We don't sell seats.
We license validated sites.

What this document is

One source of truth for how Qualitum is priced, sold, delivered and split with partners. It replaces every earlier pricing sheet, deck slide and verbal quote. Where a number elsewhere contradicts a number here, this document wins until it is formally revised.

How to use it

It is built as a guide, not a report. Pick a chapter from the contents on the left, read it end to end, then press Next at the bottom to continue. Chapters stand alone — you can read the partner model without reading the market analysis — but they are ordered so that reading straight through works as a course, and the deck and narration script are built from the same order.

What changed from earlier versions

• Per-site, not per-seat. Users are unlimited. The licence is bound to a site or an enterprise programme, and the fair-use envelope is counted in validation packages per year.
• Qualitum always implements. Partners sell and support; they do not deliver. The implementation fee is split with them 50/50 anyway.
• The pilot is paid. The free step is a ninety-minute teardown. The Proof of Value costs €20,000 and is credited against the first licence year.
• Delivery list prices went up to fund that 50/50 split. Anything quoted from an older sheet is re-quoted before signature.
• Every open question is closed. Chapter 16 records the seven decisions and why the alternative lost.

How the whole thing works

Before any of the detail: this is the entire commercial model, in the order money actually moves. If you can hold this chapter in your head you can hold a first call without notes.

The four things a client actually buys

How a deal moves

Nothing here is invented. It is the standard motion for regulated enterprise software, compressed. Each step is small enough that the buyer can say yes without a committee, and each one produces the evidence needed for the next.

Who does what

Where the money goes

One worked example, so the shape is concrete. A mid-size CDMO buys one Standard site through a Certified Partner: Proof of Value, licence, standard deployment, validation pack and light tailoring.

In year one the partner earns two thirds of what we do, because they carried the sale and we carried the delivery cost. From year two the position inverts permanently: the licence renews, the project fees do not, and our share is almost all margin. Say that to partners out loud — it is the deal, and hiding it makes them suspicious of the renewal terms.

What is deliberately outside the licence

• Infrastructure. Their cloud, their account, their bill. We deploy into it.
• Model tokens. Their model contract in the default deployment. Where we host, an envelope is included and overage is published. Either way it is single-digit percent of what they pay us — see chapter 11.
• Their own validated state. We supply the evidence; they remain accountable for the decision to use it.
• Custom code. Configuration is theirs to change freely. Code is not. That boundary is what keeps the licence from leaking — chapter 10.

The vocabulary

Every term used in the rest of this guide, in one place. If a partner cannot use these words correctly on a first call, they are not ready to hold one.

A category that just got validated by a $172B buyer

Digital validation stopped being an emerging category in June 2026. Thoma Bravo agreed to take Kneat — the category leader — private for approximately C$650M in cash, at a 40% premium to its unaffected share price. Kneat's ARR at 31 March 2026 was C$76.4M, so the transaction cleared at roughly 8.5× ARR for a business growing ARR 20% year over year at 78% gross margin.

What the numbers tell us about buyer behaviour

• The land-and-expand unit is the site, not the seat. Kneat's investor materials describe site-by-site expansion; MSD runs 27 sites; a European CDMO signed in April 2026 across two dozen production sites starting with equipment validation. Buyers approve one site, prove it, then roll out. Our licence unit should match how they buy.
• Deal sizes are large and the floor is high. Mid-six to seven-figure annual deals are reported for top-20 pharma and major CDMOs; smaller for biotech. Veeva benchmarks: ~$325k annual licensing plus ~$300k implementation for a 500-person biotech; ~$2.15M licensing plus $1.2M+ implementation at 5,000+ employees.
• Services are a second market, not a rounding error. The CSV services market is forecast at ~$4.5B in 2026 growing to $7.4B by 2032. A single GMP system validation project runs $25k–$150k depending on complexity and GAMP category. This is the pool our implementers already live in.
• Incumbent implementation is slow and expensive. Kneat publishes 8–12 weeks average time to live. ValGenesis realistically runs 3–6 months for a single-site biotech and 9–18 months for a mid-cap multi-site rollout. Veeva makes implementation mandatory — there is no self-service path in a regulated environment.
• Change-of-control risk is live. Buyers signing multi-year Kneat deals in 2026 are being advised to price in the ownership change. Private-equity ownership historically means price discipline and upsell pressure. That is a real, current, sayable objection for our sellers.

