Frontline & Deskless Workforce Statistics 2026: 7 in 10 Frontline Exits Are the Employee's Own Choice

Frontline turnover is usually explained as something that happens to an employer: seasonal demand, a soft labour market, an industry that has always churned. The federal data does not support that reading. When you take the US Bureau of Labor Statistics' Job Openings and Labor Turnover Survey and split separations into the ones the employer initiated and the ones the employee did, frontline industries turn out to be the industries where departures are most likely to be the worker's own decision — and desk-based industries, which churn less overall, turn out to churn for the opposite reason.

That one split changes what the numbers are evidence of. An industry losing 4.9% of its jobs a month to separations of which 71% are quits has a retention problem. An industry losing 4.1% a month of which only 51% are quits has a restructuring problem. Both show up as "high turnover" in a headline. Only one of them is fixable by anything a people team does.

This page assembles what is actually published and verifiable about frontline and deskless workers in 2026: the turnover matrix by industry, two derived metrics computed here with the arithmetic shown, the benefits and rewards access gap between occupation groups, and the measured relationship between recognition and staying. Every figure is attributed to a named release with its reference period. The statistics that circulate widely but should not be cited as published are listed separately in section 10 rather than repeated as fact.

1. How many frontline and deskless workers are there?

There is no official answer, and this is the first thing to get right if you are citing a number. No national statistics agency publishes a "deskless" or "frontline" category. Neither term is an occupational classification; both are descriptions of working conditions that cut across classifications. A hospital nurse, a warehouse picker, a shop assistant and a site electrician are all frontline, and they sit in four different industry supersectors.

The figure everyone quotes — that 80% of the world's 2.7 billion workers do not sit at a desk — comes from Emergence Capital, a venture firm, as part of its deskless workforce research. Its accompanying survey covered approximately 1,500 deskless workers across 20 industries, and the firm's best-known framing was that despite the size of this population, less than 1% of venture funding went to companies building technology for it.

It is a reasonable estimate and the proportion is almost certainly in the right region. It is also, importantly, a 2020 private-report estimate that has been recirculated without a date for six years. If you cite it, cite it with its origin and its age. If you need a figure that a statistics agency will stand behind, use industry employment instead and say which industries you are counting.

2. The turnover gap: what JOLTS actually says, by industry

JOLTS reports hires, quits and total separations each month as a rate: the count during the month as a percentage of employment. The table below is the August 2026 release, which is preliminary, seasonally adjusted, and the most recent available at the time of writing. The last three columns are not in the BLS release — they are derived here, and section 3 and section 4 explain each one.

Source: US Bureau of Labor Statistics, Job Openings and Labor Turnover Survey, August 2026 (preliminary), tables 2, 3 and 4, seasonally adjusted. Derived columns computed as described in sections 3 and 4. Transportation, warehousing and utilities is omitted from the derived columns because only its quits rate (1.9%) is quoted here.

The spread is the finding. A hospitality employer and a bank are operating in the same labour market under the same macroeconomic conditions, and one loses three and a half times as large a share of its workforce to voluntary departures every month as the other. Nothing about that gap is explained by the business cycle, because both figures are drawn from the same month.

Health care is the quietly interesting row. At 2.0% it looks moderate, close to the all-private average, which is how it is usually reported. But health care and social assistance is one of the largest employers in the US economy, so a 2.0% monthly quits rate represents an enormous absolute number of departures, and — as section 3 shows — two thirds of its separations are voluntary. For a sector with structural staffing shortages and credential-gated hiring, that mix is the problem, not the rate.

3. Derived metric: the Voluntary Exit Share

JOLTS publishes quits and total separations as separate tables. It does not publish the ratio between them, and that ratio is where the frontline story actually lives. We compute it here and call it the Voluntary Exit Share:

Voluntary Exit Share = quits rate ÷ total separations rate. It answers a question the headline turnover number cannot: of everyone who left this month, what share chose to?

For accommodation and food services in August 2026 that is 3.5 ÷ 4.9 = 71%. For retail, 3.0 ÷ 4.4 = 68%. For information, 1.1 ÷ 2.5 = 44%. For professional and business services, 2.1 ÷ 4.1 = 51%. The all-private benchmark is 2.1 ÷ 3.5 = 60%.

Read the two charts together and the usual framing inverts. On raw separations, professional and business services (4.1%) is barely below retail (4.4%) and well above manufacturing (2.5%) — it looks like a high-churn sector. On the Voluntary Exit Share it is near the bottom of the table at 51%, because almost half of its separations are layoffs, discharges and end-of-contract events. Information is the clearest case: a 2.5% separations rate against a 1.1% quits rate means 56% of departures in information were not the employee's choice, the highest involuntary share of any group here.

