← Back to Sources and assumptions

TeamTalk — Commercial Research Inputs for the GTM Guide

Compiled 28 August 2026. All prices verified against primary vendor/regulator sources on that date unless marked ESTIMATE or UNVERIFIED.

FX used throughout: GBP/USD 1.3583, GBP/EUR 1.1666 (Frankfurter, 2026-08-28). USD costs are converted at this rate; re-run the model if sterling moves more than 5%.

Product assumed: Laravel 11 + Livewire 3, Reverb/Pusher realtime, MySQL, Redis, Forge/AWS eu-west-2. Integrations: Twilio Verify (phone-number login), Mux (video + captions), OpenAI (translation), VAPID web push, S3 (images).

Reading order for a founder short on time: §1.8 (the verification decision), §2.9 (the tier card), §3.3 (why this is a compliance purchase), §4.4 (the demo), §5.3 (what year one realistically looks like).

1. Unit cost model

1.1 Twilio — verification and SMS (UK)

Notes. A GSM-7 message is 160 characters per segment; any emoji or non-Latin character forces UCS-2 and drops the segment to 70 characters, roughly doubling cost. Keep OTP messages plain ASCII and short. Alphanumeric sender IDs work in the UK and cost the same as long codes; they raise deliverability and brand trust but cannot receive replies. UK carrier surcharges apply on top of the listed rate for some routes — treat £0.045/segment as the planning figure rather than £0.0412.

1.2 Mux — video

Mux restructured its pricing: encoding/input is now free, and there is a substantial free tier.

Important nuance on "free encoding": encoding is free on the Basic encoding tier only (on-demand, up to 4K). Plus is $0.025–$0.03125/min and Premium $0.0375–$0.047/min (Mux docs pricing). Basic is the right tier for site-manager phone video; do not let anyone switch the default. Premium also multiplies storage by 1.5×. One documented discrepancy: the pricing page shows live captions at $0.0024/min, the docs page at $0.024/min — verify before relying on live captions; it does not affect on-demand.

The number that matters: free auto-captions on demand. Captions are a genuine accessibility and comprehension feature on a noisy site and they cost nothing. Turn them on by default. The 100,000 free delivery minutes/month is account-wide, not per tenant — at the modelled usage (1.6 videos × 2 min × 70% watch rate) that covers roughly 45,000 workers before you pay a penny for delivery. Video is effectively free for the whole of year one and two.

1.3 OpenAI — translation

Current lineup and pricing (USD per 1M tokens), from OpenAI API pricing:

Batch API: 50% discount on input and output (cached-input rate unchanged). Translation of scheduled, non-urgent posts is a textbook batch workload; urgent safety posts are not.

Worked estimate: one 300-word post into 6 languages

Assumptions: English at ~1.33 tokens/word → 400 tokens of post; 150-token system/glossary prompt; output 1.25× input tokens (Polish, Romanian, Lithuanian and non-Latin scripts tokenise less efficiently than English); one API call per target language so each translation is independent and retryable.

Per call: ~550 input tokens, ~500 output tokens. Six calls: 3,300 input + 3,000 output tokens.

Recommendation: gpt-5-mini as the default, gpt-5-nano for low-stakes/social posts. At half a penny per post into six languages, translation cost is not a business input — a tenant sending 8 posts a month into 6 languages spends 4p a month. The constraint is quality and safety-critical accuracy, not price, so buy the better model. Do not build a caching or batching optimisation for translation; it will never pay for the engineering time.

Do spend the saved money on quality controls: a per-tenant terminology glossary (site names, trade terms, PPE names, "permit to work", "exclusion zone"), a mandatory back-translation check on any post flagged safety-critical, and a "show original" toggle so a worker can always see the English. Under MHSWR Reg 10 the duty is to provide comprehensible information (§3.3) — a bad translation is a compliance liability, not just a bad user experience.

1.4 Hosting — Laravel Forge + AWS eu-west-2 (London)

Forge

Forge is flat-rate; you still pay your cloud provider separately. Zero-downtime deployment is now included (Envoyer is no longer needed for a single-server deploy).

EC2 on-demand, eu-west-2, 730 hrs/month

Source: ec2.shop, eu-west-2, AWS on-demand list prices.

Graviton (t4g) is 20–27% cheaper than the equivalent t3 for identical vCPU/RAM. PHP 8.3, MySQL, Redis and Reverb all run natively on ARM64. Use t4g throughout — this is the single largest free saving in the stack. A 1-year no-upfront Reserved Instance or Compute Savings Plan takes another ~37% off; commit only once you have a paying customer.

Database: RDS MySQL vs self-hosted

eu-west-2 on-demand, MySQL, no license required:

Source: AWS Price List API, AmazonRDS/current/eu-west-2. RDS gp3 storage is $0.133/GB-month Single-AZ, $0.266 Multi-AZ (AWS Price List, eu-west-2) — so 100 GB is $13.30/month Single-AZ. EBS gp3 on EC2 is $0.0928/GB-month with 3,000 IOPS and 125 MB/s included.

Recommendation: RDS db.t4g.medium Single-AZ, with automated backups and a documented restore drill. It costs about £18/month more than self-hosting and buys automated backups, point-in-time recovery, patching and a one-click restore. That £18 is the cheapest insurance in the business, and — more commercially relevant — "managed database with point-in-time recovery" is an answer on every supplier security questionnaire you will be sent (§3.5). Move to Multi-AZ (+£39/month) only when you sign an SLA above 99.5%, not before. Self-hosting MySQL on the app server is defensible only pre-revenue.

Storage and CDN

Rough cost per 100 GB stored and served once: S3 storage £1.77/month + CloudFront egress ~£6.30 = ~£8/month per 100 GB, and the first 1 TB/month of CloudFront egress is free. Images at 400 KB each are irrelevant to the P&L: 10,000 workers each loading 3 images a month is 12 GB — inside the free tier.

Do this: serve all images and video posters through CloudFront with a long cache TTL, and store originals in S3 with a lifecycle rule to Infrequent Access after 90 days. Not for the money — for the page-load time on a 3G signal in a site cabin, which is what actually drives the activation metric in §4.9.

