This article applies to England and Wales.
Tribunal evidence is moving from paper to screen
In brief: The communications in substantial employment tribunal judgments are shifting from letters toward email and messaging apps. Between 2017–19 and 2023–25, email rose from 69% to 82% of single-claimant merits judgments in England and Wales, WhatsApp from 4% to 16%, and letters fell from 88% to 80%. Paper-only judgments halved; digital-only doubled; judgments mentioning both remain the norm.
Last updated: September 2026
By Yerty | Analysis of published Employment Tribunal decisions, 2017 to 2025.
Two thirds of substantial tribunal judgments mention both letters and digital messages, in both periods. The change is in the minority that mention only one. In 2017–19 that minority was mostly paper. Now it is mostly digital.
This analysis follows three words through nine years of published merits judgments: email, WhatsApp and letters. It reports them year by year, then jointly, then within each claim type and industry, and it tests whether a changing mix of published claims could produce the pattern on its own. The observed changes in claim mix and industry mix do not.
The headline finding
| 2017–19 | 2023–25 | Change | |
|---|---|---|---|
| 69.4% | 81.7% | +12.3 | |
| 3.6% | 16.1% | +12.5 | |
| Letters | 87.9% | 79.9% | −8.0 |
Share of substantial single-claimant merits judgments mentioning each term at least once. n = 2,029 (2017–19) and 1,738 (2023–25).
Email was already the dominant digital channel at the start of the period and has consolidated. WhatsApp has gone from one judgment in twenty-five to one in six. Letters remain in four judgments out of five, but were in nearly nine out of ten.
Method. Population is single-claimant merits decisions with written reasons, England and Wales, one judgment per case, restricted to a fixed band of 20,000 to 60,000 characters so that longer judgments do not mention more simply by being longer. Each term is a single word matched on the judgment text (email including "e-mail"; WhatsApp in any casing; "letter" or "letters" as the word itself, not the wider correspondence group). Single words are used because they match reliably: multi-word phrases in the extracted text are often separated by runs of spaces or line breaks, and a literal-space match misses 9% to 41% of the documents containing them, with the miss rate falling over time; a single word is never split, so its series carries no such bias. Scotland was run separately: email and WhatsApp move the same way there, but letters fall only 3.0 points against 8.0 in England and Wales, so the decline in letters is stated as an English and Welsh finding and the article is confined to England and Wales.
Year by year
Share of substantial judgments mentioning each term at least once, by year of judgment. England and Wales, single-claimant merits decisions, 20,000–60,000 characters.
Click a legend entry to hide a series. n per year is shown on hover; the table below carries every figure.
| Year | n | Letters | ||
|---|---|---|---|---|
| 2017 | 596 | 66.6% | 2.3% | 88.8% |
| 2018 | 617 | 69.9% | 2.6% | 89.1% |
| 2019 | 816 | 71.1% | 5.3% | 86.4% |
| 2020 | 621 | 76.3% | 4.7% | 83.7% |
| 2021 | 991 | 77.1% | 9.9% | 86.0% |
| 2022 | 797 | 75.9% | 13.0% | 84.4% |
| 2023 | 676 | 80.0% | 14.9% | 81.4% |
| 2024 | 562 | 81.9% | 17.3% | 78.8% |
| 2025 | 500 | 83.8% | 16.4% | 79.2% |
The shape is a broad trend rather than a step. Email rises across the period, with one small dip in 2022, and levels in the low eighties. WhatsApp is flat until 2019 and rises across the following five years, with a dip in 2020. Letters fall across the period, with small rises in 2018 and 2021. No single 2020 discontinuity is visible; the transition was under way before the pandemic and continued through it, which is all the series establishes.
2017 is a partial year. The 2025 WhatsApp dip sits in the year where the length band retains the smallest share of judgments (see below) and should not be read as a turn until 2025 matures.
Mostly substitution, with some addition
Marginal rates cannot say whether digital references are replacing paper ones or accumulating alongside them. The joint distribution can.
Share of judgments mentioning digital terms (email or WhatsApp) and paper terms (letters), both, or neither.
Provisional: from the first-cut population (n 2,039 / 1,960). To be re-issued on the re-run population (request 34).
Digital here is email, WhatsApp, screenshots or text messages; paper is letters. [From the first cut, n = 2,039 and 1,960; to be re-issued on the re-run population before publication.]
The biggest change is substitution: judgments mentioning paper correspondence without digital communications have become much less common (22.7% to 11.6%), while digital-only judgments have become much more common (8.8% to 16.8%). There is some accumulation as well, since the share mentioning both rises from 64.8% to 67.6%. The share mentioning neither does not move. Mixed digital-and-paper judgments remain the norm throughout.
