Market and competitors

Twenty vendors checked against their own pricing pages on 28 August 2026. The short version: the category costs about £2.75 per user per month^S14, the credible direct competitor is Blink, and nobody offers translated video captions or UK data residency.

Three findings that change the map

1. Beekeeper no longer exists as an independent vendor. beekeeper.io 301-redirects to lumapps.com. LumApps: "Beekeeper is now part of LumApps… We are building a unified product experience that will roll out in phases over the next 12 to 18 months." LumApps acquired Beekeeper, announced 1 July 2025, backed by Bridgepoint, combined entity valued over $1bn. Every Beekeeper account in a renewal window for the next 18 months is a displacement target.
2. SafetyCulture has rebranded to Mitti. safetyculture.com 301-redirects to mitti.com. Battlecards saying "SafetyCulture" are stale. Its Heads Up module is the single most direct competitor to TeamTalk' core loop — and at $24/seat/month it is priced so that operatives never get licensed.
3. HammerTech is the competitor the brief missed. It runs toolbox talks with attendance tracking and sign-offs, prices on annual construction volume plus active job sites rather than seats, and gives subcontractors free access — so the seat-cost objection that defeats Mitti does not apply to it. It does not translate. That single gap is where the fight is.
4. Yapster was acquired by Sona (announced 17 Jan 2023) and the brand is gone. Sona now sells workforce management, not comms. Yapster's UK chain base — BrewDog, Caffè Nero, Krispy Kreme, Next, Welcome Break, Dunelm, Wagamama, Ann Summers, Safestore — is orphaned. That is a named displacement list.

Market map

TeamTalk competes across four layers, and the objection is different in each.

The GTM fact that matters: in construction, catering and logistics the incumbent is almost never a product. It is a site WhatsApp group on a foreman's personal phone. Layers A–C are who you get compared to in procurement; layer D is what you actually replace.

Feature comparison

Legend: ● full/native · ◐ partial, paid tier, or undocumented depth · ○ absent · ◆ committed but not shipped or certified (TeamTalk only; nothing aspirational carries a ●) · ? unverifiable

3.1 The construction sub-market, where the real fight is

Layer B vendors do not sell into construction; layer C vendors do. This is the table that matters for Barratt London and Fernham Homes.

Read the two blank rows. Nobody in the construction stack translates, and nobody captions video. Everybody has a site layer we don't. That is the trade, and it argues for integrating (Biosite/Insite for access, HandsHQ for RAMS, Procore for the project) rather than building.

TeamTalk column reflects the shipped data model: post_actions with confirmed_at and reminder tracking, hazard_reports, post_translations / video_caption_translations, push_subscriptions, 70+ seeded languages, tenant scoping, translation gating on posts.

The construction sub-market

Layer B vendors do not sell into construction; layer C vendors do. This is the table that matters for Barratt London and Fernham Homes.

Read the two blank rows. Nobody in the construction stack translates, and nobody captions video. Everybody has a site layer we don't. That is the trade, and it argues for integrating (Biosite/Insite for access, HandsHQ for RAMS, Procore for the project) rather than building.

TeamTalk column reflects the shipped data model: post_actions with confirmed_at and reminder tracking, hazard_reports, post_translations / video_caption_translations, push_subscriptions, 70+ seeded languages, tenant scoping, translation gating on posts.

Pricing landscape

Median published price for a frontline comms app ≈ £2.75 per user/month. The number to beat is £2.95 (Blink Core). Blink puts advanced multi-language publishing in Pro at £3.75 and lists reporting and export as a separate paid item (confirm before quoting). TeamTalk is not priced per user (Chapter 3), but the buyer will divide anyway: a full Site plan works out at £1.80 per enrolled worker per month, and the Company bands at £1.30–£2.20 at their ceilings, with translation, translated captions, confirmed reach, export and reminders all included and no minimum. On that arithmetic it undercuts every credible competitor except a stripped Connecteam Basic — which has no translation on that tier, no captions, no acknowledgement export, no UK entity and no UK/EU data residency.

What "per user per month" means when 80% of users are frontline

This is the central commercial insight of the category and belongs in every pitch.

