From Report to Pipeline

Why this document exists

SWIFT did not publish The Space Economy Jobs Report to publish a report. The stated purpose is a working talent pipeline for employers. This document takes the report as the first input to that pipeline and asks what has to happen next for it to produce one.

The operating model

Building a talent pipeline for employers requires five steps in order. Each step consumes the previous step's output.

The chain has one structural property worth stating plainly: it is subtraction. Step 3 subtracts step 2 from step 1. Subtraction requires both operands to be expressed in the same units. That single requirement is where the chain currently breaks.

Where the chain breaks

Break 1 — Step 1 is not complete

The report delivers occupations and credentials, not skills. "Welder," "technician," "AWS D17.1" are labels for bundles of skills, not skills themselves. A pipeline cannot be planned from a list of job titles, because a job title cannot be trained, measured, or subtracted.

This is not an oversight in the report. It is the state of the entire field — no one deconstructs postings into skills, because the tooling to do it at scale did not exist. It is the specific thing PowerTechs does.

Break 2 — Steps 1 and 2 are denominated in different units

Demand, as the report captures it, is expressed in certifications and standards. Supply — the way any workforce agency, college, or employer describes a talent pool — is expressed in degrees, credentials, and years of experience.

This is why no one in the sector has a number for the gap, only a qualitative sense that one exists. Step 3 is currently unexecutable, which means steps 4 and 5 are being run on intuition.

Break 3 — Step 5 assumes every gap is trainable by the institutions on hand

SWIFT's inventory of pipeline instruments is community colleges, technical schools, and regional training centers. That inventory closes gaps that a curriculum can close. It does not close the others — and the report does not distinguish between the two.

This is the most expensive break, because it explains the report's own central paradox.

The welding paradox is a diagnosis, not an anomaly

The report establishes that the shortage is not of people who can weld, but of welders who meet exacting standards. Follow it to the end:

1. If the binding constraint were technical, certification would relieve it. Train more, certify more, fill the roles.
2. Texas colleges teach welding. Certified welders exist. The report confirms ample national supply.
3. The roles remain open.

This matters directly and immediately for SWIFT's mission: the strategy as written — align colleges to industry certifications — executed perfectly, produces more certified welders and the same open roles. On the report's own largest single gap (11.9% of postings), the plan does not reach the constraint.

We think we can show exactly where the constraint hides. Below is one real posting from a Texas employer.

Worked example: Firefly Aerospace, Flight TIG Welder II, Briggs TX

We deconstructed one real posted role from a company inside SWIFT's own network.

The posting contains 31 discrete requirement lines: 12 responsibilities, 9 required qualifications, 10 desired qualifications.

The closest is a schedule-flexibility line — an availability condition, not a capability — listed last under Desired.

What the posting says the job is

The summary states that the welders work closely with Fluid Systems Engineers to ensure the build is executed to instruction. That is the job, in Firefly's own framing.

The screen does not screen for the thing the summary says the work is.

The second layer, hiding inside layer-1 language

Every line below is a real requirement from the posting. The left column is what Firefly wrote. The right column is what it actually demands of a person.

What this proves

Firefly describes the job accurately in prose, then screens on 31 technical items, none of which touch the layer that determines whether the hire holds up. This is not a Firefly failure — it is what every employer does, because no instrument exists for the other layer.

Why SWIFT should care more than Firefly does

A Texas community college building a welding curriculum from this posting will build to all 31 items. It will teach AWS D17.1, GTAW, borescope inspection, precision measurement, tube fabrication, cleaning to aerospace standards. Every one of them correctly.

This is Break 3, in one document, at one employer, twenty minutes from SWIFT's office.

Hypothesis worth testing on the full dataset

We noticed something in a second Firefly posting we do not have enough data to claim. In the Welding Engineer III listing — a degreed role — the durable layer is described: self-direction, internalizing intent, filling gaps, driving improvement to completion without supervision, working with stakeholders. Written in prose, unnamed as skills, unscreenable — but present.

In the welder posting, that language is absent entirely.

If that pattern holds across the 5,000 postings, it means the non-degree track — half of the report's demand — receives the least articulation of the layer that determines its success. That would be a significant finding for SWIFT, and it is testable on data you already hold.