Per-seat, per-module, and a services bill that rivals the licence

Three structural weaknesses we can price against

Where pricing is heading in 2026

The wider market has already moved off pure seats. Seat-based pricing fell from 21% to 15% of SaaS companies in twelve months while hybrid models rose from 27% to 41%. Hybrid — a predictable base fee with a variable or capped usage layer — is now the dominant enterprise standard for AI products, because pure outcome pricing requires attribution nobody can agree on and pure consumption pricing makes budgets unforecastable. Our model is deliberately hybrid: a fixed licence per site, with a fair-use envelope rather than a meter.

The Qualitum Site Licence

One annual fee per licensed site. Unlimited users. All agents included. The client configures, maintains and owns their data; Qualitum owns the platform and keeps it in a validated, regulator-current state.

Defining "site" — this is the clause that protects the model

A loose definition is how a global pharma turns eleven sites into one licence. The contract defines a Licensed Scope as either:

1. A Licensed Site — one physical manufacturing, laboratory or packaging facility operated by the client at a single address, holding its own GMP authorisation, site registration or equivalent establishment identifier. Buildings within one contiguous campus under one authorisation count as one site.
2. A Licensed Programme — one enterprise system validation programme not bound to a physical site (an ERP, MES or LIMS rollout, a CSV/CSA programme across an IT estate). Priced on the same bands.

The second definition matters more than it looks. A large share of validation spend is enterprise IT, not equipment — one publicly reported Kneat customer used the platform for all CSV documentation on an SAP S/4HANA deployment and reported $81M in savings. A pure per-site licence would have no way to price that deal.

Why unlimited users is the right trade

The economic argument is that we should never be paid less because our agents replaced a reviewer. The commercial argument is stronger: seat counting is the single biggest source of friction in a validation rollout, because QA reviewers, engineering, external CQV contractors and equipment vendors all touch the workflow occasionally. Every one of those is a seat negotiation in a Kneat or Veeva deal. Removing it makes our deployment faster than theirs before we write a line of configuration — and speed to first value is the thing we are actually selling.

Recommended end-client pricing

All figures are recommended list, in EUR, ex VAT. The annual licence is quoted to the client as one number; internally it comprises the operating licence (80%) and Assurance & Upgrades (20%). Partners earn the shares set out in §09 and are free to sell above list — the uplift is theirs.

Annual licence

What the limits actually mean

Two numbers in the licence look like restrictions and get asked about on every call. Both are deliberately generous and both are reconciled once a year, never metered mid-term. Here is exactly what they count.

Project and service fees

How this compares

Conversions use approximate mid-2026 rates and are directional. We are consistently positioned below the incumbents — never above — while carrying no per-seat ceiling.

They own the outcome. We own the machine.

The hardest question a partner will get is: "If we own our data and we can configure it ourselves, what exactly am I renewing?" Here is the answer, and it is a strong one.

What happens at non-renewal

The change-control clause — read this carefully

Giving the client the right to modify their own configuration is commercially right and technically fine, but it has a regulatory edge. The platform is a GAMP 5 Category 4 configured product: we validate the product, the client validates their configuration. The contract must say that in those words. Two consequences:

• They will need help doing it, which is exactly what the Validation Pack and Annual Periodic Review are for. The right to self-serve creates the service revenue rather than removing it.
• Under the EU AI Act, a deployer who substantially modifies an AI system or changes its intended purpose can become a provider. The licence therefore defines an envelope of permitted configuration — inside it they are a deployer and we carry provider obligations; outside it they take on provider status themselves. Partners should sell staying inside the envelope, not warn about leaving it.

Qualitum implements. Always.

Delivery is not delegated. Every deployment, validation pack and corpus tailoring is executed by the Qualitum team, on Qualitum's method, to Qualitum's evidence standard. The partner sells it, owns the client relationship, runs first-line support — and takes 50% of the implementation fee.