• A 400-person housebuilder has perhaps 35 people who need a desk-grade licence and 365 who need to receive a notice and press a button. Every vendor priced per uniform seat forces the buyer to choose between paying office rates for site labour or not licensing the site labour at all. They choose the second — which is exactly why Mitti's Heads Up, Teams and Procore's comms do not, in practice, reach operatives. The compliance gap is created by the pricing model, not the feature set.
• Vendor user counts systematically under-count the workforce. Vendor A "has" 2,000 UK frontline users at a customer employing 9,000. The other 7,000 are on WhatsApp.
• Churn destroys per-named-seat economics in the ICP. Agency labour on a housebuilding site can turn over 30–50% a year. Annual per-seat licensing means paying for leavers. That is why Ocasta sells per site, Procore prices on turnover, and Mitti offers "site-based pricing" at Enterprise — three independent vendors reaching for the same fix.
• Minimums are the hidden floor. Oneteam's 100-user minimum, Flip's €4,950, Workvivo's $20,000/250 employees and Staffbase's ~$20k + implementation are not rate cards — they are exclusion criteria. A 60-worker site (Fernham Homes runs about 220 people across the year, in sites of that size) cannot buy four of the eight direct competitors in this document, and Notify would want a 3–5 year contract.
• Three independent vendors have already abandoned per-seat for exactly this reason, and they are the ones that sell into construction. Procore prices on annual construction volume with unlimited users; HammerTech prices on construction volume plus active job sites and gives subcontractors free access; Ocasta sells per site — "everyone at that location gets access." Mitti offers site-based pricing at Enterprise. Notify gives free reporting to any employee. This is not a pricing fashion; it is the sector telling you that headcount is the wrong unit.

Implications for TeamTalk pricing:

1. Price so that licensing 100% of the workforce is the default, not a budget decision. Sub-£2 achieves this; £3+ does not.
2. Offer a per-site / unlimited-operatives option for construction. "£X per site per month, unlimited workers" removes the single biggest procurement friction in the sector and matches how a housebuilder thinks about cost.
3. Never price admin seats separately. The buyer counts the 400, not the 35.
4. Bundle the export. Blink appears to charge for it (verify). It is the artefact the buyer is actually purchasing.
5. Cost floor to model: Twilio Verify $0.05 per successful verification + UK outbound SMS $0.056 ≈ $0.106 (~8p) per onboarding/login event (Verify · SMS GB), plus OpenAI translation cost per post × language count. Both are per-event, not per-seat — which is precisely why TeamTalk can sustain £1.50 where a vendor carrying an enterprise sales and CS org cannot go below £2.95.

Positioning

Where TeamTalk wins

Where TeamTalk loses

The honest gaps

1. "Confirmed" proves an affirmative act, not comprehension. Do not sell it as legal cover. Sell it as materially better evidence than anyone else can produce — which is true — and offer an optional comprehension question on critical notices. Note that Mitti already claims the stronger version ("proof they've been read and understood" via digital signature); be careful not to be out-claimed on your own ground.
2. Machine translation of safety instruction carries residual risk. The defensible framing is "translated automatically, reviewable before publish" — never "certified translation".
3. Web push is best-effort. iOS PWA push requires the user to add the site to the home screen, and permission can be revoked. The record is robust; the delivery is not guaranteed. SMS fallback for critical notices is the answer and should be priced in from day one.
4. Blink is genuinely good and genuinely cheap. £2.95 with mandatory reads, CSV export, phone/group-code activation, near-miss forms, offline mode, Cyber Essentials Plus and a UK logo wall including Stagecoach and the NHS. The wins against it are real but narrow: captions, QR onboarding, export included, translation-at-base-price, UK residency and price. Do not pretend they are wider.
5. Single-market concentration. UK-only is a strength in the pitch and a weakness on the cap table.

The one-sentence position

The construction-specific version: "HSE says you must give workers comprehensible information. On a London site where a third to three-quarters of your operatives don't have English as a first language, a toolbox talk in English isn't comprehensible — and a signature on a sheet doesn't prove it was. TeamTalk translates the talk, translates the captions on your site manager's phone video, and gives you the per-worker sign-off sheet, for every operative on site, not just the ones with a licence."

Who you are actually against, by deal shape

Five messaging angles against WhatsApp

Each is framed as a question, because the point is to let them reach the answer.

1. "Whose phone is your safety record on?"