The pattern holds inside every claim type
Evidence vocabulary could vary with the kind of claim, and the mix of claims published could have changed between the two periods. Both are true. Neither explains the finding.
| Claim type | n 2017–19 | n 2023–25 | Letters | ||
|---|---|---|---|---|---|
| Unfair dismissal | 1,470 | 1,216 | 67.8% → 80.7% (+12.9) | 3.4% → 16.2% (+12.8) | 91.9% → 85.4% (-6.5) |
| Unlawful deduction from wages | 558 | 652 | 68.5% → 80.2% (+11.7) | 4.8% → 22.4% (+17.6) | 86.4% → 74.2% (-12.2) |
| Breach of contract | 638 | 616 | 67.9% → 80.7% (+12.8) | 4.2% → 16.4% (+12.2) | 90.6% → 80% (-10.6) |
| Disability discrimination | 320 | 454 | 70% → 79.5% (+9.5) | 2.2% → 10.8% (+8.6) | 88.8% → 78.6% (-10.2) |
| Working time | 278 | 410 | 66.2% → 80.2% (+14.0) | 3.6% → 22% (+18.4) | 83.1% → 75.1% (-8.0) |
| Race discrimination | 244 | 315 | 76.2% → 81.3% (+5.1) | 3.7% → 13.7% (+10.0) | 80.7% → 76.8% (-3.9) |
| Public interest disclosure | 153 | 195 | 78.4% → 79.5% (+1.1) | 7.2% → 16.4% (+9.2) | 86.3% → 79.5% (-6.8) |
| Sex discrimination | 172 | 181 | 72.7% → 76.8% (+4.1) | 7.6% → 17.1% (+9.5) | 86% → 81.2% (-4.8) |
| Age discrimination | 86 | 129 | 79.1% → 82.2% (+3.1) | 1.2% → 12.4% (+11.2) | 89.5% → 78.3% (-11.2) |
| Redundancy | 111 | 120 | 64.9% → 77.5% (+12.6) | 1.8% → 9.2% (+7.4) | 92.8% → 83.3% (-9.5) |
| Religion or belief | 48 | 60 | 77.1% → 76.7% (-0.4) | 8.3% → 10% (+1.7) | 77.1% → 73.3% (-3.8) |
| Maternity and pregnancy | 99 | 56 | 71.7% → 78.6% (+6.9) | 11.1% → 26.8% (+15.7) | 87.9% → 85.7% (-2.2) |
| Contract of employment | 48 | 55 | 62.5% → 74.5% (+12.0) | 0% → 25.5% (+25.5) | 81.3% → 70.9% (-10.4) |
| Health and safety | 35 | 34 | 71.4% → 70.6% (-0.8) | 5.7% → 11.8% (+6.1) | 71.4% → 73.5% (+2.1) |
Provisional: from the first-cut population (n 2,039 / 1,960). To be re-issued on the re-run population (request 34).
A judgment can carry several claim types and counts in each row it belongs to. Every cell has at least 30 judgments. [From the first cut; to be re-issued on the re-run population before publication.]
WhatsApp rises in fourteen claim types out of fourteen. Email rises in twelve; the two flat rows are the two smallest. Letters fall in thirteen.
Wages and working-time judgments lead on WhatsApp, at over 22% in 2023–25, which is consistent with messaging apps being where shift patterns and pay queries are now discussed. Discrimination claims sit lower on WhatsApp but show the same direction.
Reweighting. The claim mix did shift between the periods: unfair dismissal is 7.4 points smaller a share of the later period, working time and disability discrimination each 2.6 points larger. Holding the 2023–25 judgments to the 2017–19 claim mix moves email by 0.03 points, WhatsApp by 0.09 points and letters by 0.78 points. Because a judgment can carry several claim types, share-based reweighting on overlapping rows is an approximation; a version on mutually exclusive claim profiles is being run. The within-family table above does not depend on it.
A second composition test: industry
The claim-type test above has a weakness: claim types overlap. Industry does not, since each case carries one primary respondent industry, so it gives an independent and cleaner check on whether the shift is an artefact of which workplaces reach a published judgment.
It is not. Holding the 2023–25 judgments to the 2017–19 industry mix moves email by 0.6 points, WhatsApp by 0.4 and letters by 0.1. And the WhatsApp rise appears in every one of the eighteen industries with enough judgments to report, including the desk-based ones where it starts lowest: technology 2.7% to 6.0%, the public sector 1.5% to 6.6%. The levels differ between industries in the way anyone would expect; the direction does not. Where the change was largest is less expected, and is the subject of the companion piece, WhatsApp evidence by industry: the fastest rises are in healthcare, education and the public sector, and retail is the one large sector where letters did not give way.
The length band, and what it does to later years
The fixed band controls length within a year. It does not hold a fixed share of the population across years.
Share of merits judgments inside the 20k–60k character band
Median length of all merits judgments (characters)
Two measures, two charts: the band retains a falling share of judgments while the median judgment grows longer.