The routing rule

This is the practical output. Once a gap is deconstructed into layers, it routes:

GAP IDENTIFIED
    │
    ├── LAYER 1 — technical / procedural
    │      Closed by: curriculum + practice + certification
    │      Instrument: community colleges, technical schools, apprenticeships
    │      SWIFT has this. It works.
    │
    └── LAYER 2 — durable
           Closed by: repeated exposure under realistic conditions, with feedback
           Instrument: simulation, or years on the floor, or prior formation
           SWIFT has none. Neither does anyone else.

Durable skills are trainable. This is the correction we would offer to the common framing. They are not innate, and they are not soft. The military trains them continuously and deliberately. They are simply not trainable by the instruments currently in SWIFT's inventory, because they do not transfer through a course. They accumulate through repetition under pressure with correction.

Without this rule, a pipeline plan routes durable gaps into colleges, where they will not close, and the resulting time-to-close estimate is wrong by an order of magnitude — in the optimistic direction. That is the most dangerous kind of planning error: it looks like a plan.

Step 4: time-to-close, and where it compresses

Time-to-close is the step nobody can currently answer, and it is the step that determines whether a pipeline plan is a schedule or a wish.

Our estimates come from practitioner expertise, not from published data — no validated dataset on durable-skill time-to-close exists in this sector, ours included. We offer the framework as the contribution and treat the numbers as what the pilot calibrates. We would rather say that plainly than present estimates as measurements.

What we can state structurally:

This is the entire basis of what PowerTechs does, and it is why the timeline compresses. Absent an instrument, layer 2 accumulates at the speed of the job — years on the floor, if the employer is patient enough to carry someone that long, which the report shows they are not. A simulation delivers the same mechanism at a controlled rate, and produces a measurement as a byproduct.

Step 2: the talent pool not counted

The report inventories Texas's educational institutions. It does not mention veterans, the military, or transitioning service members.

We would submit that Texas's largest source of layer-1-and-layer-2-complete technical talent is not in the inventory. Look again at the Firefly posting through this lens:

This is the point we would most want SWIFT to take. The military is not a source of raw candidates needing durable-skills training. It is the one institution in the country that trains durable skills deliberately, at scale, and by design. A transitioning aircraft structural maintainer has spent years doing precisely what the Firefly posting's unnamed layer requires.

So they land in the quadrant the current screen cannot see: layer 2 complete, layer 1 uncredentialed. Firefly's 31-line screen reads them as unqualified — while the thing that would make them succeed is invisible to it.

This is a rare shape for a workforce problem. The expensive half is already built and free; the cheap half — certification — is exactly what SWIFT's existing inventory of colleges is good at.

And the placement mechanism already exists and costs the employer nothing. Under DoD SkillBridge, a service member in their final 180 days trains full-time with a civilian employer while the Department continues paying salary, allowances, and benefits; industry partners are prohibited from paying wages. PowerTechs is an approved SkillBridge provider.

Prototype: two weeks

Not a platform. One role, deconstructed end to end, so the method can be judged on evidence rather than description.

Scope: welder / precision technician postings from SWIFT's 27 surveyed employers.

Input needed from SWIFT: the subset of postings for this role family, plus the taxonomy structure. Roughly 200–500 postings. If the subset is not readily extractable, we can start from public postings and the value is only slightly reduced.

Output — a Skill Demand Profile, one page:

The "Implied — never stated" column is the whole message. Employers require these skills in fact, do not name them in the posting, therefore they do not reach the curriculum, and therefore the pipeline does not produce them.

Success criterion: does the method account for the paradox the report identified but could not resolve? If yes, it scales to all 27 employers and becomes step 1 of the pipeline. If no, SWIFT has spent two weeks and learned something real.

Then, immediately testable: whether the durable-layer score predicts x-ray reject rate and standards-qualification pass rate better than certification status. One employer, one criterion. That is the validity study, and it is also the pilot.

What we are asking for

1. The welder-family posting subset and the taxonomy structure — the two-week prototype input.
2. Sandy Barker's review of the deconstruction. Someone who designed simulators at NASA and built training and compliance at Firefly is unusually well placed to say whether we have the layer-2 attributions right. We would want that read whether or not anything else proceeds.
3. A conversation about data ownership before the pilot, not after. We would rather raise it early and plainly: who owns the measurements, who may publish, what happens to the data if the partnership ends. Easy to settle now, expensive to settle later.