DET — Living Roadmap¶
What this page is — strategic memory for DET
The single source of truth for DET: research → strategy → expectations → return → report. DET is the group's captive transport arm — it moves auction buys and equipment for the other entities and fills spare capacity with outside loads, billing at arm's-length rates. Every fire triages against this — direction that persists, not tasks that scatter.
Markers: 🟢 Confirmed · 🟡 Assumption · 🔴 Gap · 🎯 Target.
Group thesis — DET is the reliability floor under the salvage engine. ATG buys salvage trucks at auction that third-party haulers handle poorly (refuse difficult loads, no-show, damage non-running units). DET being captive — ATG's driver, ATG's rig, ATG's schedule — removes that friction, and the outside/backhaul freight that fills the gaps subsidizes the driver's wage so the captive capability is cheap. It is vertical integration to solve a labor-and-reliability problem, not a per-mile freight company 🟢. Narrative depth → entity page.
0 · MARKET RESEARCH & ANALYSIS — the why (as of 2026-06)¶
- Two service modes — do not conflate 🟢: drive-away (truck runs; driver flies out, drives it back loaded — one-way miles, no deadhead) and trailer haul (truck flatbedded on the Kenworth + Ledwell step-deck — round trip, the empty outbound leg is baked into the rate).
- The problem DET solves 🟢 — third-party haulers refuse difficult salvage units, no-show, lack the knowledge to load a non-running truck safely, and charge premiums for awkward loads. Each is a real cost a per-mile rate never captures: a truck that can't be transported can't be dismantled or rebuilt; a no-show is a delayed build and capital idling at an auction yard accruing storage.
- DET's three value buckets 🟢 — (1) avoided third-party cost (what ATG would pay haulers, done at internal cost); (2) avoided friction cost (delays, no-shows, damage — the "princess drivers in flip flops" cost, hard to quantify but real); (3) optionality (DET can be summoned on demand to hit an auction window; third-party can't). A per-mile model only addresses bucket 1.
- Market rate context 🟢 — DET's rates sit inside Southwest regional carrier ranges: drive-away $3.00–$4.50/mi (United Road, DAS benchmarks); trailer haul $4.50–$7.00/mi loaded (flatbed/step-deck). This is the §482 arm's-length support for intercompany billing.
- Seasonality 🟡 — DET demand is pulled by the salvage pipeline (auction wins) and the group's build cadence, so it tracks water-truck seasonality (peak May–Sep) at one remove. Validate once a full year of haul data lands.
- The outside-freight thesis 🟡 — Phoenix is a strong flatbed market; the model assumes backhaul fills (on salvage/haul trips) and dead-day local freight can keep the driver working 3–4 idle days a week and largely cover his base wage. This is a hypothesis to validate at scale, not a proven figure — degree of coverage depends on Phoenix freight conditions, Ledwell load compatibility, and how aggressively dispatch can schedule while preserving salvage priority.
- 🔴 Gaps: addressable Phoenix flatbed/equipment-freight $ market; whether the LA-port empty-container lane (modeled ~$58K/yr single-lane gross margin at 100 hauls, spot-mid $3.75/mi) is worth activating vs. higher-value loaded backhauls — empty-container repositioning is a low-margin lane; make-vs-buy threshold (DET cost/mi vs. third-party rate) — the headline economic case, still unproven.
1 · CHARTER / INTENT — V2MOM (the how)¶
- Vision 🟢 — be the reliability floor under the salvage engine: move ATG's auction buys and equipment more reliably and cheaply than third-party haulers, and turn otherwise-empty miles into incremental margin with outside/backhaul freight.
- Values 🟢 — salvage moves are absolute priority (no freight commitment may compromise one); HOS legality is absolute; driver retention (home-based, not burned out) is the value proposition, not an afterthought; every haul carries ≥30% target margin or it escalates.
- Customer 🟢 — the primary (captive) customer is the dealership STE, which pays arm's-length rates for auction-to-yard and equipment moves. Secondary: outside fill-freight (backhaul + dead-day) that fits the trailer and lane mix.
- Positioning 🟢 — "captive demand floor + fill-freight upside": STE captive volume guarantees a utilization baseline; any outside backhaul revenue is incremental margin on top.
