When One Thesis Is Real and the Other Isn’t
Findings
Overview
THE SITUATION
A mid-market logistics company had spent 18 months developing an internal hypothesis: their proprietary dispatch data — accumulated over a decade of carrier operations — was the foundation of a competitive intelligence product. Phase 1 was an internal ML-assisted dispatch optimization platform. Phase 2 was an external SaaS product sold to regional freight brokers.
The venture lead had built a compelling internal case. The CEO had approved a sprint to validate it before Stage 2 investment. The CFO had set a clear financial gate: 24-month full payback, positive contribution margin within 18 months.
WHAT THE SPRINT INVESTIGATED
Over 14 working days, the sprint ran five parallel evidence streams:
• Four internal stakeholder interviews — venture lead, engineering lead, CFO, VP Operations
• Two independent external expert interviews via GLG — a technical expert with direct experience in comparable data systems; a qualified buyer with budget authority and no prior relationship with the client
• A bottom-up market sizing rebuild from primary and secondary sources
• A full unit economics model across three scenarios: optimistic, base case, and bear case
• Five adversarial analytical techniques (the Black Box Bias Audit) — designed to surface structural weaknesses the venture team’s own analysis could not see
Finding 1 — The data foundation requires a phase the plan doesn’t include
The engineering lead’s 18-month build estimate assumed the company’s data was ready for ML training. An independent technical expert — with no prior relationship with the client — identified that the behavioral attribute layer of the data had approximately 40–60% coverage. Before a single line of model code could be written, 12–18 months of data normalization work was required. This phase appeared nowhere in the venture plan. Total timeline to a production-grade internal tool: 30–36 months from board approval. Bias-adjusted build cost: $2.6M–$3.5M.
Finding 2 — The external market is structurally closed.
The Phase 2 thesis assumed regional freight brokers would purchase dispatch intelligence from a carrier. Two independent qualified buyers — interviewed separately, with no connection to each other — gave the same answer: they would not. The reason is structural, not a pricing or communications problem. Brokers do not purchase competitive intelligence from a company whose core business competes with them for carrier capacity. Confirmed by two independent Tier 2 sources.
Finding 3 — The unit economics fail at the market level, not just the model level.
The sprint rebuilt the market from the bottom up. Year 3 serviceable obtainable market: approximately $45,000 in annual recurring revenue. Required ARR to justify Stage 2 investment: approximately $438,000. The market is roughly ten times too small to support the external SaaS thesis. Not a growth problem — a structural ceiling
LTV:CAC (optimistic) 0.78:1
Required: 3:1 Year 3 reachable market ~$45K ARR Required: ~$438K
Build timeline (total)30–36 months Plan assumed: 18 months
Finding 4 — The internal platform thesis holds, conditionally.
Phase 1 — the internal dispatch optimization platform — rests on a defensible host thesis. The company is the only entity with the company’s data. The internal efficiency case is unambiguous if the data foundation proves out. The venture is not a bad idea. Phase 2 is a broken configuration of a real data asset.
Recommendations
Phase 1: CONDITIONAL GO — The internal platform is viable.
Three condition categories must carry named owners before capital is committed: an explicit governance decision on engineering resource sequencing; a full data foundation audit before build begins; and named owners on implementation risk thresholds. A CONDITIONAL GO is a real GO — with conditions that are real. If conditions cannot carry named owners, the verdict moves to NO GO.
Phase 2: NO GO — Two independent structural grounds, neither addressable within the current venture scope
The unit economics fail the 3:1 LTV:CAC gate in all three scenarios. The addressable market is ten times too small for Stage 2 investment. These are not execution risks. A NO GO verdict always names what comes next: in this case, a structurally different commercial path — direct data licensing to a technology partner rather than direct SaaS sales to brokers. Different buyer, different economic unit, different value proposition. Worth a scoping exercise on its own terms.
Interested in a diagnostic for your own venture?