Frontline industries are the mirror image. They lose more people and a larger share of those people are walking out. That is the definition of a retention problem, and it is the part of turnover that responds to pay, scheduling, management quality and whether anyone notices good work — rather than to the demand forecast.

4. Derived metric: the monthly churn index

Separations are only half of the workload churn creates. Every departure is followed by a requisition, a hire, an onboarding and a ramp, and the hires rate is published alongside the separations rate. Adding the two gives what we call the monthly churn index: the number of employment transitions per 100 jobs in a month.

Monthly churn index = hires rate + total separations rate. For accommodation and food services in August 2026: 5.2 + 4.9 = 10.1 transitions per 100 jobs, in one month.

More than one job in ten in US hospitality changed hands or changed state in a single month. Financial activities managed 1.8 + 1.9 = 3.7. The ratio between them is 2.7×. Annualised naively, the hospitality figure implies on the order of 121 transitions per 100 jobs a year — which is to say that the average hospitality job is involved in more than one hire or separation event per year.

The practical use of this index is in sizing the operational load rather than the attrition rate. A district manager with 400 frontline staff at a churn index of 10 is personally absorbing roughly 40 employment transitions a month — interviewing, inducting, rostering around gaps and re-training. That is the real cost centre, and it is invisible in a turnover percentage.

5. The rewards access gap, measured by the BLS

The second federal dataset worth putting next to the turnover numbers is the BLS National Compensation Survey, published as Employee Benefits in the United States. Its March 2026 edition, released on 25 September 2026, reports benefit access by occupation group — and it contains the single most useful statistic in this whole report for anyone running a rewards programme.

37% of private-industry workers in service occupations had access to nonproduction bonuses, against 61% in management, professional and related occupations. Nonproduction bonuses are the BLS category that covers the discretionary, non-piece-rate payments most recognition and incentive programmes run through: holiday bonuses, referral payments, attendance and safety awards, spot awards. On the BLS's own measure, a frontline worker is roughly 1.65 times less likely to work somewhere that can give them one at all.

Source: US Bureau of Labor Statistics, National Compensation Survey, Employee Benefits in the United States, March 2026, released 25 September 2026. Figures are percentages of workers with access to the benefit, which is a higher number than participation in every case where both are published.

Line this up against section 3 and an uncomfortable pattern appears. The occupation groups with the highest Voluntary Exit Share are the same groups with the least access to the rewards that influence whether someone stays. The population most likely to leave by choice is the population least likely to be given a reason not to.

6. The part-time penalty is sharper than the occupation gap

Frontline work is disproportionately part-time, and part-time status is where the benefits data turns brutal. In the same March 2026 release, 23% of part-time private-industry workers had access to medical care, against 87% of full-time workers — a 3.8× gap, and a wider divide than any occupation-group comparison in the dataset.

This matters for programme design in a way that is easy to miss. If your recognition or rewards programme is delivered through benefits infrastructure — a payroll bonus code, a benefits portal, a pension-linked award, anything that assumes an employee is enrolled in something — then a large share of your frontline population is structurally excluded from it before you have designed a single award. The eligibility question is settled by the employment contract, not by the programme.

7. Engagement and recognition: the measured retention link

Gallup puts global employee engagement at 20% in 2025, and reports 31% of US employees engaged and 17% actively disengaged as of May 2026. The two figures are not comparable with each other — different populations, different fieldwork — and should not be presented as a trend line, a point section 10 returns to.

The recognition evidence is more useful, because one study tracked the same people over time rather than asking them how they felt once. Gallup and Workhuman followed the career paths of nearly 3,500 employees from 2022 to 2024. Their findings:

Source: Gallup and Workhuman, recognition and retention research, nearly 3,500 employees tracked 2022 to 2024. The five pillars the research scores recognition against are that it be fulfilling, authentic, personalised, equitable and embedded in the culture.

The 22% and 55% figures are the ones to pair with the BLS access gap. Recognition is not scarce because employers dislike it; it is scarce because most employees are not receiving any that meets a quality bar, and in frontline occupations the mechanism to deliver it often does not exist. That is a solvable problem, and it is a different problem from engagement.

A 45% reduction in observed two-year turnover is a larger retention effect than most compensation interventions can demonstrate, and recognition is the cheaper of the two. The constraint on frontline recognition is almost never budget. It is delivery.

On why frontline workers leave, McKinsey's retail and frontline research consistently identifies career development, uncompetitive compensation and inspiring leadership among the leading drivers of departure, with flexibility ranking first in several of its workforce surveys. We cite that direction of travel rather than specific percentages here, because the figures differ between its reports and the underlying survey details are not consistently published alongside them.

8. The deskless technology gap

Emergence Capital's deskless survey, for all that its headline figure is dated, produced three findings that have held up as the most-quoted description of the deskless technology problem: 75% of deskless workers spent at least half their working time using some form of technology, 83% were primarily provided with desk-bound devices to do work that is not done at a desk, and 78% treated the availability of technology as an important or major factor when looking for their next role.