1.5 Realtime: Pusher vs self-hosted Reverb

Recommendation: Reverb, self-hosted, from day one. The workload is asymmetric and cheap — a site operative holds an open connection for seconds while reading a post, not all day, and the message volume (8 posts/month/tenant) is trivial. Reverb on the existing queue server costs nothing incremental, keeps every byte in eu-west-2, and removes a named US sub-processor from the DPA — which is worth more in a procurement review than it is in engineering time. Keep Pusher configured as a fallback driver so you can switch with one env var if Reverb becomes an ops problem.

Realtime pricing detail (for the Pusher/Reverb decision)

Source: Pusher Channels pricing.

The binding constraint is concurrent connections, not messages. A single 5,000-worker tenant would exhaust Pusher's $299 Business plan on connections alone. Reverb has no per-connection fee. Two operational facts to design around (Laravel Reverb docs): the default stream_select event loop caps at ~1,024 connections — install ext-uv before you exceed that; and raise ulimit -n and nginx worker_connections to 10,000. A single box tops out near 28,000 connections on ephemeral ports; beyond that, scale horizontally with REVERB_SCALING_ENABLED=true over Redis pub/sub.

1.6 Email, monitoring, domains

Recommendation: SES + Sentry Team + UptimeRobot Solo. Total £34/month. SES at 10p per 1,000 emails makes email genuinely free at this scale — you will send perhaps 2,000 transactional emails a month. Renew domains at Cloudflare Registrar (at-cost, no markup) rather than at a registrar that triples the price on renewal.

1.7 Alternative hosting comparators

Stay on Forge. It is $19/month against a stack you already know, and the alternatives save at most £10/month while costing you a migration.

1.8 The one cost decision that actually matters: verification policy

Everything in this model is rounding error except phone verification. At UK rates one Twilio Verify SMS costs 7.8p. At the brief's assumed rate of one verification per user per month, verification is 58–82% of total COGS at every tenant size — more than servers, translation, video, captions and storage combined.

Recommendation: policy C. Long-lived, device-bound refresh tokens; re-verify only on a new device, cleared browser storage, or a 180-day gap. This is simultaneously the cheapest and the best user experience — a site operative should never be asked for a code to read a safety briefing. Keep Twilio Verify rather than rolling your own: the $0.05 fee buys SMS-pumping fraud protection that a solo founder otherwise learns about via a £4,000 bill. Revisit policy D only above ~50,000 verifications/year, where the saving exceeds £1,500/yr.

Corollary for the price list. If a customer insists on SMS delivery of posts rather than push — some HSEQ leads will, for genuinely good reasons on sites with no wifi — that is 4.1p per message per worker. Eight posts a month to 500 workers is £165/month of pure pass-through against a £950/month subscription: 17% of revenue, gone. Either price SMS delivery as a metered add-on at 8p/message (≈2× cost, and the markup is defensible) or exclude it in the order form. Web push via VAPID costs nothing, which is precisely why it must be the default and why "does it work without an app?" is the demo's most important 30 seconds.

1.9 Worked cost-per-user-per-month model

Assumptions (state these on any slide that uses the output)

Fixed platform cost

Per-tenant variable cost (independent of headcount)

Per-user variable cost

Cost per user per month, by tenant size

Scenario C — recommended verification policy

Scenario A — one verification per user per month, as briefed

Scenario A costs 91–94% more per user than Scenario C for a worse login experience. That is the whole finding of this section.

Gross margin at the candidate prices from §2.9

Effective per-user prices in the recommended tier card: £3.19 (Company 501-band floor), £2.20 (Company ≤250), £1.90 (251–500 ceiling), £1.60 (501–1,000 ceiling), £1.30 (1,001–2,000 ceiling).

Scenario C

Scenario A

The honest caveat about "fixed cost per user"

The tables above spread fixed cost across 3,000 platform seats. Until you have those seats, the fixed cost lands on whoever is paying:

Your first pilot customer is gross-margin negative or marginal at any sane price. That is fine and expected — £161/month is not a business risk, it is a rounding error against a founder's time. Break-even on infrastructure is roughly 75 paid workers at £2.20, i.e. reached by the first real customer. From tenant #2 onward the model is above 94% gross margin at every size and price in the tier card.

Three conclusions for the GTM guide

1. Cost is not a pricing input for this product. COGS is 3–14 pence per user per month against prices of £1.30–£3.19. Price on the value of the compliance record and on what the market bears (§2), never on cost-plus.
2. The only variable worth engineering against is SMS. Verification policy and any SMS-delivery option are the entire cost story. Everything else — servers, translation, video, captions, storage, realtime — totals under £200/month at Stage A.
3. Free-tier headroom is enormous. Mux gives 100,000 delivery minutes/month free (≈45,000 workers at modelled usage) and free on-demand captions; CloudFront gives 1 TB/month; AWS gives 100 GB/month egress; VAPID push and Reverb are free. Do not architect around costs you will not incur for two years.

1.10 UK compliance, insurance and payments — the non-COGS opex

These do not belong in gross margin, but they are the costs that gate deals (§3.5) and they must be in the cash plan (§5.4).

Insurance limits are a sales input, not just an overhead. UK construction and FM supplier contracts routinely demand minimum cover levels; buy limits that clear the thresholds your target accounts ask for (see §3.5) rather than the cheapest policy, because raising a limit mid-procurement costs you three weeks.

Stripe UK

Two decisions this drives. (1) B2B cards are overwhelmingly commercial cards at 2.8% + 20p, not 1.5% — a £19,200 annual invoice on a company card costs £537.80 in fees. (2) Take annual contracts by Bacs Direct Debit or bank transfer, not card. Bacs is capped at £4.00 per collection, so the same £19,200 costs £4, saving £534 per customer per year. Offer card only for the £2,500 pilot fee, where speed matters more than the £70 of fees. For a plc that insists on invoice-and-BACS through its AP system, take that — it is free.