- Revenue streams 🟢 — intercompany STE hauls (drive-away $3.50/mi, trailer haul $5.00/mi, Rate Schedule v1.0 eff. 2026-06-11, Brandon-authorized) · outside fill-freight (🟡 modeled, not yet booked).
- Obstacles / methods 🎯 — hold ≥30% margin per haul; stand up the AI dispatch agent to optimize the single truck's calendar (salvage priority → backhaul fills → dead-day fills) at a cost a human dispatcher for one truck could never justify; resolve the activation blockers (insurance, DOT authority, IFTA, ELD, HOS strategy) before any outside commercial mileage.
🔴 Activation blockers — non-negotiable before outside freight
DET bills STE captive hauls today, but running outside commercial freight requires items not yet confirmed: commercial auto liability + cargo + physical-damage insurance (brokers typically require $1M auto / $100K cargo — until in place, ANY commercial haul is catastrophic personal exposure and the liability-isolation structure does not protect against operating uninsured), DOT authority (USDOT/MC/BOC-3), IFTA, ELD, driver-qualification verification, and an HOS strategy for trips exceeding a single-driver window. Owner: Brandon. (🟢 that these are required; 🔴 on current status of each.)
2 · OPERATING MODEL & GO-TO-MARKET — the how it earns¶
- The rig 🟢 — one owned tractor (2015 Kenworth T680,
15D52TR) + a Ledwell step-deck (intercompany lease from STE, $1,200/mo). The Kenworth is a fixed asset — never a job class on an invoice. - Job costing 🟢 — QB class = the transported truck's stock number, giving per-truck job P&L straight from the class report. Stock numbers verified against the STE inventory workbook before posting — the authority is the stock number, never the class code or description.
- Rig-pool overhead allocation 🟢 — Insurance and Equipment Lease are rig costs, allocated each
period to the jobs that physically used the rig, revenue-weighted (a proxy until per-job mileage/hours
are tracked). Tolls are direct-classified at source; interest/bank/payroll-tax stay unclassified G&A.
(AM-adopted 2026-06-24; script
am_det_alloc_truckin.py.) - The AI dispatch agent 🟡 — the scheduling problem (salvage priority + proactive backhaul capture at moment-of-auction-win + dead-day fills, all under hard HOS and home-basing constraints) is too continuous for a human dispatcher on a single truck but ideal for a substrate agent. Structural edge: DET finds backhauls proactively at auction-win with days of lead time, where traditional carriers scramble reactively the morning the truck is empty. Designed, not yet stood up.
- Driver-livelihood constraint 🟡 — salvage moves alone don't fill a full-time calendar; the fill freight sustains the driver's income and therefore the retention that IS the value proposition. Likely structure is a guaranteed base wage + freight upside (trades some DET margin for retention).
- Single-driver fragility 🟡 — one driver (James, documented; employment arrangement, income needs, and return-at-activation all 🔴 unconfirmed) is one point of failure; a second driver is a future decision gated on salvage volume + freight utilization.
- Intercompany pricing 🟢 — §482 arm's-length; commercial-carrier comparables should be captured per haul (broker/carrier quote screenshot at time of service) for related-party tax defense.
3 · TARGETS — LEAD / LAG, RAG (the contract)¶
KPIs are defined once in the Metrics Glossary (DET/transport table) —
cited here, never redefined. RAG per the glossary's standard thresholds;
cost-per-mile, OR and deadhead % are "lower is better" (Green = at/under target). Where no real number
is in the sources, the target is 🔴 set and the actual — (pipeline pending #55) — never fabricated.
