Revenue Architecture for Vertical SaaS — The Complete Operator Guide in 2027

The two architectures you are actually choosing between

Every vertical SaaS operating team eventually faces the same fork, and most teams pick by accident rather than on purpose. The fork is not "SMB versus enterprise" or "product-led versus sales-led." It is whether the company's revenue architecture is seat-primary or transaction-primary, because that single choice cascades into pricing, comp, org design, board reporting, and ultimately the multiple the company trades at.

Architecture A — seat-primary with payments as an accessory. The subscription line is the business. Payments exist, but they are sold as a convenience after go-live, owned by a product manager, and reported as "other revenue" on the board deck. The revenue engine looks like horizontal SaaS: AEs carry ARR quota, CS carries logo retention, and the payments attach rate is a metric nobody is comped on. This architecture is easier to run in year one because the entire GTM playbook is borrowed — from Salesforce, from HubSpot, from every SaaS operating book written since 2015. It is also the architecture that caps out. A seat-primary vertical SaaS company with $25K ACV and 78% gross margin is, from a valuation standpoint, a horizontal SaaS company that happens to have industry-specific fields.

Architecture B — transaction-primary with software as the wedge. The subscription anchors the relationship and secures the system-of-record position, but the revenue architecture assumes the transaction layer will eventually be the larger number. Toast is the public proof point: roughly $5B of fintech revenue against $936M of software revenue in its 2024 fiscal year — a ratio on the order of 5:1 in favor of the non-software line. That ratio is not an accident of a low-margin restaurant vertical; it is what happens when a company designs its GTM so that every new logo is underwritten, onboarded, and comped for payments from the first conversation.

The trade-offs are real and worth naming honestly. Architecture B carries risk Architecture A does not: underwriting exposure, chargeback liability, processor concentration, regulatory surface area (money transmission, lending licensure in some structures), and a materially more complex revenue recognition conversation with your auditors. It also means your gross margin percentage goes *down* even as gross profit dollars go up — payments typically nets 60–75% margin against 75–80% for software, and a board that reads gross margin percentage as a quality signal will need re-education before the first quarter of blended reporting.

There is a third, less-discussed option worth mentioning because a meaningful number of teams land there: the partner-referral architecture, where you never touch the payment flow and instead route processing to Stripe, Adyen, or a specialist and book a revenue share. This is genuinely correct for some verticals — particularly ones with low transaction density, unusual settlement requirements, or where a dominant incumbent processor already owns the operator's rails. The economics are thinner (you capture a slice of a slice), but you avoid the balance sheet, the underwriting team, and the compliance build. Teams in legal tech, life sciences, and parts of construction have made this call deliberately and been right.

The neighboring decision that gets bundled into this one — and shouldn't — is whether to build the AI agent layer as a metered line or fold it into tier pricing. That is a separate architectural fork with its own economics, and answering it correctly requires knowing whether your agent actions have variable cost that scales with usage. Most do in 2027, which argues for metering, but bundling into a premium tier is defensible when the agent's job is to make the seat stickier rather than to be its own revenue line.

How to decide between them

The decision is mechanical if you answer four questions honestly about your vertical, and it does not depend on your ambition or your investor's preference. It depends on the operator's transaction profile.

Question one: what is the annual payment volume flowing through a typical customer? This is the gate. If a median customer processes $800K–$3M a year in card volume, transaction-primary is almost certainly correct — at a net take of roughly 80–110 basis points after interchange and network fees, a $1M-volume customer contributes something in the $8K–$11K range of payments gross profit annually, which can rival or exceed a $10K–$25K software ACV on a gross profit basis. If the median customer processes under $200K, the arithmetic collapses and you should be seat-primary with a referral relationship for the payments that do exist.

Question two: does the money already move through a system you control or could control? Restaurants take card payments at the point of sale, which is a terminal Toast already ships. Home services companies collect at the job site through an app the field tech already carries. Those are natural rails. A vertical where the operator invoices on 45-day net terms through their bank, and where you have no presence at the moment of payment, is a much harder wedge — you're asking them to change a banking relationship, not to use a button that is already on the screen they're already looking at.