2. Pricing benchmarks and models

All competitor figures below were verified against vendor pricing pages on 28 August 2026. Three market facts reframe the competitive set:

2.1 Published per-seat pricing — frontline comms

Oneteam is the only vendor that publishes its hidden costs, and they are instructive: existing-integration setup €500 one-off, new HR/WFM integration build €2,500 one-off, white-labelled app €200/month per platform (iOS and Android charged separately).

2.2 Vendors with no public pricing — and what they actually charge

2.3 The adjacent threat: comms bundled into rota tools

Deputy's Messaging+ at ~£1.44/user/month is the cleanest available market price for "comms bolted onto a workforce system we already own". Expect a caterer or FM buyer to hold that number up. The answer is not price — it is that none of these produce an evidential briefing record or translate into a worker's language.

2.4 The "we have Teams" objection, priced

UK GBP, ex-VAT, annual commitment, en-GB Microsoft pages, 28 Aug 2026:

Viva Employee Communications & Communities requires a qualifying base licence, and M365 F1 qualifies. So the genuine Microsoft frontline-comms stack is:

That is above Blink Core (£2.95), above every Thrive tier, and above our own recommended Company-tier effective prices. Two consequences: (a) £3.84 is the real price ceiling for a frontline comms seat in the UK — above it you are being compared to a SKU the buyer's IT team already has on the price list; (b) the Teams objection is not a price objection, so never answer it on price. Answer it on: who is actually licensed today (usually a small fraction of site operatives), whether it needs an app install and a password (it does), whether it translates into the worker's language (it doesn't), and whether it produces a per-person evidential briefing record (it doesn't).

2.5 What UK construction and FM buyers already pay for software

Three things this proves for the pricing narrative:

1. UK construction and FM buyers pay £36–£45 per user per month without blinking — but only for supervisors and office staff, never for every operative. The per-seat ceiling is a function of who is licensed, not of sector willingness to pay.
2. HandsHQ, a successful UK construction SaaS, does not sell per seat at all — flat monthly tiers banded by headcount, effective rate falling 55% as the band widens.
3. Procore has explicitly rejected per-user pricing and sells that rejection. That is the strongest public evidence that per-seat is the wrong instrument in construction.

2.6 Pricing model prevalence and why per-frontline-user is resisted

Per-user is the default, but every vendor specialising in genuinely multi-site frontline operations has moved off it — YOOBIC to per-store, Ocasta to per-site-by-default, HandsHQ to headcount bands, Procore to volume. Four of those six sell into construction, retail or hospitality. That is the pattern to cite.

Why the resistance is real in these sectors specifically:

What buyers accept instead, with live proof points: per site (Ocasta, YOOBIC) · tiered headcount bands (Thrive, HandsHQ) · platform fee + low marginal rate (Connecteam) · unlimited users on a business-volume metric (Procore). No live proof point for per-active-user (MAU) pricing was found in this category — it sounds attractive but nobody in the market is doing it, and it makes the buyer's bill unforecastable, which is the objection you were trying to solve.

2.7 Minimum seats and floors — the gap we price into

The category splits into two postures with nothing in between: enterprise vendors gate at 250–1,000 seats and floor ACV at ~$20k; self-serve vendors have no floor and no compliance story. A UK housebuilder regional division with 400 site operatives and 30 office staff sits exactly in the gap and cannot buy the enterprise products at all. That is the market opening.

2.8 Discount and uplift norms (observed, not estimated)

Staffbase implementation at $15k–$75k can exceed a full year of licence fee against its own $28,544 median ACV. That is the single most exploitable weakness at the enterprise end of this category, and Ocasta is already exploiting it.

2.9 Recommended pricing for TeamTalk

Three tiers. All prices ex-VAT, GBP, 12-month minimum term, implementation £0. Monthly billing available at +15%.

Tier 1 — Site

For single-site or few-site operators: a small housebuilder, an independent caterer, one FM contract.

Why per-site: it is the unit the buyer already budgets in, it removes the "every new subbie costs me money" objection, and it has direct precedent — Ocasta Engage publishes per-site as its default model for location-based teams, and YOOBIC prices by store.

Tier 2 — Company (the tier that should win most deals)

Unlimited sites, banded by enrolled worker count. Re-band annually, not monthly — never mid-term.

Adds: divisions/regions hierarchy, role-based admin, scheduled + mandatory posts, compliance export API, monthly reach report, named onboarding.

Two things to watch with bands. First, the cliff: a customer at 251 workers pays £950 while one at 250 pays £550 — a 73% jump for one extra person. Hold the line in negotiation, but give yourself one legitimate release valve: quote the band that matches expected enrolment 12 months out, and re-band only at renewal. Second, the band floor is the price a buyer actually compares. At 251 workers the effective rate is £3.79/user/month, which sits just under the £3.84 Microsoft F1 + Viva reference price (§2.4). That is deliberate but tight — never let a band floor exceed £3.84, or you hand IT a reason to say "we already own that.

Tier 3 — Enterprise

2,000+ workers, or any buyer requiring SSO, a security review or a bespoke SLA.

Where these tiers sit against the market

The structural gap we are pricing into: enterprise frontline vendors gate at 250–1,000 seats and ~$20k ACV; self-serve vendors have no floor and no compliance story. A UK housebuilder regional division with 400 site operatives sits exactly in between and is structurally underserved.

Pricing rules to hold

Pilot offer structure

"60-Day Proof of Reach" — a paid, capped, time-boxed pilot with a pre-agreed conversion price.

Why paid, not free. A free pilot has no internal owner, no budget line and no deadline, so it drifts and dies at the first site-manager holiday. £2,500 is below almost every UK plc division's procurement threshold, which means a Divisional MD or H&S Director can approve it on a purchase order without a supplier-onboarding cycle — that is the real reason for the number. The refund guarantee removes the risk without removing the commitment.

Do not offer: open-ended trials, unlimited-site pilots, pilots without named success criteria, or pilots without a pre-agreed conversion price. Put the year-1 price on the pilot agreement itself so the renewal conversation is a signature, not a negotiation.

