apart, in the same state, in the same political climate. Our case files score one of them 90 out of 100 and the other 2.

The 90 is Meta's Lebanon campus, inside a purpose-built district where a 2023 council vote made the use by-right. The 2 is Google's Franklin Township petition, withdrawn in September 2025 before a hostile council vote it was not going to survive. Nobody needs decimals to act on that spread. But almost nothing a screening team looks at is a 90-versus-2 spread, and this post is about what we do with everything in between: why RealClear leads its site memos with an outcome band rather than a score, and why a memo that cannot earn its band gets held rather than shipped.

What is an outcome band?

An outcome band is the recommendation register a site memo leads with. Four positions: Advance, Monitor, Hold, Kill. Advance means the record supports moving into deeper diligence. Monitor means the record is live and the posture is watchful. Hold means do not commit further budget until named questions resolve. Kill means the record already contains the ending.

The number still exists underneath, but the doctrine in the product is explicit: read the band, not the number. The band is the hero of the memo; the score is a subordinate detail. That is a design decision about honesty, and it is worth explaining, because it cuts against how most analytics products behave.

Why lead with a band instead of a score?

Because a band matches the actual precision of the evidence, and a naked score lies about it.

A score of 59 implies the instrument can tell 59 from 62. It cannot, and neither can anyone else, because entitlement risk is not a continuous physical quantity. It is a read on documents, votes, procedures, and people. What the record can actually support is a coarser and more useful statement: this site's risks are structural and priced, or live and unresolved, or terminal. Pretending to single-point precision is exactly the kind of confident overstatement the rest of our methodology exists to prevent.

Our Saline Township file is the working example. It carries a 59, and the number alone would tell you almost nothing. Is 59 good? The band-level read is what a development team can use: this was a litigation-and-settlement approval path, not a clean rezoning win. The township denied 4-1, the developer and landowners sued, and a consent judgment authorized the project subject to roughly $14 million in commitments and detailed operating restrictions, with a state power case still contested months after the land question closed. The honest register for that record is watchful, with the reasons named. A team that read "59" and binned it as "middling, proceed" would have missed everything that mattered; the file exists precisely because the path was messy in ways the number cannot carry.

There is a second reason, and it is about incentives. A score invites false comparison across memos ("take the 61 over the 58"), which manufactures precision out of noise. A band forces the conversation back to the record: what would move this site from Monitor to Advance? That question has a documentable answer. "What gets it from 58 to 61?" does not.

What makes a band move?

Evidence, on the record, with dates. The Franklin Township file shows a read moving in real time: at filing in March 2025, on structural facts alone (a discretionary rezoning, flood overlays written into the parcels' designations, a council call-down as a second veto point), the site read at 25 out of 100. Elevated risk, but no organized opposition yet visible. By August 2025, with formal remonstrance filed, a resident coalition organized, and the district councilor working the call-down, the read had collapsed to 2. The band moved because the public record moved. Five months, same parcel, different answer.

That is not the instrument being fickle. That is the instrument doing its job. A site read is a photograph of a record that is still being written, which is why memos carry dates and why monitoring exists. Anyone who hands you an entitlement score that never changes is handing you a horoscope.

This is also why a delivered memo is not the end of the engagement. The sources a memo actually relied on (the specific dockets, agendas, and filings behind its claims) become the watch list, so that when the record moves, the band conversation can reopen with the new document in hand rather than a vague sense that something changed. Monitor is a posture, not a shrug.

Why hold a memo instead of shipping a hedged one?

Sometimes the record cannot support any band at all. The jurisdiction is thin in the public sources, the key documents are not yet retrievable, the claims that matter cannot be tied to evidence. The tempting move, the move most generated research makes, is to ship anyway: write fluent prose with hedges spread through it like insulation, so the formatting implies a confidence the sourcing never had.

We hold the memo instead.

This is the honest-hold doctrine, and we have written about it as a review-gate design before: every memo passes deterministic checks and an automated review before it can publish, and the gate's answer is genuinely allowed to be no. The band framing explains why that gate has to exist. A band is a claim about the record. Advance without evidence is not a cautious guess; it is a fabricated claim wearing a label. If the evidence cannot earn a band, there is no such thing as shipping the memo "with caveats." There is only shipping a hollow one.

So the deliverable contract runs one way. Either the memo ships with a band the cited record can defend, claim by claim, or it does not ship and we say why: here is what the record lacks, here is what would resolve it. For a data-center team deciding where the next site visit goes, "we cannot band this jurisdiction yet, and here are the three missing documents" is more useful than a polished maybe. It is also rarer, because it costs revenue to say and most tools are built never to say it.

The test of the whole design is simple to state. If our bands are honest, then over time sites we band Advance should rarely die at the podium, sites we band Kill should rarely get built as proposed, and the holds should convert to bands once the named evidence arrives. That ledger is checkable against outcomes in the public record, and I expect it to hold up; the Franklin Township and Lebanon files are early entries. If it stops holding, the methodology is wrong and we will change it in public.

Ask any research vendor you use the same question. What would make you hold the report? If the answer is nothing, you already know how much the confident ones are worth.

A RealClear site memo is a source-cited research summary, not legal advice. It can contain errors and should be verified independently before any investment decision.