What moves the number

Capacity is the constraint, and we manage it openly

If we deliver every deal, then delivery capacity — not sales — is the growth ceiling. A standard site consumes roughly 35 engineer-days across deployment and validation pack, so six engineers of standing capacity is on the order of twenty to twenty-five sites a year, not fifty. Three rules follow, and partners are told all three on day one rather than discovering them at signature.

• Go-live slots are published a quarter ahead and allocated on signature date. A signed order books a slot; a verbal commitment does not. The scarcity is real, so we use it rather than apologise for it.
• Channel recruitment is capped to delivery capacity — no more than one Certified partner per two engineers of standing capacity. The channel will outrun delivery long before it outruns the product, and a partner with no slot to sell into churns and then talks.
• The hiring trigger is written down: when booked slots pass 70% of the following two quarters, we recruit. Waiting for 100% guarantees a two-quarter gap, because a delivery engineer in this domain takes a quarter to become useful.

The two levers that raise the ceiling without hiring are Deploy · Lite — steer clients there wherever their templates genuinely fit — and corpus tailoring, which converts bespoke effort into a repeatable asset.

Why delivery list prices are where they are

Deployment and validation are delivered entirely by us, and the partner still takes half the fee. So the list prices are set from the far end: the half we retain has to clear 40% gross margin on fully loaded delivery cost, or the channel is being funded out of our own delivery P&L. That is the whole reason deployment lists at €40,000 rather than the €26,000 an earlier sheet carried, and the validation pack at €40,000 rather than €28,000. It is a real price increase, and it is not negotiable downward without also changing the partner share — the two numbers move together or not at all. Fully loaded cost per line is shown in the master table in chapter 13, so you can see the arithmetic rather than take it on trust.

Margin policy — two floors, both must hold

• Gross margin on Qualitum's own revenue: floor 70% on licence, floor 40% on delivery. Licence COGS per site per year is ~€8–14k (L3 support, platform ops, model operations, renewal management).
• Qualitum's share of end-client licence value: floor 65%. At 65% retention and ~12% COGS the contribution is ~53% of end-client list — the 50% target now holds at every tier and every band, which was not true when partners took 35–40%.
• Delivery margin is deliberately thinner than licence margin. It buys the annuity.

You sell and support. We build and validate. We split it.

Market benchmarks: referral partners in software typically earn 10–15% of first-year contract value; partners who own the customer relationship earn 20–35%; and hybrid structures pay a first-year share plus a smaller recurring one. Our model sits inside that range on the licence and adds something the benchmarks do not have — half the implementation fee, for work we perform.

The support boundary, defined

The 50/50 delivery split only works if this is precise, so here it is. It goes into the partner agreement verbatim, and it is settled before a Certified partner signs rather than after the first argument about a ticket.

Rules that keep the channel honest

• Deal registration, 90 days. First to register with a named contact and a qualified opportunity owns the account. Renewable once on evidence of progress.
• The licence contract is always Qualitum ↔ end client. The partner earns margin but the licence and the EULA are direct. This protects the client (never orphaned if a partner leaves) and protects us (a partner cannot hold a renewal hostage).
• Renewal invoicing is ours. The partner share is paid automatically on collection.
• No sublicensing, no white-labelling the platform. Partners may white-label their own service wrapper. The product is Qualitum in front of the client and in the audit trail — which is what makes the compliance story hold.
• Account ownership is reviewed annually. The 25–30% renewal share is paid for keeping the client live, supported and current. If support goes quiet, the share reverts to us at the next renewal — and the client is told who to call instead.
• Support scope is written down — see the table above. L1 and L2 are theirs, L3 is ours, and the severity matrix and escalation rule sit in the partner agreement verbatim. This is the one place a 50/50 delivery split goes wrong, so it is settled before signature and not left to goodwill.
• Certified partner recruitment is capped to delivery capacity — one per two engineers of standing capacity. Signing partners we cannot give go-live slots to is how a channel destroys itself in its first year.

How we stop the licence leaking

Unlimited users plus client-side hosting plus a right to self-modify is, on paper, an invitation to give the product away. It isn't, provided six things are true.

Everything below the licence line

Infrastructure and model consumption sit outside the licence. We are agnostic by design — that is the sovereignty argument — and there are three ways a client can run us.