Question three: is there a dominant incumbent processor with switching friction you cannot overcome? In some verticals the operator's processing relationship is bundled with equipment leases, franchise agreements, or association buying programs. Fighting that head-on burns sales cycles. Better to model payments as an 18–36 month land rather than a 90-day attach, and to price the software so it stands alone in the interim.

Question four: can you carry the compliance and risk build? Embedded payments at scale requires underwriting, KYC/KYB, chargeback operations, and — if you touch lending — either a bank partner or a licensing strategy. That is a real headcount and legal spend that a Series A company should think hard about before committing.

The output of this decision is not a slide. It is three concrete artifacts you should be able to produce within two weeks of making the call: a comp plan that pays for the behavior the architecture requires, a board KPI set that reports the architecture honestly, and an onboarding sequence that moves the customer to the target state inside the attach window.

One adjacent scenario worth planning for: companies that start seat-primary and later convert. This is common and survivable, but it is a re-platform of the go-to-market, not a pricing change. The rep who has spent three years selling seats does not become a payments seller because you added a line to the comp plan. Expect a 2–3 quarter productivity dip, expect to lose some percentage of the existing team, and expect the back-book conversion rate on existing customers to run well below the new-logo attach rate — because the existing customer already solved payments somewhere else, and you are now asking them to unsolve it.

Concrete numbers behind each option

Architecture arguments get resolved by arithmetic, so here is the arithmetic at the line-item level.

Software subscription pricing, by vertical shape. ServiceTitan's published Pro tier for home services runs in the neighborhood of $398 per user per month. Toast's restaurant plans start at a modest per-terminal base — in the $69–$165/month range depending on tier — with hardware and add-ons layered on top. Mindbody's boutique fitness plans have historically listed in the $129–$349/month band. Procore prices against annual construction volume rather than seats, roughly a fraction of a percent to 1% of the volume the customer runs through the platform, capped per company. Veeva sits at the far end — enterprise life sciences pricing negotiated per module, per organization, nowhere near a public list price. The pattern across all five: the software line is priced to be *defensible and boring*, not to be the growth engine.

Payments economics, unpacked. The posted rate a merchant sees is not what you keep. A typical card-present rate lands around 2.49% + $0.15; card-not-present, around 2.9% + $0.30; ACH on B2B invoice rails in the 0.8% range with a per-transaction cap. Out of the card rate, the overwhelming majority flows to interchange (to the issuing bank) and network assessments (Visa/Mastercard). The vendor's net — the number that actually reaches your P&L — typically lands somewhere in the 80–110 basis point range depending on mix, average ticket size, and how much of the stack you own versus rent from a processor. On $1M of annual customer volume, that is roughly $8K–$11K of gross payments revenue, against which you carry processing infrastructure cost, fraud loss, and chargeback operations, netting toward 60–75% margin.

The stacked customer, at maturity. A useful mental model for a mature transaction-primary customer at the three-year mark: roughly $25K of software subscription, plus $10K of payments gross revenue, plus $3K from capital products, plus $4K from metered AI agent actions. Call it low-to-mid $40Ks of annual revenue from a logo that landed at $10K–$25K. That expansion is not price increases; it is surface area. This is why the land-then-stack math works and why net dollar retention above 110% is achievable in vertical SaaS when it is difficult in horizontal categories — the expansion vector is new revenue lines, not seat growth in a customer whose headcount is flat.

Capital and lending economics. The merchant-cash-advance-style product — an advance repaid via a daily skim on processing volume — is where verticals with strong payments attach go next. Toast, ServiceTitan, and Mindbody have all launched capital products on this shape. Two structures exist: refer to a specialist partner and book a fee, typically a low-single-digit percentage of the advance amount, or take balance sheet risk and earn the full spread on factor rates that commonly sit in the 8–14% range for these products. The referral structure nets 25–45% margin with essentially no risk; the balance sheet structure earns far more and requires capital, an underwriting function, and a genuinely different conversation with your investors. Most companies should start with referral and revisit once they have 18 months of repayment data on their own customer base.

