More Factory Output Can Mean Fewer Factory Jobs
Reshoring and output growth do not restore the old employment ratio. Compare employment per unit with the skill mix before treating a new plant as a return.
Factory-automation traces good output, paid hours, and capital economics through a line; workforce-transformation reads the occupational composition that remains when those ratios change.
A new domestic plant can be a real investment and still restore neither the old number of direct jobs nor an accessible path into the technical work it creates.
The ribbon is cut outside the new factory.
The parking lot is smaller than the old one was.
There are new panels on the roof, fresh guardrails around the cells, and a sign using the words reshoring, capacity, and local jobs in the same sentence.
Maybe your old plant closed. Maybe a line returned from abroad. Maybe the company expanded one building while another one emptied.
Do not decide what the new building means from the size of the crowd at the opening.
Count what it makes.
Then count the labour required for each sellable unit.
The building may be a real investment.
The jobs may be real too.
Neither fact restores the old employment ratio.
The Parking Lot
Headcount is the number most likely to appear in a press release, because it travels well.
It is also incomplete.
A plant with more people can be using less labour per unit if output rose faster than employment.
A plant with fewer people can be producing more if machines, fixtures, material flow, and quality control have reduced the hours needed for each good unit.
The productive question is not how many jobs there are.
Ask how many paid hours and people this plant uses for the output it actually sells.
Write down two periods.
The old line.
The new line.
For each period, record annual good output, not gross cycles.
A good unit meets specification and can be sold without rework.
Keep first-pass yield, rework, scrap, downtime, changeover loss, and customer returns beside it.
Then record direct production hours.
Divide direct production hours by good output.
That is direct production hours per good unit.
Do not put an output figure from one period under a headcount from another.
Do not compare a nameplate capacity claim with last year’s actual production.
Keep the product, unit, and period attached to every number.
The ratio is not a forecast.
It is a record of labour intensity.
It tells you whether a factory is asking fewer hands to make each sellable thing.
BLS defines manufacturing labour productivity as output per hour, and publishes output and hours as separate index series alongside it.
Those three series move close together and are not the same series.
Treat that as a rule for the plant in front of you.
Output.
Hours.
Productivity.
Keep all three.
Two Columns
Take out a sheet of paper.
Draw two columns: old line and new line.
Across the top, write the period and the product or product family.
A mixed line can make any ratio dishonest if the old plant made a simple variant and the new plant makes a complex one.
Now fill the rows.
Good output. Record units that meet specification and can be sold without rework.
Direct production hours. Record paid hours spent making, loading, assembling, inspecting, packing, or otherwise moving the product through the line.
Maintenance hours. Record preventive maintenance, breakdown response, calibration, spare-parts work, tooling, and controls support.
Engineering hours. Record programming, product revisions, process design, quality analysis, commissioning support, and the labour required to keep variants running.
Wages. Record the wage bill or, if that is unavailable, the wage level and paid hours for each role family.
Do not treat a higher technical wage as a payroll increase or decrease without its hours.
Calculate three things.
Direct production hours divided by good output.
Maintenance hours divided by good output.
Engineering hours divided by good output.
Then calculate the technical-role share: maintenance, engineering, controls, programming, quality, and related roles as a share of stated headcount, paid hours, or wage bill.
Name the denominator.
Do not switch it halfway through the comparison.
None of this is validated.
It is a way of deciding whether jobs returned describes the work that returned.
If direct production hours per unit fall while the technical share rises, call the result workforce transformation.
Do not call it a simple jobs return.
The work has changed its composition.
That distinction matters, because a lower direct labour figure can coexist with more maintenance, more engineering, more supplier sorting, or more customer effort.
A fall of 6 operator-hours per 100 units can coincide with 2 additional maintenance-hours.
The difference does not vanish because it appears in another department.
Externalized work is still work.
The Fast Cell
Do not let cycle time choose the denominator.
A cell can complete a 30-second cycle and still fail the operating test if it waits for material, rejects too much, or needs frequent intervention.
Ask what the machine does when the feeder is empty, a double part arrives, a sensor is obscured, pressure drops, power fails, a guard opens, a barcode is rejected, the network disappears, or a tool is changed incorrectly.
These are not edge cases once production begins.
They determine whether a plant has bought capacity or a demonstration.
An industrial robot system includes more than the moving arm: the robot, end effector, control system, power sources, sensors, and monitoring or sequencing interfaces all belong to it.