UK GDPR. A site WhatsApp group makes the foreman's personal handset a processing location for personal data of every worker in it — names, numbers, photographs, and often incident and health detail. The organisation remains the controller and carries the liability, but has no access, no retention control, no records of processing for that channel, and no way to respond to a subject access request or an erasure request. Do not tell them what that is under the UK GDPR; ask two questions and let them draw the conclusion: "If a worker asked for every message about them, could you produce it? If they asked you to delete it, could you?" Ask them directly: if a worker asked you today for every message about them, could you produce it? If they asked you to delete it, could you? TeamTalk is one controlled tenant, UK-hosted, where an admin can search, export and delete.

2. "The foreman left on Friday. Where did the group go?"

Leavers. When a manager leaves, the group, its history and every worker's mobile number leave with them — and if they are the group admin, they may retain access until someone remembers to remove them. Conversely a departed worker stays in the group indefinitely, receiving operational information about a site they have no right to see. There is no offboarding process for a WhatsApp group because there is no system of record. In TeamTalk, offboarding is one status change: access ends, the history stays with the company, and the worker's confirmation record remains on file as evidence.

3. "Show me who read the asbestos notice."

No audit trail. Blue ticks tell you a message was displayed on a device. They can be switched off by the recipient. They cannot be exported. They do not distinguish "delivered to a phone in a locker" from "read by the operative". And they say nothing about the seventeen people never added to the group — or, on a Business App broadcast list, the people who silently never received it because they hadn't saved your number. When an HSE inspector, an insurer, a principal contractor's auditor or a tribunal asks for evidence that a named worker was informed of a named hazard on a named date, a screenshot of a group chat is the weakest answer available. TeamTalk produces a per-post and per-worker report, timestamped, printable, with non-responders listed and chased automatically. CDM 2015 reg. 13(4) requires the principal contractor to ensure "a suitable site induction is provided" — nothing in the regulations credits you for a message you cannot prove landed.

4. "Your Romanian bricklayer just got a safety notice in English."

No translation. A WhatsApp group broadcasts one message in one language. WhatsApp's on-device translation (shipped ~Sept 2025) is recipient-initiated, per chat, on the reader's own device — the sender cannot broadcast once and guarantee everyone reads it in their language, and gets no confirmation that translation happened. It also does nothing for the photograph of a laminated notice, which is how most site information is actually shared, and nothing for the video the site manager recorded on his phone.

The regulator has already written this argument for you. MHSWR 1999 reg. 10 requires "comprehensible and relevant information" on risks and protective measures. HSE, updated June 2024:

HSE goes further and names the remedy — its list of acceptable means includes "use professional translation software or free online tools" and non-verbal communication including video. And the scale is not marginal: UK contractors surveyed by the University of Reading / RICS in 2026 reported non-English-speaking operatives at an average of 34% of their workforce, reaching 75%; ONS put non-UK nationals at 40% of London's construction-of-buildings workforce; CIOB reports migrant workers at up to 50% of the London onsite workforce, "many of [whom] speak little or no English."

English-only on a multilingual site is a defensible-sounding practice that becomes indefensible the moment there is an incident. TeamTalk publishes once and every worker receives it in their own language — including the captions on that video.

5. "It's free until it costs you."

The cost that isn't on the invoice. WhatsApp's price is zero and its cost is contingent: a failed safety prosecution, an uninsured claim, a tribunal where the only evidence of instruction is a deleted group, a data protection enquiry you cannot answer, or a lost tender because the principal contractor's PQQ asked how you evidence communication of safety information to a multilingual workforce and you had nothing to attach. And if they try to do it "properly" via the WhatsApp Business Platform, it stops being free: per-message pricing since 1 July 2025, a daily toolbox talk that probably falls into the marketing category — the most expensive tier, the one excluded from volume discounts, and the one whose UK rate rose again in July 2026 — plus a BSP, template pre-approval and a developer. TeamTalk at £1.50–£2 per worker per month, under £25 a year, is not competing with free; it is competing with the cost of not being able to prove anything.

On the "but they like WhatsApp" objection: don't try to ban it. WhatsApp is fine for "the gate code changed". It is not fine for anything you would ever need to prove. Draw the line at notices that need a record — a rule a site manager can actually enforce, and precisely the scope TeamTalk sells.

How big is this, honestly

How many UK companies look like the ICP, what they are worth at list price, and what a realistic year-three share of that is. Every step shown.