What tokens actually cost

Partners ask this on every call and usually assume it is the risk in the deal. It is not. The table below is built from a blended model price of $5 to $12 per million tokens — the low end being a mid-tier model with cached retrieval, the high end a frontier model running full multi-pass sampling on every document. Real deployments land in the middle.

Compliance: who is who, and what we charge for it

This is a pitch, not a disclaimer. The client cannot outsource accountability to a regulator — nobody can — so the offer is not “we take the risk off you”. It is: we make the evidence cheap, current and inspection-ready, and we stand next to you when someone asks for it.

Priced compliance items

• Included in the licence: platform validation evidence, regulatory-change updates, quality agreement, supplier questionnaire responses, one remote vendor audit per year.
• Validation Pack — €40,000/site, one-off at deployment. The client's own IQ/OQ/PQ evidence for their configuration.
• Annual Periodic Review — €18,000/site/yr. Sold as the thing that keeps self-maintenance safe.
• Inspection support retainer — €12,000/yr, plus €1,600/day for attendance.
• EU AI Act conformity file — €16,000, one-off, refreshed at €6,000/yr.
• On-site vendor audit hosting — €4,800/day beyond the included remote audit.

The ladder

Every enterprise software company that has won in a regulated category has run the same motion: a small, cheap, credible first commitment that produces a number the buyer can take to their finance director, followed by a site, followed by the network. Kneat ran exactly this and was bought for C$650M. We are not inventing a motion; we are compressing it.

The Proof of Value, defined

It is a paid, fixed-scope services engagement — not a licence, not a trial, and not an obligation to buy anything afterwards. It exists to convert a belief into a measured number the client can defend internally.

Commitment, cancellation and the credit

When €20,000 sits above their no-tender threshold

Ask it on the first call: “what is your no-tender threshold for software and services?” In most pharma organisations €20,000 sits below it. Where it does not, we do not split the purchase across two invoices to get under it — deliberate threshold-splitting is a finding internal audit actively looks for, and being the vendor who proposed it is a poor first impression to make on a quality organisation. The deal takes a different shape instead:

• Extend the teardown rather than shrinking the PoV. Half a day instead of ninety minutes, still free, still their own documents — now two or three packages, with their QA reviewer in the room. It does most of what the PoV does, and it costs us a morning.
• Then go straight to the lead site, with the first three weeks gated. The deployment's first milestone becomes an acceptance gate against the same three success criteria. Miss the gate and the client may terminate, paying only that milestone. Same risk transfer, one purchase order, no threshold problem.

A discounted half-scope PoV was considered and rejected: most of the cost is the environment build and the corpus load, neither of which halves, so a €9,500 version cannot clear the margin floor. What this route does cost us is the budget owner a paid pilot creates — so the champion has to be stronger before we take it. Qualify accordingly.

Why it is paid rather than free

• Free pilots do not convert in this category. Enterprise spend on generative AI tripled to $37B in 2025, yet MIT research found roughly 95% of generative AI pilots failed to show measurable P&L impact within six months. Almost all of them were free, unscoped and owned by nobody.
• A paid pilot creates a budget owner. Someone senior has to sign — and that is precisely the sponsor you need for the licence conversation twelve weeks later. Phase-gated pilots convert dramatically better than big-bang deployments.
• The credit removes the only real objection. They are not spending €20,000 on a trial; they are pre-paying the first slice of a licence they can still walk away from.
• The free step already exists. It is the Teardown. Anyone who will not give ninety minutes and one historical document was never going to buy.

Success criteria worth agreeing to

Pick three, in writing, before kick-off, measured against the client's own historical baseline which they supply: for example a 40% or better reduction in authoring and review hours on the sample packages; 100% requirement-to-test traceability in the generated output; and QA acceptance of the generated content by their own reviewer without structural rework. Agreeing the baseline is half the sale — it forces the client to quantify a cost they have been carrying without measuring.

The buyer, and what actually persuades them

Four objections and the answers

Cost, margin, partner price, client price

Everything on one sheet. Line items are at recommended list. “Our cost” is fully loaded internal delivery cost, not marginal cost. Each partner tier shows what Qualitum keeps and what the partner earns from the same client invoice.