The work around that system becomes visible during commissioning.
Someone stages material.
Someone verifies lots.
Someone replenishes consumables.
Someone reads an alarm.
Someone performs preventive maintenance.
Someone changes a tool, inspects output, and recovers after a fault.
OSHA’s 2024 manufacturing narrative review identified 550 robot-related incidents.
Treat that as a reason to include setup, recovery, guarding, tool changes, and bypassed sensing in the labour account.
Do not write unattended as if it means nobody carries the risk.
Your old role may not disappear in one clean motion.
It may be broken into loading, audit, fault clearing, quality review, maintenance callout, and vendor escalation.
Count the pieces where they land.
Reshoring
Returning production to a country can be sensible without recreating the old factory.
It can reduce inventory exposure, shorten lead times, improve response time, locate engineering nearer to a customer, protect product control, or reduce the risk of a supply interruption.
It can also use an automated line that needs far fewer direct assembly hours than the overseas process it replaces.
That is not a contradiction.
It is the capital economics of a plant whose purpose is output and reliability.
The Reshoring Initiative reports that U.S. manufacturing costs can remain 10% to 50% above offshore competitors, depending on the case.
That is an advocacy organization’s estimate, not a universal cost premium.
Keep the range with its producer.
The comparison includes more than hourly wages.
Freight, tariff exposure, inventory, quality escape, lead time, intellectual-property risk, supplier coordination, and disruption cost all sit in the calculation.
The company is not necessarily choosing between one worker and one machine.
It is choosing a system of production.
The same organization’s 2024 U.S. report counted 1,412 reshoring and foreign-direct-investment cases and 244,940 announced jobs.
It also says actual hiring typically lags announcements by 12 to 24 months.
Write announced beside the number.
Do not quietly convert it into filled jobs.
Do not convert it into production volume.
Do not convert it into a job yield per unit.
In the same data, 67% of those job announcements were in electrical equipment and computer or electronics industries.
That concentration describes a recent investment mix.
It does not tell you how many jobs a particular reshore creates.
Name the thing the number measures before you use it.
An announcement measures an announcement.
Where the Gain Goes
Automation often creates capacity before it creates savings.
The company can use it for higher volume, shorter lead times, additional shifts, more variants, more reliable delivery, or a constrained labour market.
If demand grows, output can absorb labour in packing, internal logistics, maintenance, quality, supplier management, and downstream service.
If demand is flat, the same investment can reduce labour input per unit and expose a reduction in paid hours, replacement hiring, or headcount.
Neither outcome is implied by the machine alone.
The financial model should count capital recovery, energy, consumables, maintenance, support, labour, depreciation, financing, and the cost of work shifted elsewhere.
A robot invoice is not the project cost.
Nor is your wage the whole alternative.
Unit labour cost is labour compensation divided by real output.
Unit cost is broader. It includes material, energy, capital recovery, maintenance, and overhead allocated to units that meet specification.
Customers pay for good units.
The machine does not get credit for defective volume.
From 1987 to 2023, BLS measured annualized labour-productivity growth of 7.2% per year in computer and electronic products and 0.1% per year in apparel manufacturing.
Those are output-per-hour measures, not profitability results for a particular automation purchase.
The contrast is the warning.
Product complexity, demand, process stability, labour content, and capital intensity change the result.
Over that same long period, labour productivity rose in 23 of 24 manufacturing and mining industries.
In 2023 alone it fell in 17 of 21 manufacturing industries.
Do not turn a productivity series into a promise.
Use it as a ratio with a sector and a period.
When cost per good unit falls, the gain can become a lower price, higher margin, new investment, higher wages, shorter hours, increased output, dividends, or some combination.
Productivity does not choose the destination.
Competition and demand affect price and output.
Ownership and bargaining affect pay.
Staffing policy affects hours and jobs.
Ask where the gain went.
The Technical Share
Look at the technical share without romanticizing it.
Automation shifts cost toward installation, programming, calibration, condition monitoring, repair, spare parts, tooling, and controls support.
A few hours of unplanned downtime at a bottleneck can consume the saving calculated from many operator-hours.
That makes maintenance capability a productive asset.
It also makes access to that capability a labour question.
U.S. industrial machinery mechanics, maintenance workers, and millwrights had a median annual wage of $63,510 in May 2024.
Use that as a reminder that higher-paid technical work has a wage structure.