3. Selling into UK construction, catering and facilities

3.1 Who actually buys

There are four candidate budget owners and they behave very differently. The champion is almost always HSEQ; the economic buyer is almost always Operations. Internal Comms owns the budget only in large plcs with a real IC function, and IT is a gate, never a buyer.

Approval thresholds — ESTIMATE, reasoned. Published delegated-authority schedules are not disclosed by UK housebuilders, so these are reasoned planning figures to test on every discovery call, not sourced facts:

The structural fact — with an important exception. UK housebuilders and contract caterers are federated, and a regional division usually buys faster than group does. A divisional MD who has run a successful pilot is a far better advocate to group than you will ever be. Default: sell to a division, and never open cold at group HQ, where you get routed to a procurement queue with no champion and no proof.

⚠️ The exception that affects your named warm lead. Barratt Redrow centralises roughly 95% of procurement across around 160 group-level agreements. Divisional autonomy is a thin 5% sliver — enough for a small pilot on a purchase order, not enough for a rollout. Plan Barratt London as an explicitly two-stage sale: stage one is a divisional pilot funded from divisional discretionary spend, deliberately scoped and priced to sit inside that sliver; stage two is a group-level agreement that the divisional MD and HSEQ lead advocate for on the strength of the pilot's numbers. Do not price or forecast the divisional pilot as if it were the deal. Ask on the first call: "what can this division sign on its own, and what has to go to group?" — the answer varies enormously between housebuilders and is the single most useful piece of qualification you can get.

Contract catering is different in one important way: the caterer often operates on a client's site under a client contract. That means (a) the caterer's own head office buys for its whole estate, which is a bigger and slower deal; and (b) a site-level win can be blocked by the client's IT policy. Ask early: "whose site is it, and whose rules apply?"

3.2 Sales cycle lengths

These are reasoned estimates, flagged as such. No published UK-specific benchmark for this vertical and deal size was retrievable. Use them for pipeline planning and replace them with your own data after ten deals.

Planning consequences. A deal started in September that is not a small private builder will not produce revenue in the same financial year. Front-load the pipeline with small, fast buyers — they pay for the compliance certificates and produce the case studies that make the slow buyers possible. And add a full calendar quarter to every plc estimate for the InfoSec and legal queues, which are wait time rather than work time.

3.3 The compliance case: why proof-of-receipt is a statutory purchase, not a nice-to-have

This is the single most important section of the sales narrative. TeamTalk is not an internal-comms product being sold to construction; it is a records product that happens to look like an internal-comms product. Price it, position it and sell it accordingly.

The statutory chain

The language duty — the strongest, least-contested wedge

HSE's own guidance to employers of migrant workers identifies that "language barriers limit their ability to communicate effectively with other workers and supervisors", that workers may not "understand all of the risks" or have only "limited knowledge of the British health and safety system", and that employers must provide "appropriate health and safety training and instruction" and verify workers have the necessary skills including language proficiency. HSE itself publishes safety material in 19 languages including Polish, Portuguese, Romanian, Lithuanian, Latvian, Slovak, Czech and Albanian — which is, in effect, HSE conceding that English-only communication does not discharge the duty on a multinational site.

Sales use: "Reg 10 says the information has to be comprehensible. HSE publishes its own guidance in 19 languages. If you brief a Romanian steel fixer in English and he signs a sheet, what exactly have you evidenced?"

What "proof of receipt" is worth in an enforcement context

Enforcement and civil claims turn on evidence, and the evidence in this sector is famously bad: a paper toolbox-talk sheet with a scrawled signature, no record of what was actually said, no record of comprehension, and frequently no record at all for subcontract operatives who were on site that week. TeamTalk replaces that with a per-person, per-message, timestamped record that includes the language the worker actually read it in and an affirmative confirmation action.

The five properties that make the record defensible — build and demo all five:

1. Immutable — confirmations cannot be edited or backdated by a site manager
2. Per-person — named individual, not "the gang"
3. Timestamped to the second, server-side, with the delivery time as well as the confirmation time
4. Content-versioned — the record links to the exact text/video the worker saw, not to the current version
5. Exportable — PDF/CSV a solicitor, insurer or HSE inspector can read without access to the system

Building Safety Act 2022 / golden thread: for higher-risk buildings (18m+ or 7+ storeys residential), duty holders must maintain the "golden thread" of building information — accurate, up-to-date, accessible information about the building and how it is being managed. Where a comms record evidences that a safety-critical change was briefed to the people doing the work, it is golden-thread-adjacent. Do not overclaim this — position it as "your golden thread covers the building; this covers the people", and let the buyer make the leap.

3.4 The "WhatsApp problem" — the four arguments that land

The provision that turns this from a comms pitch into a compliance pitch

HSWA 1974 s.40 reverses the burden of proof. In a prosecution for failing to do something "so far as is reasonably practicable", "it shall be for the accused to prove... that it was not practicable or not reasonably practicable to do more than was in fact done."

Say this sentence in every discovery call: "In an HSE prosecution you don't have to be proved to have failed. You have to prove you did enough. What's your evidence?" The regulations say what must be communicated; s.40 makes it your client's job to prove they communicated it. A per-person, timestamped, language-stamped confirmation record is exactly that proof, and a paper sheet with a scrawled signature is a much weaker version of it.

MHSWR 1999 reg 12 extends the information duty to the employees of outside undertakings working in the host's premises — the direct statutory answer to "subcontractors aren't our employees."

A dated reason to buy before October 2026

The Employment Rights Act 2025 brings the strengthened harassment duties — including the "all reasonable steps" standard and third-party harassment liability — into force in October 2026 (per the Government's Implementing the Plan to Make Work Pay roadmap, updated 25 August 2026). An employer that cannot evidence what it communicated to its workforce, in a language they understood, is in a materially worse position on "all reasonable steps" from that date. This is a real, dated, externally-imposed deadline — the most useful thing a first-time seller can have. Verify the commencement date before putting it in writing to a customer; commencement dates move.

