Location Decision Memo

What These Candidates Are: points sampled along a route line, not parcels

Nobody proposed these 19 candidates. They were sampled along the supplied route polyline, and each one's name is its milepost: the candidate reading "mile 180" is the point 180 miles along that line. It is not an address, a parcel, or a property anyone has visited.

19 points along 347.1 miles of corridor: one every 20 miles from mile 0 to mile 340, and a final point at mile 347.1 where the route line ends (7.1 miles after the one before it).

An interval point is a screening position on a corridor. It is not a parcel selection, not a site anyone has inspected, and not an entitlement finding. Ranking these points says which stretches of the corridor are worth examining. It does not say where to build, and no ranking here survives contact with a title search, a zoning check, or a site visit.

The route line is a simplified polyline. A milepost on it locates a stretch of corridor rather than a point on the roadway, and distances along it are approximate.

Run Limits: what every reading in this pack assumes, stated once instead of on each row

These conditions are not properties of any one candidate. They constrain the run that produced all 19 of them, and the comparison above assumes every one of them holds. They are listed here rather than repeated in each row because a column that says the same thing in every cell stops being read, and what a reader then misses is a caveat rather than a duplicate.

• This interval sweep re-solves the study by candidate using distance-proxy market costs and a stated unemployment rate; measured road routing, live BLS LAUS/OEWS refresh, repeated viability, payroll, robustness, and forecast-register sections are parked for the rerun-cache lift.
• Candidate was re-solved through the study engine; rank stability propagation is still a separate gate.

Ranked Locations

Why This Candidate Leads On Current Evidence

On the components these two share, Total market capture puts US-97 20-mile sweep mile 200 0.0171 ahead of US-97 20-mile sweep mile 220; Annual output puts it 0.0074 behind, Self-cannibalization share a further 0.0031, and Net new capture a further 0.0010. The Annual output gap on its own is wider than the 0.0056 margin, so the lead holds only because the differences running the other way are larger together. The Consequence column below reads each component on its own rather than against the runner-up, so a component it describes as increasing rank can still be one this candidate trails on.

Score rates this component for this candidate on a 0-to-1 scale where 1 is the best any candidate achieved and 0 the worst; it is a ranking aid calibrated against the other candidates in this comparison and against nothing outside it. Weight is relative rather than a share: the weights that carry a score total 1.3, and the candidate's score divides by that total.

What Could Change The Answer: 0.0056 of weighted score separates the top two

• A small movement in weighted components could move US-97 20-mile sweep mile 220 ahead of US-97 20-mile sweep mile 200; run correlated propagation before making this a final siting recommendation.

No resampling stands behind these conditions. They are read off the size of the margin and the components that make it, so nothing here has tested how far the evidence under this order can move, and no variable in this pack has been varied at all.

Which Components Could Flip This Order

These are the components the order between US-97 20-mile sweep mile 200 and US-97 20-mile sweep mile 220 is exposed to, least movement first. Movement is stated on the 0-to-1 component scale the tables above score on, and the margin each one has to cover is 0.0056 of weighted score.

• Net new capture. It carries 0.27 of US-97 20-mile sweep mile 220's weighted total, so 0.0208 of movement on this one component covers the whole 0.0056 margin: US-97 20-mile sweep mile 220 gaining that much, or US-97 20-mile sweep mile 200 losing it. US-97 20-mile sweep mile 220 sits at 1.0000 and the component tops out at 1, so it cannot gain that much; US-97 20-mile sweep mile 200 falling from 0.9964 does it. It rests on derived evidence, the shared market solver.
• Total market capture. It carries 0.27 of US-97 20-mile sweep mile 220's weighted total, so 0.0208 of movement on this one component covers the whole 0.0056 margin: US-97 20-mile sweep mile 220 gaining that much, or US-97 20-mile sweep mile 200 losing it. US-97 20-mile sweep mile 220 scores 0.9364 on it now and US-97 20-mile sweep mile 200 1.0000, so either move stays inside the 0-to-1 component scale. It rests on derived evidence, the shared market solver.
• Annual output. It carries 0.23 of US-97 20-mile sweep mile 220's weighted total, so 0.0243 of movement on this one component covers the whole 0.0056 margin: US-97 20-mile sweep mile 220 gaining that much, or US-97 20-mile sweep mile 200 losing it. US-97 20-mile sweep mile 220 scores 0.0639 on it now and US-97 20-mile sweep mile 200 0.0318, so either move stays inside the 0-to-1 component scale. It rests on estimated evidence, the per-candidate study re-solve.
• Study release quality. It carries 0.15 of US-97 20-mile sweep mile 220's weighted total, so 0.0364 of movement on this one component covers the whole 0.0056 margin: US-97 20-mile sweep mile 220 gaining that much, or US-97 20-mile sweep mile 200 losing it. US-97 20-mile sweep mile 220 scores 0.5000 on it now and US-97 20-mile sweep mile 200 0.5000, so either move stays inside the 0-to-1 component scale. It rests on estimated evidence, the per-candidate study re-solve.
• Self-cannibalization share. It carries 0.08 of US-97 20-mile sweep mile 220's weighted total, so 0.0728 of movement on this one component covers the whole 0.0056 margin: US-97 20-mile sweep mile 220 gaining that much, or US-97 20-mile sweep mile 200 losing it. US-97 20-mile sweep mile 220 scores 0.8720 on it now and US-97 20-mile sweep mile 200 0.8318, so either move stays inside the 0-to-1 component scale. It rests on derived evidence, the shared market solver.