Do not use it to claim that any particular plant pays more, or that every former assembler can enter it.
The plant must budget training hours, backfill, assessment, wage progression, and a role when the training ends.
If it does not, reskilling is a transition cost shifted onto you.
A course that lets an operator reset an alarm may reduce callouts.
A credential without access to controls work does not create a new occupation.
Ask for the route in writing.
Which role is open after training?
Who gets paid time to train?
Who covers the old shift?
Who assesses competence?
What wage applies on completion?
What happens to the workers who do not receive a slot?
The answers show whether a technical share is a workforce plan or a press-release adjective.
In one manufacturers’ outlook survey, 72% of respondents named attracting and retaining a quality workforce among their top two business challenges.
That tells you firms have a staffing problem.
It does not establish an accessible ladder from your line job to the scarce role.
Keep the two facts separate.
Ladder Thinning
The loss usually arrives in the hiring plan before it arrives in the redundancy notice.
A factory can retain experienced workers, stop replacing routine departures, collapse several junior steps into one experienced technical role, or require entrants to bring credentials the old line once supplied through paid work.
That is entry-ladder thinning.
Stable headcount can hide it.
BLS projects 18.34 million annual occupational separations across all occupations for 2024–34: 10.19 million transfers to another occupation and 8.16 million labour-force exits.
A declining occupation can still have vacancies, because separations can exceed projected employment decline.
Do not tell a young worker that a vacancy proves the old ladder still exists.
Ask how many junior roles were posted.
Ask how many were filled internally.
Ask what training was paid for.
Ask whether the new technical roles are internal, vendor roles, contractor roles, or jobs in another region.
A national industry-occupation matrix is a check on a local plant’s role mix, not a local forecast.
The local evidence is the plant’s own output, vacancies, retirements, quality records, contractor use, and occupational composition.
Production, transportation, and material-moving occupations represented 17.3% of employment for Black workers and 8.7% for Asian workers in BLS’s 2023 occupational distribution.
Those are descriptive shares, not characteristics of people.
They show why a changed factory role mix does not distribute exposure or access evenly.
The affected place loses more than payroll when the ladder narrows.
It can lose supplier work, instructors, equipment support, union knowledge, and the informal route through which people learned the industry.
An advanced role in a distant market does not restore that local function.
The Long Count
Study the long record before you accept a revival headline as a labour-market verdict.
U.S. manufacturing employment fell from 19.6 million in June 1979 to 12.8 million in June 2019: a reduction of 6.7 million jobs, or 35%.
That is not one factory’s story.
It is a warning against treating manufacturing as one uniform occupation.
Within nondurable manufacturing, apparel and textile industries lost 81% of jobs across those forty years, while food manufacturing rose about 8%.
One sector label carried sharply different outcomes.
The same holds inside a plant.
The relevant record is not only total employment.
It is the mix of roles, the paid hours attached to each, the output base, who has access to the remaining work, and whether the next cohort can enter.
Do not hide a classification change inside a trend either.
Census analysis of the SIC-to-NAICS transition found that reclassification increased the measured services share by roughly 36% and moved about 1.4 million jobs into services without any change in economic activity.
That is 7.6% of manufacturing employment.
If a chart changes classifications without a concordance, it can manufacture a transformation that did not occur.
Check the codes.
Check the period.
Check the denominator.
Then check where the work went.
And do not spend the evening deciding whether the ribbon cutting is good news or bad news.
It is very likely both: more reliable domestic output, and fewer accessible direct jobs per unit.
Those two facts have never needed each other’s permission.
Minimum Viable Action
Before you sleep tonight, find one public production or capacity number for the plant making the promise.
Write it beside the current headcount.
Write the period beside both.
Then write what the plant makes.
You do not have employment per unit yet.
You have stopped letting one number impersonate it.
What you will be able to say afterwards
- Calculate direct production hours per good unit instead of treating headcount as a production measure.
- Separate a reshoring announcement, announced jobs, actual hiring, and production volume.
- Record maintenance and engineering hours when direct assembly hours fall.
- Measure the technical-role share alongside total paid hours and wages.
- Test whether a local jobs claim includes contractors, suppliers, and the entry roles that formerly fed the plant.
- Keep output bases and periods attached to employment-per-unit and occupational-composition series.
Questions this raises
Does a new factory mean the old factory jobs are back?