Primary sources for every provision cited above: HSWA 1974 s.2 · HSWA 1974 s.40 · MHSWR 1999 reg 10 · MHSWR 1999 reg 12 · CDM 2015 reg 13 · CDM 2015 reg 15 · HSE migrant workers guidance · HSE employer duties for migrant workers

The "this is live right now" hook

The Independent Review of Non-corporate Communications Channels in Government was announced on 2 July 2026 (led by Professor Sir Anthony Finkelstein) and expressly covers disappearing messages. It is weeks old and it is the sharpest available proof that this is a current governance issue rather than a vendor talking point. The Cabinet Office's NCCC guidance (30 March 2023) already tells departments at ¶37 to "ensure that NCCCs are not routinely used to process personal information", and at ¶27 sets out offboarding obligations.

⚠️ Get the ICO citation right — this is easy to get wrong and embarrassing to be corrected on. The ICO report on private messaging at the Department of Health and Social Care is Behind the screens, published 11 July 2022 by Commissioner John Edwards. DHSC received a reprimand, not a fine, for breaches of UK GDPR Articles 5(1)(e), 5(1)(f), 25 and 32. There is no 2024 ICO report on this. (PDF)

How to deploy this in a meeting: never attack WhatsApp as a product. Say: "Keep WhatsApp for the banter. Move anything you'd have to defend onto a system that produces a record." That framing wins; "WhatsApp is dangerous" does not.

3.5 Procurement gates — what will actually be asked for

Prepare all of this before the first plc conversation. Every item you cannot produce on the day adds one to three weeks of elapsed calendar time, and elapsed time is what kills first-time-founder pipelines.

The single highest-leverage preparation: a one-page security summary plus a pre-filled InfoSec questionnaire, sent unprompted immediately after the demo. It converts a six-week IT queue into a two-week one, and it makes a solo founder look like a company.

3.6 Seasonality and budget cycles

The two verticals are in opposite cycles — reprioritise accordingly

This is the most important strategic finding in this section. The brief assumes housebuilders are the primary vertical. On current data, contract catering is the easier sale in the next 12 months — growing, mobilising new contracts, and structurally more multilingual than construction. Keep the housebuilder warm leads (they are warm, and a Barratt logo is worth a great deal), but weight the outbound list toward catering and FM. The 40/40/20 housebuilding/catering/FM split used in §4.2 is aligned to where budget actually exists in 2026–27.

Construction shutdowns — verified dates

Source: Build UK / CIJC Annual Holidays 2026.

The selling calendar

Financial year ends — verified

Ask "when does your financial year start?" on every discovery call and diary the budget conversation for six weeks before it.

Contract catering runs on client contract cycles, not calendar years

Mobilisation is the single best moment to sell into catering. Ask every caterer: "what are you mobilising in the next six months?"

Live right now: education catering re-tenders are being decided in August 2026 with January–April 2027 starts, typically on a 3+1+1 shape. Also note CH&Co has been Compass-owned since 30 April 2024 (brand retained) — do not pitch it as an independent, and Crown Commercial Service is now the Government Commercial Agency (gca.gov.uk).

⚠️ Mitie is in an offer period — a recommended cash offer from OCS announced 21 July 2026. Discretionary software spend freezes during an offer period. Deprioritise Mitie until completion.

3.7 What a housebuilder regional-division pilot looks like

Scale of the opportunity — verified. A regional division of a volume housebuilder runs 9–14 active sites (verified across five UK housebuilders). A single housing development site has tens of operatives on site on a given day, the large majority of whom are subcontract, not direct employees — UK construction has an unusually high share of self-employed and subcontracted labour, and a substantial share of the workforce, particularly in London and the South East, does not have English as a first language. (The workforce percentages are directional; verify current ONS and CITB figures before putting a specific number on a slide.)

Scope the pilot at 3 sites, not 8–12. A whole division (9–14 sites) is the right commercial target and is what the research supports as a coherent unit — but it is a rollout, not a pilot, and one founder cannot support it while also selling. Pilot 3 sites, price the division. One site gives you an anecdote with no statistical comfort; three gives you a spread and a defensible number; the remaining sites are the expansion you have already priced in the pilot agreement.

The success-metric anchor to use. Barratt publishes a 92% mandatory-training-completion KPI. Where a prospect has an equivalent published number, set the pilot's confirmed-reach target against their own public commitment rather than against a target you invented. "You publish 92%; can you evidence it per person, per briefing, in the language they read it in?" is a far stronger frame than any benchmark you bring.

Two failure modes to design against. First, the pilot dies from the manager end, not the worker end — if nobody posts, workers stop opening. Chase posters weekly, by phone. Second, subcontractor enrolment is where the numbers go wrong — the site manager has a list of directly employed operatives and an incomplete one of everyone else. Get the subcontractor list in week one and treat it as the critical path.

4. Founder-led sales mechanics

4.1 Outbound benchmarks

Caveat first. These are real, sourced figures, but none are specific to UK construction, catering or FM — no such dataset exists publicly. The closest proxy is Lavender's Operations function data, which matters because your buyers are Operations and HSEQ, not IT. Saleshandy also flags Real Estate among the lowest-replying industries, which is the nearest adjacency to construction. Plan at the pessimistic end and replace every figure with your own after ten deals.

Cold email

The single biggest lever is list size, not volume:

This vindicates the 100-account list in §4.2 and is the most actionable finding in this section. A hand-built list of 100 named UK contractors, worked properly, should materially outperform the 3.7% platform average. Also note: for companies of 50–500 employees, Manager/Director titles reply best — which is exactly the HSEQ lead and the divisional Ops Director.

Sequence design

Two-contact multi-threading is already built into the cadence in §4.2 — the +93% figure is why it is not optional.

Cold calling — and the UK is the best market in the world for it

Actual SDR activity levels (Bridge Group 2025, 351 B2B companies): 44 calls + 41 emails + 19 LinkedIn touches + 8 texts = 112 activities/day for 4.1 quality conversations/day. Phone-centric teams get 4.6 quality conversations/day vs 3.4 for email-centric. Only 60% of SDRs hit quota — a record low. Calibrate your own expectations against that: a full-time specialist manages four real conversations a day.