| KPI (→ glossary) | LEAD/LAG | 🎯 Target (RAG) | Horizon | Owner | Source of truth |
|---|---|---|---|---|---|
| Operating ratio (OR) | LAG | < 93 | Monthly | DET | QB P&L — pending #55 |
| Revenue per mile (RPM) | LAG | 🔴 set (within regional bands: drive-away $3.00–4.50, haul $4.50–7.00) | Monthly | DET | Ops log + QB |
| Cost per mile (CPM) | LAG | 🔴 set (🟡 model ≈ $3.04/mi on the LA-port lane — assumption-heavy) | Monthly | DET/AM | QB job cost |
| Deadhead / empty-mile % | LEAD | 🎯 beat industry ~28–35% (proactive backhaul is the lever) | Monthly | DET | Dispatch log — stand up |
| Truck utilization | LAG | 🔴 set (rev-miles ÷ truck ÷ week) | Monthly | DET | Ops log |
| Per-haul margin | LAG | ≥ 30% (revenue − COGS) ÷ revenue | Per job | DET/AM | QB job P&L by class |
On CPM: the ≈$3.04/mi figure comes from the LA-port empty-container cost model, built on May 2026 industry data and assumed diesel prices, not DET's realized cost basis. It is a 🟡 planning number, not a booked actual — do not launder it to 🟢.
4 · RETURN — actuals (the books)¶
Fed monthly after close. Each cell cites source + as-of. — = pipeline not yet wired. Revenue, margin
and OR are gated on the GL-detail pipeline (TaskLoop #55, core.gl_fact); job-cost and per-job
margin wire from QB now.
| KPI | 🎯 Target | Actual | Variance | As of | Source / status |
|---|---|---|---|---|---|
| Operating ratio (OR) | < 93 | — | — | — | QB P&L — pending gl_fact (#55) |
| Revenue per mile | bands | — | — | — | ops log + QB — wire |
| Cost per mile | set | — | — | — | QB job cost — wire |
| Deadhead % | < ~30% | — | — | — | dispatch log — not stood up |
| Truck utilization | set | — | — | — | ops log — wire |
| Per-haul margin | ≥ 30% | 30.5% (DRWY0604, Monticello drive-away) | +0.5 pt | 2026-06-09 | QB job P&L — one closed job; 2 open (HAUL0611, HAUL0616) COGS TBD |
What's real today 🟢: DET is billing STE against Rate Schedule v1.0 with June 2026 activity — one closed drive-away (
DRWY0604, $7,500 rev / $5,212 COGS / 30.5% margin) and two open trailer hauls pending COGS. That is the entire booked history; everything else awaits the monthly pipeline.
5 · REPORT — monthly review (the loop)¶
After each close: diff Targets vs Return, name the misses, set next month's actions, append a dated entry to the decision log. Humans read it here; agents ingest it.
Latest month: (pending first close)¶
- On track: — · Behind target: — · Research that changed: — · Actions (owner → by when): — · Decisions logged: —
How the monthly report is produced
- AM confirms DET books closed. 2. Rig-pool allocation runs (
am_det_alloc_truckin.py), P&L by Class refreshed. 3. Job pulls RETURN metrics (Metabase → table) once #55 lands revenue/OR. 4. An agent pass writes Latest month: variances, misses, 3–5 actions, marker changes in §0/§2. 5. Dated review → decision log; "Last review" updates. 6. Quarterly: refresh the cost model and work the 🔴 gaps (addressable market, make-vs-buy threshold, outside-freight coverage).
6 · AI OPERATIONS — how AI serves DET's goals¶
- DET dispatch agent = the scheduling brain — continuously knows driver position/HOS, the salvage pipeline (from the Auction Manager via the truck-spine event bus), and load boards; books/recommends loads that maximize earning while keeping salvage priority, HOS legality, and home-basing as hard constraints. Makes single-truck optimization economically viable that a human dispatcher never could.
- Cost-estimate + margin gate — every quote runs an estimated COGS check; margin < 30% escalates to Brandon before an invoice posts. Automates the ≥30% discipline.
- Rig-pool allocation script —
am_det_alloc_truckin.pyposts the monthly classifying JEs so per-job P&L stays true. - AI KPI: deadhead % driven down (proactive backhaul capture) + per-haul margin held ≥ 30%.
How AI uses this page¶
Direction for every DET-facing agent. Work traces to a Targets row; Return says whether it moved; Report recalibrates monthly; every 🔴 (activation blockers, make-vs-buy threshold, RPM/CPM/utilization target-setting, deadhead-log stand-up, driver arrangement) and every 🟡 (outside-freight coverage, dispatch agent, CPM model) is standing to-do — research to run or an assumption to prove or kill. The reliability floor under the salvage engine, kept honest and pointed at outcomes.