Predicting which subcontractor a general contractor will hire
From prior hires alone: right first guess 59.1% of the time, in the top three 78.4%.
Method
Strictly chronological. For each hire we use only that general contractor's earlier hires — never anything that happened later. Candidates are ranked by past frequency.
Result — Philadelphia, 17,019 mechanical hires
| Condition | Top-1 | Top-3 | n |
|---|---|---|---|
| any prior history | 56.8% | 73.2% | 13,865 |
| 3+ prior hires | 59.1% | 78.4% | 10,788 |
| 5+ | 60.3% | 80.1% | 8,993 |
| 16+ | 60.7% | 83.4% | 4,809 |
| random from 1,804 active firms | 0.06% | — | — |
A prediction is possible for 81.5% of hires. Excluding cases where the general contractor self-performs (26.6%), top-1 is 48.4% and top-3 is 67.2%.
Two scales, not one
Loyalty makes the event more predictable and the package harder to win. A general contractor who gives two thirds of the work to one firm reaches a mechanical permit in 69.7% of cases; one who spreads the work reaches it in 45.6%. These are shown as separate numbers and never combined into a single score.
Direct answer
Our analysis uses the public records and sample shown below to measure a historical pattern, not to predict one future job with certainty. The answer is useful only with its jurisdiction, sample and time horizon attached; therefore each result should guide research priority and be verified against the issuing authority before commercial action.
How to use this research
These results are descriptive measurements from the stated records, market and observation window. Use them to choose what to investigate first, then verify an individual project with the issuing authority. A historical result is not a guarantee of a future permit, award or hiring decision.
Conclusion
The practical value is a narrower and more explainable research queue. The evidence is strongest when its jurisdiction, sample and horizon remain attached to the claim; that is why Projects Ahead does not turn local measurements into a single national score.