LinkedIn

Conversion through the funnel

Gartner's 6–10 buying-group figure matches the 4–7 stakeholders you should expect at a plc division (§3.2). The referral data is the strongest argument in this report for exhausting your three warm leads before scaling outbound — and for asking every pilot customer for two introductions at signature.

Paid vs free pilots — the numbers behind §2.9

A 5–8× spread between committed and uncommitted trials is the entire quantitative case for the paid pilot in §2.9. One adjustment worth making to that design: SaaStr recommends 60 days with a 30-day review rather than 90. Consider shortening the pilot to 60 days — it forces the decision inside one budget quarter and reduces the chance of a summer or Christmas shutdown swallowing the middle of it. Once you have two or three reference customers, SaaStr's stronger recommendation is to replace pilots entirely with a 60-day opt-out annual contract, which they report churns less than paid pilots.

Founder-led sales

No published benchmark exists for "conversations needed to reach your first 10 customers" — SaaStr explicitly declines to give one. Derived estimate from the figures above: at 2–4% cold reply, ~30–50% of positive replies becoming held meetings, and a first-time founder's realistic 15–25% meeting-to-customer rate in a new category, expect roughly 1,000–2,000 targeted cold touches per closed customer, and 60–100 discovery calls to land 10 customers — falling to perhaps a third of that via warm introduction. This is consistent with the 1,600-touch plan in §4.2 and the base case in §5.3.

The only outbound number you control is activity. If your activity is materially below 1,600 touches over four months, no conversion-rate improvement will save the year.

4.2 The cadence to actually run

Target list: 100 named accounts, not 1,000. In UK construction and catering the addressable set is small and everyone knows everyone; a sloppy mass send burns the market permanently. Split 40 housebuilders/main contractors, 40 contract caterers and hospitality operators, 20 FM contractors — weighted toward catering because it is growing 9.3% while private housing output is forecast down 10% in 2026 (§3.6). Target the regional/divisional office, never group HQ first (with the Barratt exception in §3.1).

Two named contacts per account: the HSEQ/H&S lead (champion) and the Ops or Divisional Director (budget). Do not start with IT.

Volume discipline: ≤30 cold sends per day from a warmed domain. 100 accounts × 2 contacts × 8 touches = 1,600 touches over ~4 months. That is roughly 20 touches per working day — one hour a day, done properly, forever.

Warm-first rule: work the 3 warm leads to a decision before the outbound engine matters. Warm intros convert at a large multiple of cold across every published benchmark; three warm leads worked hard are worth more than the first 60 cold accounts. Ask each of the three, at the pilot signature, for two introductions — that is where deals 4, 5 and 6 come from, not from the cold list.

UK legal position — read this before the first send

(Practical summary, not legal advice. Primary sources: PECR 2003 regs 2, 21, 22, 23, 26; UK GDPR Arts 6, 14, 21; the ICO's "Direct marketing: a guide", "Guidance on direct marketing using electronic mail", "Guidance on direct marketing using live calls" and "Business to business marketing".)

Why B2B cold email is lawful — and exactly where it stops being lawful. PECR reg 22(1) applies only "to the transmission of unsolicited communications by means of electronic mail to individual subscribers". Corporate subscribers are outside it entirely.

This matters directly in your sectors. Main contractors, housebuilders, caterers and FM firms are overwhelmingly Ltd/PLC/LLP. But small groundworks and trades subcontractors, and many independent catering operations, are sole traders or ordinary partnerships. Segment your list against Companies House before sending — an active company number is the clean discriminator, and the entire legal defence rests on it.

Reg 23 applies to everyone, corporate included. Every marketing email must: not disguise or conceal the sender's identity; provide a valid address for opt-out; and be identifiable as a commercial communication as soon as received. Add your registered company name, number, registered office and VAT number (E-Commerce Regs 2002 reg 6).

UK GDPR sits on top and is where founders actually get caught.

• Lawful basis: legitimate interests, with a documented Legitimate Interests Assessment. The Data (Use and Access) Act 2025 now expressly names direct marketing as a possible legitimate interest, but the balancing test still applies — it is not one of the new "recognised legitimate interests". The ICO publishes a sample LIA template.
• Article 14 is the trap. You did not collect the data from the individual, so you owe them privacy information within one month of collection, or at the latest at your first communication — and you must tell them the source you got it from. The clock runs from collection, not from sending. Scrape in January and send in April and you were already non-compliant in February. Collect and contact in the same month.
• Article 21(4): the right to object must be brought to their attention at the latest at first contact, clearly and separately from any other information.

Cold calling is the asymmetry that catches people out. PECR reg 21 bans marketing calls to any number on the TPS or CTPS registers — and unlike email, this applies equally to corporate subscribers. A limited company you may lawfully email may be one you may not lawfully call.

• Screen every number against TPS and CTPS, and re-screen every 28 days. A registration becomes active after 28 days, so a stale screen lets new registrations through. The duty is yours — buying a "TPS-screened" list does not transfer liability, because you instigated the call.
• Screening is cheap: £11.25 + VAT for 1,000 numbers, £35 for 5,000 (TPS Services). Registering your own numbers on CTPS is free.
• Reg 21(A1) requires you to present a Calling Line Identity that is a number someone can actually call back. Withheld or unreachable CLI is a standalone breach the ICO has fined for.

⚠️ Penalties changed on 5 February 2026. The widely-quoted £500,000 PECR maximum is out of date — and the ICO's own guidance pages have not all caught up. DUAA 2025 s.115 replaces PECR's penalty regime with UK GDPR levels: marketing breaches now carry up to £8.7m or 2% of worldwide turnover, and failure to comply with an information or enforcement notice up to £17.5m or 4%. (The common shorthand "PECR fines are now £17.5m" is imprecise — 2% is the marketing-breach maximum. Confirm with counsel before relying on this in writing.)