No. A new domestic factory can restore production, shorten supply chains, and create direct roles while using fewer direct assembly hours for each good unit than the earlier operation. The employment result depends on output demand and staffing policy. A reshored automated line can make more units with a smaller direct-production workforce and a larger relative need for maintenance, engineering, quality, and external support.
What number shows whether a factory is using less labour for what it makes?
Direct production hours per good unit shows labour intensity. The calculation divides direct production hours in a stated period by good units meeting specification in the same period. Good output matters because gross cycles can include scrap, rework, or defective units. Headcount alone cannot show whether a plant employs fewer people because output fell or because each sellable unit now requires fewer hours.
Why is headcount not enough to judge a reshoring announcement?
Headcount lacks a production denominator and a period. A plant can add workers while output rises faster, leaving fewer paid hours or workers per unit. It can also retain headcount while reducing entry roles through attrition. A credible account carries good output, paid hours, headcount, wages, and capital expenditure together, then distinguishes an announcement from actual hiring and from realized production.
What should be counted after a manual station becomes an automated cell?
The operating record should compare good units, first-pass yield, scrap, rework, planned and unplanned downtime, changeover loss, intervention count, direct labour hours, maintenance hours, engineering hours, contractor hours, wages, vacancies, injuries, and customer returns against the pre-change baseline. This reveals whether labour is saved, moved to another cost centre, converted to capital services, or used to support higher output.
Can a factory save operator hours while creating more technical work?
Yes. An automated cell can reduce repetitive direct labour while increasing installation, programming, calibration, condition monitoring, repair, spare-parts, tooling, controls, quality, and engineering work. A fall in operator-hours does not establish a fall in all human work. The relevant comparison records maintenance and engineering hours with direct production hours, then identifies whether the added work is internal, contracted, or supplied by vendors.
What does a reshoring jobs announcement actually prove?
It proves that an organization has announced a project and its expected jobs under its reporting method. The Reshoring Initiative counted 1,412 reshoring and foreign-direct-investment cases and 244,940 announced jobs in its 2024 U.S. report, while stating that actual hiring typically lags announcements by 12–24 months. An announcement is not a count of filled jobs, output, or employment per unit.
Why can a plant produce more without hiring in the old ratio?
Automation can lower direct labour hours per good unit, improve repeatability, add capacity, shorten lead times, and support more variants or shifts. If demand grows, total employment can rise while labour intensity falls. If demand is flat, the same capacity can permit fewer paid hours or fewer replacement hires. Output growth and job growth therefore describe different outcomes and need separate measures.
Does higher factory productivity prove that workers received the gain?
No. A reduction in cost per good unit can become lower prices, higher margins, new investment, higher wages, shorter hours, increased output, dividends, or a combination. Competition and demand affect prices and output; ownership and bargaining affect pay; staffing policy affects hours and headcount. Productivity is a ratio, not an instruction for how a factory distributes its gain.
How can a factory look stable while its entry jobs disappear?
A factory can retain incumbent workers while ceasing to replace routine departures, reducing junior hiring, or requiring entrants to arrive with credentials that earlier jobs provided on the job. This thins the entry ladder before an occupation vanishes. Vacancies can remain in a declining occupation because separations exceed projected decline, so vacancy counts need to be read with entry hiring, training access, and promotion data.
Is technical upgrading an automatic route from an assembly job to a better role?
No. Training changes access only when it is connected to paid time, assessment, wage progression, and a real role with access to the relevant equipment or controls work. A course that permits an operator to reset an alarm can reduce callouts; a credential without a job route does not create a transition. Employers choose who receives training, a guaranteed role, a credential, or only notice of a shrinking job family.
What does the technical share of a factory workforce show?
The technical share shows the portion of paid hours, headcount, or wage bill held by defined maintenance, engineering, controls, quality, programming, and related roles in a stated period. Rising technical share alongside falling hours per good unit indicates workforce transformation rather than a simple return of old employment. The series must keep its occupational definitions, output base, and period visible to remain comparable.
Why should contractors and suppliers be included in a local jobs claim?
Automation can move work from direct assembly to equipment suppliers, integrators, contractor maintenance, upstream sorting, metrology, logistics, and engineering services. Excluding those workers can make labour look eliminated when it has been moved; counting them as identical local plant jobs can overstate the direct benefit. A credible claim separates construction, equipment suppliers, direct roles, component roles, logistics, and induced spending.