All figures in EUR. The Proof of Value credit is applied to the client's licence invoice and comes entirely out of Qualitum's share — the partner's licence percentage is always calculated on full list, which is why the Certified year-one total shows €121,600 rather than €141,600.

Change any number

The calculator below starts on that same example. Every quantity, price and cost is editable, and so are the three partner percentages. Set a quantity to zero to remove a line.

Worked cases

Note the shape of case A: in year one the partner earns two thirds of what we do, because they carried the sale. From year two the position inverts permanently and our share is almost all margin. Say that to partners out loud — it is the deal, and hiding it makes them suspicious of the renewal terms.

The five things their QA will ask for

Chapter 11 commits us, in one line of a table, to five deliverables under GMP: a quality/technical agreement, a supplier qualification pack, validation evidence, change notification and audit access. They are included in the licence, which means they are pure cost, and they have to exist before the first signature rather than after it. This chapter says what each one is and what building it involves.

1 · Quality / technical agreement

A bilateral document signed by QA on both sides, not procurement, sitting alongside the commercial contract. Required by EU GMP Chapter 7 on outsourced activities and by Annex 11 §3.1, which obliges the regulated company to have a formal agreement with any third party providing a computerised system.

2 · Supplier qualification pack

What their QA needs in order to put us on the approved supplier list. Driven by Annex 11 §3.2 and by GAMP 5 supplier assessment. This is the single artefact that decides whether being qualified takes two weeks or eight, so it is a sales asset as much as a compliance one.

3 · Validation evidence

Careful with the naming. This is not the €40,000 Validation Pack we sell — that is the client's own IQ/OQ/PQ for their configuration. This is our product-level evidence, given away inside the licence, proving the platform itself is built under control so the client can leverage it rather than re-test the platform layer themselves.

• Software development lifecycle procedure and its records
• Requirements and design specifications
• Risk assessment (FMEA) at product level
• Unit, integration and system test protocols and executed results
• Requirements-to-test traceability
• Version control, release records, release notes, defect and known-issue log
• Configuration management and environment qualification
• A Part 11 / Annex 11 technical controls assessment — audit trail, electronic signature, access control, timestamps, record integrity against ALCOA+

This is the long pole. Three to four weeks with engineering, and it cannot be written by anyone who is not inside the codebase.

4 · Change notification

A contractual commitment not to break their validated state without warning. It needs a classification scheme with lead times, published, and identical for every client.

Add a version support window — n-1 supported for eighteen months — and a deprecation policy. Then state plainly that there are no silent model swaps. That single sentence does more work in a QA meeting than the rest of the annex combined.

Two pages, half a day to write. The one thing to get right is consistency with the configuration envelope in chapter 10: if clients may modify freely and we also promise change control, the boundary between configuration and code has to be identical in both documents.

5 · Audit access

Define it, or it becomes unlimited — and forty clients auditing us individually is a headcount problem long before it is a compliance one.

• One remote or postal audit per year included. On-site beyond that at €4,800/day, already priced in chapter 06.
• 30 days notice, reduced to 5 working days for-cause following a major deviation or incident.
• Scope limited to the service they buy. Escorted. No access to other clients' data, environments or records — which is itself a control they should want.
• Regulator access: we commit to supporting inspection by their competent authority or the FDA, including a pre-approval inspection, on reasonable notice.

Build order, and what it costs us

Why this sits in a pricing guide and not a compliance one

All five items are given away inside the licence, and four of the five carry a recurring cost: the evidence has to be refreshed at every release, the response bank has to be maintained, the quality agreement has to be reviewed on cycle, and every client is entitled to an audit a year. That is a standing cost against every licence line, for as long as the licence renews.

Which is the point. When a partner asks why the licence cannot be discounted by fifteen per cent to close a quarter, this is a large part of the answer, and it is a better answer than “policy”. It is also the strongest thing to say to a client comparing us on price alone: the cheaper quote in front of them almost certainly does not include a supplier who will still be producing current validation evidence in year three.

Frequently asked

Grouped by who asks. Click a question to open it. Answers here are the approved wording — if you find yourself improvising an answer to something on this list, tell us and we will add it properly instead.