LinkedIn: two separate risks. (1) A pitch in a LinkedIn DM or InMail is electronic mail under PECR — the ICO's guidance expressly includes private social-media messages and in-app messages. Whether an employee's personal LinkedIn account is an "individual subscriber" is unresolved, so keep the first touch non-promotional (connection request with genuine context, no pitch) and move the actual offer to email or phone. (2) Separately, LinkedIn's User Agreement §8.2.2 and §8.2.13 prohibit scrapers, bots and automated messaging outright. Every popular automation tool breaches this on its face. That is contract law, not data protection — the ICO will not fine you, but LinkedIn can permanently terminate the account your entire pipeline lives in, and it takes your Sales Navigator subscription with it. Do it manually.

The suppression list is your single most important control. Never delete an opt-out — suppress it, or you will re-import and re-offend next quarter. Keep one suppression list across all channels: an email opt-out must also stop the call and the LinkedIn message (Art 21 is channel-agnostic). Act within 48 hours.

Records to keep: a dated LIA per campaign; a data provenance log (source, URL, date, method — Art 14 makes you disclose it); evidence of the Art 14 notification and its timing; the corporate/individual classification with the Companies House number; dated TPS/CTPS screening receipts; and a sent-message archive. Also: pay the ICO data protection fee (Tier 1, £52 — see §1.10).

4.3 Discovery call script outline (25 minutes)

Goal of call 1: qualify, quantify the pain, book the demo with the right people in the room. Do not demo on call 1 unless they insist.

Disqualify fast if: fewer than ~60 frontline workers; no named safety or ops owner; "we're happy with WhatsApp" from the actual budget holder (not from IT); a group-level IT freeze; or they want you to also do scheduling, T&A and payroll.

Three questions that predict a close (ask all of them):

1. "What do you hand the HSE?" — if they have no answer, you have a compliance purchase, not a nice-to-have.
2. "How many of your operatives don't read English comfortably?" — if >15%, translation is the wedge.
3. "Who chairs your monthly H&S review?" — that person is your economic buyer.

4.4 Demo flow (30 minutes, in this order)

Demo the proof, not the posting. Buyers assume posting works; they do not believe the receipt.

Demo rules: use their real site names and their real language mix in the demo data; never demo on a laptop simulator when a real phone is available; never show settings screens; if asked "can it do X", answer yes/no/roadmap-with-a-date and move on.

4.5 Objection handling

4.6 CRM and stack for a solo founder

Keep it to four tools. Do not buy a sales engagement platform in year one.

Current prices for the shortlist (verified 28 Aug 2026):

Recommended solo stack: HubSpot Free + Smartlead Base + Sales Navigator Core + Lusha Starter + TPS screening ≈ £160–200/month. Upgrade only when the free CRM's 1,000-contact or 2,000-send ceiling actually binds.

On Notion-as-CRM: don't. It has no email sync, no call logging, no dedupe, no enrichment and no email tracking — every interaction is manual entry, which is exactly the discipline a founder in the field drops first. HubSpot Free does strictly more, at the same price of £0.

Non-negotiable hygiene: SPF, DKIM and DMARC on the sending domain before the first send (mandatory under the Google/Yahoo bulk-sender rules since Feb 2024, with Gmail escalating to permanent rejections from Nov 2025); keep your spam rate below 0.1% and never above 0.3%; a real signature with a UK registered address and company number; a working one-click unsubscribe (RFC 8058).

4.7 Paperwork essentials

Order form (2 pages, must contain):

1. Customer legal entity + company number, and the specific divisions/sites in scope
2. Subscription tier, unit of charge (per site / per band), quantity, and the definition of the counted unit
3. Term start, initial term, renewal mechanic, notice period to terminate (30–60 days for a first customer, not 90)
4. Price, currency, VAT stated separately, annual uplift clause (CPI or a stated cap)
5. Invoicing schedule and payment terms — ask for 30 days, expect 45–60 from a plc
6. Named customer project owner and named technical contact
7. Onboarding scope: number of sites, training sessions included, data import
8. Signature blocks + a line incorporating your MSA and DPA by reference and by URL

Also needed before the first enterprise call:

SLA — what the market actually commits to

(These figures were not fetched directly — verify each on the vendor's legal page before quoting one to a buyer.)

What to offer: 99.5% monthly, with credits at 10% / 25% / 50% of the monthly fee at <99.5% / <99.0% / <95.0%, capped at 100% of one month's fee. Measure monthly, not annually — a 99.9% annual commitment permits 8.76 hours of downtime in a single outage; monthly caps it at ~43 minutes. Service credits must be the sole and exclusive remedy — that clause is the entire point of publishing an SLA. Claims must be made by the customer within 30 days. Salesforce publishing no SLA at all is useful precedent when a buyer pushes for 99.99%.

Exclusions you must have, because this product depends on things you do not control: scheduled maintenance in a published window announced 48 hours ahead; emergency maintenance; force majeure; the customer's own network, devices and site connectivity; third-party push notification delivery (APNs/FCM), SMS gateways and app-store availability; beta features; suspension for non-payment; customer misconfiguration. Measure "platform availability", never "message delivered to the worker's handset" — you cannot control mobile signal in a site cabin.

Support targets: P1 (service down) 2 business hours, P2 (major feature broken) 1 business day, P3 4 business days, 08:00–18:00 UK Mon–Fri. Do not offer 24/7 you cannot staff. An SLA you breach quarterly is worse than a modest one you always hold.

Limitation of liability

Standard position: 100% of fees paid in the preceding 12 months (125% is a common landing spot after negotiation). Uncapped by law and non-negotiable in the UK: death or personal injury caused by negligence (UCTA 1977 s.2(1)) and fraud. Exclude indirect and consequential loss, and list the heads explicitly — loss of profit, revenue, business, goodwill, anticipated savings and data — because English courts read generic "consequential loss" exclusions narrowly.

⚠️ The UCTA trap. Because you contract on your written standard terms, UCTA 1977 s.3 subjects your limitation clauses to a reasonableness test. A cap set implausibly low relative to the fees and the risk can be struck out entirely, leaving you uncapped. A cap at 100% of annual fees on a £19,200 contract is defensible; a flat £1,000 cap on the same contract may not be. Do not get clever here.