Treat all three as 2020 survey findings from approximately 1,500 respondents across 20 industries, because that is what they are. The reason they are still worth citing is the mismatch they describe: the tooling gap for frontline workers is not an absence of technology but a mismatch of form factor. Something designed around a desk, a corporate login and an enrolment record does not reach someone who has a shift, a phone and a locker.

This is precisely why gift-card and stored-value rewards became the default frontline recognition currency rather than points platforms: a code delivered to a phone number or a printed slip needs no account, no enrolment and no corporate email. If you are building that kind of programme at scale, our bulk gift card API and the practical guide to buying gift cards in bulk for employees cover the delivery mechanics, and frontline employee recognition in 2026 covers the programme design.

9. The 2026 frontline workforce benchmark cheat sheet

Everything above, condensed to the numbers worth keeping. Each row states its source and reference period so it can be cited directly.

10. Frontline statistics to handle with care

Five numbers in this field circulate far more widely than their evidence supports. They are listed here so that this page is useful for checking a claim as well as making one.

"Retail turnover is 60% a year" (or 100%, or 130%)

Annual retail turnover figures in the 60–130% range are common in vendor marketing and almost never BLS. They are usually total separations annualised, sometimes for a single sub-segment such as part-time hourly store staff, and sometimes drawn from a client panel rather than a representative sample. The BLS retail quits rate of 3.0% a month annualises to 36%, and the total separations rate of 4.4% to roughly 53%. If you need a defensible annual retail figure, derive it from JOLTS and say which rate you annualised.

Multiplying a monthly JOLTS rate by twelve

We show the ×12 column above because readers want it, but it measures quit events per 100 jobs a year, not the share of employees who leave. A single high-churn post that turns over four times in a year contributes four quits to the numerator while occupying one job in the denominator. In high-churn frontline settings the overstatement relative to headcount turnover is substantial, which is exactly where the annualised figure is quoted most often.

Pairing Gallup's global 20% with the US 31%

These are different populations measured at different times with the same instrument. The 20% is global for calendar 2025; the 31% is US as of May 2026. Putting them in one sentence implies a US-versus-world comparison that neither figure was designed to support, and placing them on a timeline implies an increase that is an artefact of the two samples.

"80% of the global workforce is deskless" presented as an official figure

It is a venture-firm estimate first circulated around 2018 and surveyed in 2020. No statistics agency publishes a deskless category. It is perfectly citable with its origin and date attached; it is not an official statistic and should not be footnoted as though it were.

"Most frontline workers have no corporate email address"

This claim appears constantly with percentages attached, usually in the 70–80% range, and we could not trace it to a published primary source with a stated sample. The underlying observation is sound and matches the device-mismatch findings in section 8, but we have not put a number on it here, and nor should anyone citing this page.

11. What to actually do with this data

Methodology and sources

Every figure on this page is drawn from one of the sources below. Derived metrics are labelled as derived, and the arithmetic is shown inline so any reader can reproduce or dispute it. Nothing here is modelled, estimated or extrapolated beyond the stated division or addition of two published rates.

A note on what is deliberately absent. There is no cost-per-frontline-hire figure on this page, no industry-specific annual turnover percentage presented as fact, and no engagement figure broken out for frontline workers. Credible versions of all three exist in vendor research; none that we could verify against a primary source with a published sample and fieldwork period. Where we could not verify, we left the number out rather than sourcing it to the nearest article that repeated it.

Key takeaways

• Frontline turnover is overwhelmingly voluntary: 71% of accommodation and food separations and 68% of retail separations were quits in August 2026, against 44% in information. High frontline churn is a retention problem, not a demand problem.
• The raw turnover rate misleads in both directions. Professional and business services looks high-churn at 4.1% separations but is only 51% voluntary; health care looks moderate at 2.0% quits but is 67% voluntary in one of the largest workforces in the economy.
• Churn creates operational load, not just attrition. At 10.1 transitions per 100 jobs a month, hospitality managers absorb roughly one transition per ten staff every month — 2.7× the financial activities rate.
• The population most likely to leave by choice has the least access to the rewards that influence staying: 37% nonproduction bonus access in service occupations against 61% for managers and professionals, and 23% part-time medical access against 87% full-time.
• Recognition has the strongest published retention evidence of any low-cost intervention — 45% lower observed two-year turnover — and only 22% of employees say they get the right amount of it. In frontline settings the binding constraint is delivery, not budget.
• Cite carefully. The 80%-deskless figure is a dated venture estimate, annualised JOLTS rates count events rather than people, and the widely repeated "no corporate email" percentages have no primary source we could locate.