The evidence tier under each of these is what the next dollar buys. They rest on derived and estimated evidence, and the Source column above says whose. Moving the components at the top of this list onto measured evidence is the work the ranking is waiting on, and it is stated as a step under What Due Diligence Comes Next.

Not Established

• Generated interval candidates are route-screening points and do not replace parcel selection, access engineering, or entitlement review.
• A provisional leader is not a site recommendation unless rank stability is measured and established.
• Correlated candidate propagation has not been run; ranking is a deterministic comparison of the supplied or solved components.
• The ranking of these candidates is not established. It was not resampled, and the order it reports is one deterministic run. Which Components Could Flip This Order names the components it turns on and how far each would have to move.
• No win probability is established. Nothing in this run estimates how often each candidate would come first if the inputs were resampled.

What This Pack Can Still Be Used For: every candidate reading, and not the Score column

Use this comparison to narrow the field, not to make the final selection. That is a job it can do, and it is the job it was run for: the candidate evidence cleared the release gate, so this pack may be circulated, a board may carry the top of this order forward as the live shortlist, and diligence money may be committed against it. Which Components Could Flip This Order names where that money goes first and how far each component would have to move. What this pack may not do is name the site that gets built. The order among the candidates at the top has not been tested against the evidence bounds at all, so nothing here says whether it would come out the same way twice.

Each candidate's component values, units, evidence tiers and sources are readings of that site, and they hold whatever the order turns out to be; so do the conditions and blockers recorded against a candidate. The Score column does not travel with them: it is calibrated against the other candidates in this comparison, so it moves when the ranking moves, and it is waiting on the same evidence the recommendation is.

What Would Make This A Recommendation: the order has not been shown to hold

These are the gates this pack has not cleared, read from what the run recorded rather than from the prose around it.

• Rank stability is not established. This comparison was not resampled, and the order it reports is one deterministic run.
• 2 conditions on this comparison apply to every candidate rather than to US-97 20-mile sweep mile 200, and are stated under Run Limits. The comparison assumes them all.

What would open them: each of these is a different purchase

• Reduce the uncertainty on Net new capture and Total market capture. Nothing else in this comparison takes less movement to change the order: 0.0208 on Net new capture alone covers the whole 0.0056 margin, on the 0-to-1 component scale the tables above score on.

What Due Diligence Comes Next: the ranking first, then what each interval point was told to confirm

These are the steps this pack records, and it records no others: what the engine was told about each interval point, and the work the ranking itself is waiting on. The pack does not scope a diligence program and nothing here is a substitute for one. US-97 20-mile sweep mile 220's conditions are here beside US-97 20-mile sweep mile 200's because the ranking is not established and 0.0056 on the same 0-to-1 scale separates them, so neither site has been ruled out.

• Establish rank stability. This comparison was not resampled, and the order it reports is one deterministic run, and that run is what stands between a provisional leader and a site recommendation.

What Evidence To Buy Next: 2 purchases, each answering something this pack recorded

What is safe to decide now. Use this comparison to narrow the field, not to make the final selection. That is a job it can do, and it is the job it was run for: the candidate evidence cleared the release gate, so this pack may be circulated, a board may carry the top of this order forward as the live shortlist, and diligence money may be committed against it. Which Components Could Flip This Order names where that money goes first and how far each component would have to move. What this pack may not do is name the site that gets built. The order among the candidates at the top has not been tested against the evidence bounds at all, so nothing here says whether it would come out the same way twice.

What is not safe to decide now. Which interval point to build on. That question stays open until the gates under What Would Make This A Recommendation clear, and no purchase below closes it on its own.

• Traffic or travel-time study: buy measured drive times on the corridor, in place of the straight-line distance proxy the run substituted. It bears on the market-access and capture readings, which the proxy currently sets. It can move a component reading, and through it the order. Asked for by a run limit, which reads: “This interval sweep re-solves the study by candidate using distance-proxy market costs and a stated unemployment rate; measured road routing, live BLS LAUS/OEWS refresh, repeated…”
• Wage and labour refresh: buy a live BLS LAUS and OEWS pull, in place of the stated rate the run was given. It bears on the employment and payroll readings, which a stated rate currently fixes. It can move a component reading, and through it the order. Asked for by a run limit, which reads: “This interval sweep re-solves the study by candidate using distance-proxy market costs and a stated unemployment rate; measured road routing, live BLS LAUS/OEWS refresh, repeated…”

What would release the recommendation. Not any single item above. Between them these 2 purchases can move a component reading, and through it the order - and the recommendation gate does not read any of those directly. It reads whether a resampling of the component evidence returns the same order often enough. Better inputs give that test a better chance of settling; they cannot settle it, and a pack that implied otherwise would be selling a gate rather than the work.

What cannot be promised. None of these purchases is offered as a way to reach a particular answer. Each one narrows a stated uncertainty; what it does to the order is not knowable before it is done, and it may confirm the current ranking, reverse it, or leave the margin too close to call. Buying all 2 would still not make a site recommendation safe unless the rank-stability gate clears on the evidence they produce.