AI agent pricing, the 2027 line. The metered agentic line is the newest and least settled. The observable shape across vertical leaders is per-action pricing — per AI-handled inbound call, per drafted intake, per processed submission — landing in a roughly $0.50–$3.00 per action band for SMB-oriented products, or bundled as a percentage uplift on seat cost in the enterprise tier. Gross margin at scale runs 70–85% once inference costs amortize, but early-stage margin can be materially worse and you should model it honestly rather than assuming the eventual number. Realistic penetration in 2027 — the share of your customer base carrying at least one agentic billing line — is 15–25%. Treating that as a near-term majority is how forecasts break.

Sales motion cost structure by segment. The SMB inside motion runs roughly $60K base against $120K OTE — a 50/50 split — with activity expectations in the range of 350 dials weekly, 8–12 demos, and 3–5 closes monthly at $10K–$20K ACV. Fully loaded CAC lands around $3K–$7K with a 9–14 month payback when payments are attached, considerably worse when they are not. The mid-market pod pairs an AE at roughly $110K base / $220K OTE with a solutions consultant near $130K base, running 45–90 day cycles against $30K–$80K ACVs. The enterprise named-account rep carries 30–50 accounts, sits at roughly $160K base / $320K OTE against a $1.2M–$1.8M quota, and runs 6–18 month cycles at a 15–20% win rate against 7–10 stakeholders — consistent with widely cited Bridge Group benchmarks for six-figure ACV enterprise motions. Note the consistent 50/50 base-to-variable split across all three tiers; accelerators kick in at 80% of quota attainment.

Coverage ratios. Customer success staffing scales inversely with segment: roughly 1 CSM per $5M ARR for pooled SMB coverage, 1 per $2.5M in mid-market, and 1 per $500K in enterprise. RevOps typically reaches a 6–8 person team by $30M ARR in a transaction-primary architecture, larger than the horizontal equivalent because someone has to reconcile processor statements against the revenue system every single month.

Implementation details and sequencing

Knowing the architecture is a fraction of the work. Sequencing it is where teams actually fail, and the failure is almost always the same: they announce the architecture before they have built the machinery that makes it true.

Quarter one — instrument before you sell. Before a single rep is asked to attach payments, you need attach rate visible per rep, per segment, and per onboarding cohort in the same system where pipeline lives. If payments attach lives in a finance spreadsheet, it will not change behavior. Build the reconciliation loop first: processor statement → revenue system → CRM opportunity → rep credit. This is unglamorous RevOps plumbing and it takes a quarter. Skipping it means quarter two's comp plan pays on numbers nobody trusts.

Quarter two — rewrite comp and onboarding together. The comp plan and the onboarding sequence are a matched pair. If the rep is paid on payments attach but onboarding schedules the payments conversation for day 120, the rep will fight onboarding and lose. The attach window is real and narrow: attach rates fall off sharply after the first 90 days post-go-live, and by 180 days the operator has re-stitched their existing processor relationship and the conversation restarts from zero. Design the implementation sequence so the payments conversation happens *during* implementation, when the operator is already in change mode and the switching cost is being paid anyway.

Quarter three — build the capital and agent rails on top. Do not launch capital or AI agent products until payments attach is above roughly 50–65% of new logos, for a simple structural reason: both products depend on the payments rail. Capital repayment skims processing volume. Agent actions are most valuable — and most easily billed — when they touch the transaction. Launching them early means selling a product to a customer base that structurally cannot use it, which produces terrible attach data and a demoralized team.