Retention falls from 49% to 30% while the median judgment grows from about 41,000 to 45,000 characters. It falls out of the top: later years exclude a growing proportion of long judgments.
If the digital terms are commoner in long judgments, the band will understate their rise in later years, which would make the reported trend conservative. That is an inference, not a measurement. A direct standardisation over judgment length, measuring each term within finer length bands and reweighting every year to the same length distribution, is being run and will replace the band when it returns.
What this analysis does not show
A mention is not an exhibit, and not reliance. The corpus counts what judgments say. A judgment mentioning WhatsApp may be describing a message the tribunal relied on, one it excluded, or one a witness denied sending. No negation handling is applied.
It is a filtered population. Only decisions with written reasons are published, and only the substantial ones are in the band. Settled, withdrawn and briefly decided cases are absent. Multi-claimant cases are excluded throughout.
Three words are not the whole of evidence. This analysis deliberately uses single words that can be matched reliably. Wider categories (procedural documents, witness evidence, medical evidence) exist in Yerty's term index, but two of the groups feeding them were found to contain terms that do not mean what the group is named for, and they are not quoted until rebuilt. The single-word series are unaffected.
Covert recordings are not reported here. Whether references to secret recordings of workplace meetings have risen is a distinct question with a fragile measurement, and a first pass shows it flat. It will be examined separately once a proper term list has been hand-checked.
The extraction has not been validated against a hand-coded sample, so precision and recall are unmeasured.
What this means in practice
The written correspondence that substantial tribunal judgments describe is now predominantly electronic, and messaging apps are the fastest-growing part of it. Letters still appear in four judgments in five, so paper has not disappeared; it has stopped being the thing a judgment mentions on its own.
The shift is not explained by the observed change in claim mix or industry mix. It is consistent with a broader change in the communications appearing in workplace disputes, although judgment text cannot establish the underlying mechanism: changes in drafting, in what judges choose to mention, in representation or in which cases are published cannot be ruled out from the text alone. The rise in messaging-app references is not confined to the workplaces where they were always likely; it appears in every industry examined.
Frequently asked questions
How often are WhatsApp messages mentioned in employment tribunal judgments?
In 16.1% of substantial single-claimant merits judgments in 2023–25, up from 3.6% in 2017–19. Of the three communication terms analysed, WhatsApp grew fastest, and it rises in every one of fourteen claim types. In wages and working-time judgments it exceeds 22%.
Are tribunals mentioning letters less?
Yes. Letters appear in 79.9% of substantial judgments in 2023–25, down from 87.9% in 2017–19, in England and Wales. In Scotland the fall is only three points. The share of judgments that mention letters but no digital correspondence halved over the same period, from 22.7% to 11.6%.
Is digital evidence replacing paper evidence, or adding to it?
Mostly replacing, on the vocabulary of judgments. Paper-only judgments fell from 22.7% to 11.6% and digital-only rose from 8.8% to 16.8%, while judgments mentioning both rose slightly from 64.8% to 67.6%. Substitution is the larger movement; some accumulation is also present.
Could the change be because different kinds of claims are being published?
Not on the observed claim mix. It did shift, but holding the later period to the earlier mix moves email by 0.03 points, WhatsApp by 0.09 and letters by 0.78, and the pattern appears within each claim type separately. Other unobserved changes in the published population cannot be excluded from the text.
Did the shift happen during the pandemic?
No single 2020 step is visible. Email rises across the period, WhatsApp is flat to 2019 and rises across the following years, and letters fall across the period, none of them monotonically. The transition was under way before 2020 and continued through it.
Does a mention mean the tribunal relied on the message?
No. The analysis counts references in the written reasons, which can describe evidence that was relied on, excluded, disputed or merely mentioned. It measures what judgments discuss, not what decided them.
Analyse the data
Email, WhatsApp and letters are three of the terms in this index. Run the same series for any claim type, industry or term.
Intelligence Hub →Manage your evidence
If your case runs on messages and emails, keep them in one place with their dates and context before disclosure.
Case Hub →Sources
Analysis of published Employment Tribunal decisions, 2017 to 2025, drawn from the public register on GOV.UK under the Open Government Licence v3.0. Population: single-claimant merits decisions with written reasons, England and Wales, one judgment per case, 20,000 to 60,000 characters; 2,029 judgments in 2017–19 and 1,738 in 2023–25.
- Employment tribunal decisions, GOV.UK
- Employment Tribunal Procedure Rules 2024 (SI 2024/1155)
- Presidential Guidance and Practice Direction: Recording and Transcription of Hearings, 20 November 2023
Related reading: can WhatsApp messages, emails and recordings be used as evidence at an employment tribunal, employment tribunal claims: what you need to know, and how to prove unfair dismissal.