The competitor chapter sizes the competition and not the market. This fixes that. It is deliberately short, and every number below can be reproduced from a free public dataset in about two minutes.

Companion documents: the sector playbooks (this chapter) · the construction playbook (this chapter) · the value models (below) · verification-log.md.

Step 1 - Count the enterprises

Source: ONS UK Business Counts - enterprises by industry and employment size band, 2025, derived from the Inter-Departmental Business Register, reference date 14 March 2025, published 24 September 2025. Retrieved from the free Nomis API on 28 August 2026 (dataset NM_142_1, geography United Kingdom, legal status all, measure count).

Reproduce it with:

`` https://www.nomisweb.co.uk/api/v01/dataset/NM_142_1.data.csv ?geography=2092957697 &industry=146800681,146800682,146800683,146800695,146800696, 146800689,146800690,146800691,146800692,146800693, 146800720,146800721,146800650,146800651, 146800726,146800727,146800728 &employment_sizeband=6,7,8,9 &legal_status=0&measures=20100&date=latest ``

UK enterprises with 100 or more employees, in the six ICP groupings:

Nomis rounds counts to the nearest 5, so the column sums carry a rounding error of a few units. It does not move any conclusion.

Step 2 - Price each band

The ONS size bands map almost exactly onto TeamTalk's Company bands. The boundaries differ by one worker (ONS "100 to 249" against TeamTalk "≤250"; ONS "250 to 499" against "251-500"); immaterial at this resolution.

⚠️ Deliberately conservative on the top band. ONS's "1,000+" is open-ended, so it contains both the 1,001-2,000 band at £2,600 and everything above it, which is Enterprise from £3,600. All 1,025 of them are priced here at £2,600. The top band is therefore understated, probably materially.

Total addressable market, at list price, if every one of them bought:

≈ £96.4m of annual recurring revenue at list price. That is the theoretical ceiling, and it is the least useful number here.

Step 3 - The multilingual-share assumption

TeamTalk's proposition is auto-translated notices. An enterprise with a wholly English-reading workforce can buy it for the confirmed-receipt record alone, but it is a materially weaker sale. So the addressable market is the multilingual subset.

The assumption: 67% of these enterprises have a materially multilingual frontline workforce. Labelled assumption.

Where 67% comes from. Of the 200 companies on the target lists, 134 carry an evidenced multilingual flag - the majority of them (116) evidenced by presence on the UK register of licensed sponsors (workers), downloaded in full from the Home Office on 28 August 2026 (142,988 sponsor entries), matched by legal entity name; the remainder evidenced by a published workforce nationality statistic or modern slavery statement. The other 66 are recorded as "Likely - not yet evidenced", not as monolingual. Full method and per-company sources are in verification-log.md.

134 ÷ 200 = 67%.

Two reasons it might be too high, and two reasons it might be too low. State all four; a good investor or a good co-founder will find them.

Too high: the 200 targets were selected for multilingual fit, so they are not a random sample of the 8,115 - they are the top of the distribution. And a sponsor licence evidences that a company sponsors some overseas workers, not that its frontline is multilingual at scale.

Too low: a sponsor licence is needed only to recruit from outside the UK labour market. An employer whose non-UK staff hold settled status, pre-settled status or Irish citizenship needs no licence and will never appear on the register - and several of the largest employers on the list fall into exactly that category. So the 134 is a floor for evidenced, not a ceiling for actual. Independently, ONS puts the non-UK-born share of the UK construction workforce at 15.7% and of London's at 57.4% (see the construction playbook (this chapter)).

Sensitivity, because one assumption should never carry a whole model: at 50% the SAM is £48.2m; at 67%, £64.6m; at 80%, £77.1m.

Step 4 - SAM

8,115 × 67% = 5,437 enterprises, distributed proportionately across the bands:

Step 5 - Year-three SOM

The capture assumption: 0.5% of SAM enterprises by the end of year three. Labelled assumption.

5,437 × 0.5% = 27 customers, distributed as the SAM is:

Cross-check, bottom-up, because a top-down percentage is not evidence. Founder-led sales with a 5-9 month cycle and 4-7 stakeholders per deal, working a list of 200 named accounts: roughly 4-6 customers in year one, 10-14 cumulative by the end of year two, 25-30 cumulative by the end of year three, before any churn. The two methods agree, which is the only reason to believe either of them.