Licence and limits

Price and comparison

Proof of Value

Delivery and partners

Technology, data and compliance

Seven decisions, and why the alternative lost

Earlier drafts of this model carried open questions. They are closed. Each one is recorded here with the reasoning, because a partner who knows why a rule exists will hold it under pressure, and a partner who only knows the rule will trade it away in the first difficult negotiation.

1 · Assurance & Upgrades is bundled, not added on

One annual number per site. The alternative was quoting assurance as a separate line at plus twenty per cent, which looks like more revenue and is not: a separable line simply gets netted against a lower base to reach the same total the market will accept. What it does reliably produce is a line item procurement can decline at the third renewal, which is precisely when budgets tighten and when we have the least leverage. Bundled is also harder to discount, because there is nothing separable to attack.

2 · The per-site bands are the only price vocabulary

€65k / €98k / €145k per site, €420k for a network. Earlier material carried a different set of tier figures; those are retired, not reinterpreted. The alternative was keeping both and explaining that one meant customer-level spend and the other per-site. Two price vocabularies inside one company is how an account ends up holding two different quotes from us, and nobody recalls a retrofitted definition correctly under pressure in a negotiation. Customer-level spend is now a derived figure — three standard sites is €294,000 — and never a published tier.

3 · Delivery list prices rose when we committed to the 50/50 split

Deployment at €40,000 and the validation pack at €40,000. The alternative was holding the earlier, lower figures and splitting those. It does not work: we carry the entire delivery cost and hand over half the fee, so the retained half at the old prices sits below any defensible margin, and the channel ends up funded out of the delivery P&L. The increase is simply the price of having a channel, and even after it a first-year bill from us is roughly seventy per cent below a Veeva-class equivalent. Anything quoted from an older sheet is re-quoted before signature.

4 · Delivery capacity is published and the channel is capped to it

Go-live slots published a quarter ahead and allocated on signature date; no more than one Certified partner per two engineers of standing delivery capacity. The alternative was recruiting the channel first and solving delivery afterwards. A partner with no slot to sell into churns, and a churned partner talks to the same small pool of accounts we are selling to. Scarcity that is stated in advance is a selling tool; scarcity discovered at signature is a broken promise. Chapter 08 has the mechanics.

5 · The Strategic tier exists, but is not published

Referral and Certified are on the sheet. Strategic is offered by invitation, after four partner-sourced sites are live or against a pipeline we have independently qualified. The alternative was publishing all three. With no certified partners in place yet, publishing three tiers invites every conversation to open by negotiating for the deepest one, before anybody has sold anything at all.

6 · The Proof of Value stays at €20,000; a low threshold changes the shape of the deal, not the price

Where €20,000 sits above an account's no-tender threshold we do not split the purchase across invoices — deliberate threshold-splitting is exactly the pattern a pharma internal audit function looks for, and proposing it to a quality organisation is a poor first impression to make. The alternative considered was a discounted half-scope pilot; it fails on arithmetic, because most of the cost is the environment build and the corpus load, and neither of those halves. The route round it is in chapter 12: an extended teardown, then a lead site whose first milestone is a terminable acceptance gate.

7 · Support tiers are defined before the first partner signs, not after the first argument

L1 and L2 belong to the partner, L3 to us, with response times and a four-level severity matrix, written into the partner agreement verbatim — chapter 09. The alternative was settling it in practice as tickets arrived. That is how a 50/50 delivery split turns into a dispute: not over money, but over who owned a ticket at two in the morning. Where there is no partner in the account, L1 and L2 become a priced service at €18,000 per site per year, which also makes visible what a partner's renewal share actually pays for.

What is still genuinely unknown

Three things, and none of them blocks a sale. Per-package token consumption, until the first three sites are instrumented — the figures in chapter 11 are engineering estimates and are labelled as such, and what is contractual is the envelope, not the estimate. The real support load behind a Certified partner, until one has been live a quarter. And the no-tender threshold in any given account, which is now a qualification question on the first call rather than a pricing question. Each has a stated trigger that reopens it. None of them is a reason to delay a quote.

Sources & method