Enterprise buyers will push for uncapped liability on data breach, confidentiality, IP indemnity and gross negligence. Concede a bounded super-cap (e.g. 2× annual fees) rather than anything uncapped — and make sure your cyber and PI limits exceed the super-cap before you agree to it (§1.10, §3.5).

4.8 First 30 days of onboarding

4.9 Metrics to track

Retention targets — use the band that matches your ACV

⚠️ I previously suggested >110% NRR and <10% annual logo churn. That is the enterprise headline number, not yours. At £5k–£50k ACV into construction, catering and FM — high workforce churn, subcontract labour, project-based customers — you have an SMB retention profile. Anchor the plan to GRR ~85% and NRR ~100–105%, and treat 111% as the top-quartile stretch. Benchmarkit also finds deals above $100k are less CAC-efficient than the $10k–$100k range, which is a useful counterweight to the pressure to chase group-level deals before you are ready.

One genuinely encouraging structural point: at your ACV on invoice terms, involuntary (payment-failure) churn is near zero — so essentially 100% of churn is an adoption and value problem, which is the kind you can fix.

Reality-check the activation target

The vendor case studies are hand-picked flagship logos and never define "adoption". A realistic planning range for monthly active use is 40–70% of eligible workers; 75–95% is achievable only where the app gates something the worker cannot avoid — a payslip, a rota, a timesheet, a safety sign-off. In UK construction and catering specifically — high churn, subcontract and agency labour, shared devices — plan at the low end and make adoption stick by tying the app to a transaction the worker actually needs.

This does not change the §2.9 pilot criteria, because activation (has logged in at least once within 14 days) is a much lower bar than monthly active. But it does mean: hold ≥70% activation and ≥85% confirmed-receipt as pilot gates, and set the weekly active expectation at 40–55%, not the ≥50% floor implied above. Report both, and never let a customer conflate them.

The two arguments with budget behind them

Forrester's Total Economic Impact study of Firstup (August 2026; composite org of 10,000 employees, 65% frontline) puts three-year ROI at 398% with payback under six months. It is a vendor-commissioned study with a constructed composite, so do not quote the ROI figure — but the structure of the value case is the useful artefact:

Build the business case on attrition and safety, not productivity hours. Those are the two lines a UK construction or FM finance director already has a number for. And open with Simpplr's finding that only 1 in 4 organisations reach 75% of their people — it is a better first line than any feature.

5. Twelve-month plan skeleton

5.1 Lead times to plan around

Planning rule: anything that requires a plc's legal or InfoSec function takes a calendar quarter regardless of how simple it is. Start Cyber Essentials and the DPA/security pack in month 1, before you need them — they are the two things that will otherwise stall your first real deal for six weeks.

5.2 Month-by-month plan (Sep 2026 → Aug 2027)

Assumes one founder, selling and building. Sales days are the constraint, not engineering days — block three days a week for selling from month 1 and defend them.

Seasonality rules baked into the above: no decisions between 18 December and 4 January (verified CIJC shutdown). There is no longer an industry-wide summer shutdown — ask each account for its employer-designated close-down. February–April and June–July are the productive selling windows. Housebuilder budget years vary by company, so ask every prospect "when does your financial year start?" on the discovery call and diary the follow-up for six weeks before it.

5.3 Revenue scenarios

Funnel assumptions, from 3 warm leads plus outbound to 100 named targets over 12 months. Conversion rates are reasoned estimates for a first-time seller in a slow vertical — treat them as planning figures, not benchmarks, and replace them with your own after 10 deals.

Resulting revenue

Read this honestly: the conservative case is the modal outcome for a first-time seller with no reference customers, and it does not pay a salary. The base case is achievable only if all three warm leads are worked hard in the first 90 days, because outbound into UK construction from a cold start with no logo is slow.

5.4 Cash needs (12 months, one founder)

Conclusion on cash: budget £50k–£65k of runway for the first 12 months. The business is not capital-intensive — infrastructure is under £200/month and gross margin is above 95% — the entire cash requirement is founder time and the calendar length of UK construction procurement. Do not spend on ISO 27001 or Cyber Essentials Plus until a specific deal is gated on it in writing.

6. Confidence, gaps and what to re-check

6.1 Corrections applied late in this research

Four things in the original brief or in my own first draft were wrong and have been fixed above. They are listed here because each would have been embarrassing in front of a customer:

6.2 Verified vs estimated

6.3 Manual checks before any of this reaches a customer

1. Payment days for each named target at check-payment-practices.service.gov.uk.
2. Cyber Essentials tier price in GBP on iasme.co.uk in a UK browser — and confirm you are preparing against the Danzell question set (in force 27 Apr 2026), not Willow.
3. DPF "Active" status by legal entity for Twilio Inc. and Mux Inc. before relying on the adequacy route.
4. ERA 2025 commencement date — the October 2026 date comes from a roadmap; commencement dates move.
5. ICO data protection fee for 2026/27 — SI 2026/386 amends the regulations and is listed as not yet in force.
6. Mux live-caption pricing — the pricing page says $0.0024/min, the docs say $0.024/min.
7. UK WhatsApp Business per-message rates — Meta moved the UK to standalone country pricing with a higher marketing rate on 1 July 2026; any earlier figure is stale.
8. Pipedrive and Apollo list prices — both block automated fetch; check in a browser.
9. Insurance minimums — read the actual supplier terms of your first three targets rather than budgeting from CCS RM6263.
10. Your own funnel numbers — every conversion rate in §4.1 and §5.3 is scaffolding. Replace them after ten deals.

6.4 Still unverified after all four research streams

Achilles Building Confidence pricing; ISO 27001 certification costs from a UK certification body; a £-value for the UK contract catering market; any UK employment case drawing an adverse inference from un-preserved WhatsApp messages; and Covid Inquiry Module 2 content (the Inquiry's own domains were unreachable — do not claim the government response contains a messaging recommendation, because it does not).