Sensitivity: at 0.25% capture, ~14 customers and £157k; at 0.5%, £314k; at 1.0%, ~54 customers and £628k. One per cent of SAM by year three is achievable only with a second salesperson and a partner channel, and should be treated as the hiring trigger rather than the plan.

What this model gets wrong, stated plainly

1. Construction is understated, and probably badly. The IDBR counts enterprises by their PAYE and VAT-registered employees. UK construction has an unusually high share of self-employed and subcontracted labour, so a housebuilder or main contractor with 80 direct employees may have 400 people on its sites - inside the ICP, invisible to this count. The 935 construction enterprises are a floor, and the sector where the count least resembles the addressable workforce.
2. Section Q is overstated for this ICP. Division 86 (human health) contains NHS trusts, which appear in the 1,000+ band in numbers (290 of the 470) and are a public-sector procurement sale with a wholly different cycle - not the ICP the playbooks describe. Divisions 87 and 88, residential and non-residential social care, are the real target inside Q.
3. Enterprises, not sites. A 900-site pub group is one enterprise here. The Site plan at £180 exists precisely for the shapes this table cannot see, and it is excluded from the model entirely, so the model ignores a whole revenue line.
4. The top band is capped at £2,600 when Enterprise starts at £3,600 and 1,025 enterprises sit in an open-ended "1,000+". Understated.
5. List price, no discount, no churn, no ramp. Real ARR at any customer count will be lower than the tables above. Treat SOM as a target, not a forecast.
6. Sectors outside the six are excluded - agriculture, waste, utilities, education and retail all contain multilingual deskless workforces and none is counted here.

Net: the omissions run in both directions, but the construction undercount and the excluded Site plan are the two largest, and both push the same way. Treat £64.6m as a conservative SAM.

The line the model exists to produce

A £64.6m SAM does not need a second salesperson to prove itself, and 27 customers is not a market share worth defending - it is a beachhead. The useful output of this page is not the SOM. It is the threshold: at roughly 25-30 customers and £300k ARR, one founder is the constraint, not the market. That is the point at which to stop selling personally and start hiring - and it arrives, on these assumptions, in year three.

All figures verified 28 August 2026. ONS UK Business Counts is published annually each September; re-run the Nomis query when the 2026 edition lands and re-check the sponsor-register share against a fresh download.

UK market context: the multilingual workforce, and what HSE requires

This is the evidence base for the entire TeamTalk proposition. It is worth quoting verbatim in decks and on the site.

How multilingual UK construction actually is

The canonical academic reference, for anyone who wants one: Bust, Gibb & Pink, "Managing construction health and safety: Migrant workers and communicating safety messages," Applied Ergonomics, 2008.

What HSE actually says — and it is a gift

HSE, Employing migrant workers — help with language issues, updated 4 June 2024 (hse.gov.uk/migrantworkers/employer/help-with-language-issues.htm):

HSE's own listed options for effective communication include a bilingual employee interpreting, a professional accredited interpreter, a buddy system, non-verbal communication (video, audio, internationally recognised signs and symbols), simple clear English — and, explicitly, "use professional translation software or free online tools."

HSE's overview page for employers of migrant workers names construction among the high-risk industries and lists as elevated-risk factors that "language barriers limit their ability to communicate effectively with other workers and supervisors, making it difficult to understand training and instruction" and that employers "fail to check their skills for the work (including language skills) or fail to provide appropriate health and safety training and instruction."

Why this matters commercially. HSE puts the burden squarely on the employer to make information comprehensible, explicitly blesses translation software as a means, and explicitly endorses video. Nobody in the construction stack does both translation and proof. Procore has 13 UI locales and no acknowledgement tracking. Fieldwire has neither. HandsHQ and Sitepass/Quartex stop at a signature on a document. HammerTech does attendance and sign-off on toolbox talks but does not translate. The only vendor shipping both auto-translation and acknowledgement tracking is Blink — which has no construction layer at all: no RAMS, no inductions, no site model, no ACV-style commercial construct, and no video captions.

That intersection — translated broadcast + confirmed receipt + video captions + a construction-shaped price — is currently empty.

The residual research list that used to sit here is internal and lives in docs/gtm/verify-residual.md; nothing in this chapter should be pasted into a deck without checking the source in Chapter 14.