Quarter four — install the operating cadence. The cadence is the deliverable; without it the architecture is a deck. Weekly: a Monday pipeline-and-payments review with the CRO, VP Sales, VP Payments, and VP RevOps in one room for 60 minutes, and a Friday 30-minute module attach scorecard with the CRO and VP Customer Success. Monthly: the NDR cohort cut by signing quarter, the payments take-rate reconciliation against processor statements, and an AI agent penetration review by segment. Quarterly: the six-KPI board review, a comp plan true-up to reset accelerators, and — in Q3 — the annual planning kickoff for next year's seat-plus-transaction mix.

Ownership, explicitly assigned. The Chief Revenue Officer owns the integrated software-plus-transaction P&L and reports it as one number to the board; splitting the report is how seat ARR gets optimized while take rate quietly decays. A dedicated VP Payments or Head of Embedded Finance owns interchange margin, chargeback rate, attach rate, and the processor relationship. Whether that role reports to the CFO or the CRO is a genuine judgment call — both configurations exist at scale and both work. What consistently fails is leaving embedded finance inside Product, where it gets roadmapped rather than operated. The VP Customer Success owns the gross retention floor and module attach, and must be comped on those rather than on NPS or ticket SLA — the moment CS becomes a support organization, the expansion engine stops. The VP RevOps owns forecast accuracy, stage hygiene, comp administration, and the monthly reconciliation.

The go-to-market details that are specific to vertical. The buyer is not a CIO. It is the owner-operator of a one-to-fifteen location business or the VP of Operations at a multi-location chain, and their alternative to your product is paper, spreadsheets, and six or seven point solutions duct-taped together. That changes outbound mechanics concretely: these buyers are not reading LinkedIn at 9am, they are at the morning huddle. Calls land before 8:30am or after 4:30pm on a mobile number. Industry events — the trade association shows, the vendor's own user conference — carry disproportionate weight compared to horizontal SaaS, where the same spend would go to digital. And the opening line of the discovery call is an operating question ("how many minutes a week does your closer spend reconciling tip-outs?"), not a platform pitch.

Adjacent motions worth borrowing from. Two neighboring categories have solved problems vertical SaaS keeps re-solving. Payroll and HR platforms figured out the compliance-as-moat play a decade early — the observation that a product handling regulated filings on the customer's behalf has structurally higher retention than one that does not. Marketplaces solved the two-sided attach problem: how to price when the party who benefits and the party who pays are different people. If your vertical has a consumer side — patients, diners, homeowners — the marketplace playbook on consumer-side monetization is more instructive than anything in the SaaS canon.

Related questions

Should payments attach be in the AE comp plan or the CS comp plan?

Both, weighted differently. The AE should carry a payments attach component on new logos — typically 15–25% of variable — because the conversation must start pre-close. CS carries back-book conversion and module attach. Splitting it cleanly to one side creates a handoff gap right at the 90-day window.

What if our vertical's operators already have entrenched processor relationships?

Model payments as an 18–36 month land rather than a 90-day attach, price the software so it stands alone at full margin, and target the natural switching moments: equipment refresh, location openings, ownership changes, and contract renewal dates. Track those dates as a pipeline of their own.

Does AI agent revenue change the valuation multiple?

Directionally it should, because metered agentic revenue expands with customer activity rather than headcount. But penetration in 2027 is realistically 15–25% of the base, so it is not yet a material multiple driver for most companies. Report it honestly as an emerging line rather than overweighting it in the narrative.

How do you report blended gross margin without spooking the board?

Report gross profit dollars per customer alongside margin percentage, and show the cohort expansion curve. The percentage falls as payments mix rises; the dollars per logo rise substantially. Pre-brief the audit committee before the first blended quarter rather than explaining it after.

Is the seat-primary architecture ever the right long-term answer?

Yes — in verticals with low transaction density, unusual settlement structures, or dominant incumbent rails. Life sciences and parts of legal fit this. The correct move there is to compete on depth of workflow and regulated-data handling, and to accept a software-multiple business rather than force a transaction wedge that the vertical will not support.

FAQ

What opening ACV should a new vertical SaaS company target?

Target a $10K–$25K opening ACV with a designed 36-month expansion path toward $50K–$120K through payments, capital, and metered AI agent attach. The land number matters far less than whether the expansion surface exists and is instrumented from day one. A $10K land with four expansion vectors beats a $40K land with none.

Do I need a dedicated SDR team for vertical SaaS?

For mid-market and enterprise motions, yes — roughly a 1:1 SDR-to-AE ratio is standard. The SMB motion frequently runs without dedicated SDRs, with AEs self-prospecting into named operator lists built from licensing databases, association rosters, and trade show attendee lists. The prospecting data sources in vertical are unusual and often better than commercial contact databases.

When exactly should payments be attached?

Inside the first 60–90 days post-go-live, and ideally during implementation itself when the operator is already absorbing change. Past 180 days, attach rates drop materially because the operator has settled their processing relationship elsewhere and you are now asking them to undo a decision rather than make one.

What gross margins should I plan for across the stack?

Roughly 75–80% on software subscription, 60–75% on payments after interchange and risk, 25–45% on capital depending on referral versus balance sheet structure, and 70–85% on AI agents once inference cost amortizes. Blended margin declines as the transaction mix grows — plan the board narrative around gross profit dollars per logo.

How many customer success managers do I need?

Approximately 1 CSM per $5M ARR for pooled SMB coverage, 1 per $2.5M in mid-market, and 1 per $500K in enterprise where the CSM is running a genuine account plan. Comp them on gross retention and module attach, never on ticket volume or SLA.

What are the board KPIs for a transaction-primary architecture?

Six: net dollar retention (110%+ is the bar), gross retention (95% floor), payments attach rate on new logos within 90 days, magic number above 1.0 at scale, module attach index (average modules per customer), and AI agent penetration as a percentage of the base. Present NDR as a cohort table by signing quarter, not as a single blended figure.

Sources

• https://investors.toasttab.com/ — Toast investor relations, annual reports and quarterly results covering the fintech-versus-subscription revenue split.
• https://www.sec.gov/edgar/search/ — SEC EDGAR full-text search for ServiceTitan, Toast, Procore, and Veeva filings (S-1, 10-K, 10-Q).
• https://ir.veeva.com/ — Veeva Systems investor relations, annual reports on the life sciences vertical model.
• https://investors.procore.com/ — Procore Technologies investor relations, construction vertical pricing and retention disclosures.
• https://www.bvp.com/atlas — Bessemer Venture Partners Atlas, including the State of the Cloud reports and vertical SaaS theses.
• https://blossomstreetventures.com/ — public SaaS metrics compilations and comparables.
• https://stripe.com/guides — Stripe's guides on interchange, embedded payments economics, and platform monetization.
• https://www.adyen.com/knowledge-hub — Adyen knowledge hub on interchange structure and platform payment models.
• https://www.federalreserve.gov/paymentsystems/regii-average-interchange-fee.htm — Federal Reserve data on average debit interchange fees.
• https://blog.bridgegroupinc.com/ — The Bridge Group's published SaaS AE and SDR metrics research.

Related on PULSE

• [Revenue Architecture for Vertical SaaS for General Contractors in 2027 (Procore-style Multi-Module Expansion)](/knowledge/ra0113)
• [Revenue Architecture for Vertical SaaS for Electrical Contractors in 2027 (Residential vs Commercial Split)](/knowledge/ra0112)
• [Revenue Architecture for Vertical SaaS for Roofing Contractors in 2027 (Insurance Supplements, Storm Seasonality)](/knowledge/ra0111)
• [Revenue Architecture for Vertical SaaS for HVAC + Plumbing in 2027 (Financing Attach, PE Roll-up)](/knowledge/ra0110)
• [Revenue Architecture for Vertical SaaS for Lawn Care + Grounds Maintenance in 2027 (Recurring Contracts, PE Roll-up)](/knowledge/ra0109)
• [Revenue Architecture for Vertical SaaS for Pest Control in 2027 (PE Roll-up Channel, Payments, NRR)](/knowledge/ra0108)