Financial Model

01 Pricing tiers (Path C+ locked April 28, 2026)

Pricing is the financial spine of the model. Path C+ was locked on April 28, 2026 after running every alternative through the PLG and VC committees. The result is a four-tier ladder with one paid entry point and two expansion altitudes. Free is the diagnostic ritual. Pro keeps DAC past graduation. Business spreads DAC across the product org. Enterprise sells the portfolio view.

No $49 Solo tier. No $149 Pro tier. Both were rejected. ProfitWell data is unambiguous on the Solo problem (it attracts tire-kickers and depresses contribution margin). The lower Pro made the headline number look softer without changing conversion math, because the willingness-to-pay anchor for a senior PM running an agentic-era pod sits well above that range. Path C+ keeps the price honest and the tiering wide enough that the natural expansion path is Pro to Business, not Pro to a deeper Pro.

The annual discount on Pro is 20% (effective $239/mo). Annual on Business is similar at $960/mo. Enterprise is custom annual, scoped to portfolio breadth, contract length, and SSO posture. The annual lever exists to lift cash collected up front and to dampen month-2 churn on Pro, which is the highest-risk window for any reverse-trial graduation.

The four tiers map onto four distinct buyer postures, and the model assumes the mix is roughly 60% Pro / 30% Business / 10% Enterprise by the Series A milestone. Free is uncounted in the paid mix because Free is a conversion vehicle, not a revenue line. The mix shift over time is the engine of ARPU compounding: Pro is the entry point, Business is where customer success drives expansion, and Enterprise is where portfolio-level deals close after the first two have been proven. The bridge in section 7 carries that mix shift line by line.

Three altitudes anchor the willingness-to-pay argument (Poyar pricing framework). Pro replaces a Reforge or Lenny+ subscription at roughly $500/mo. Business replaces an executive coach at roughly $1,500/mo. Enterprise replaces a Head of Product Ops hire at $250K plus equity. Each tier sits below its anchor with measurable ROI, which is the structural reason the price ladder holds against discount pressure in early sales conversations.

02 Per-score unit economics

The atomic unit of cost in the Dacard model is one full diagnostic score across 88 dimensions and three frameworks (Team Operations, Development Lifecycle, Product Assessment). That single score is what the customer pays to receive, and what the engine pays to produce. Today, the loaded cost of one score is approximately $0.17.

Inference is the largest line and the only one with material upside risk. Today the primary model is claude-sonnet-4-5, with a fallback model tested end-to-end against the same eval harness. Trace capture is mandatory at every call site (see the observability rule in this codebase), so every score has a stored prompt and response, which is what enables prompt-level cost analysis and (later) fine-tuning.

The pattern-discovery pipeline is the invisible cost lever. It does not run live LLM calls. It runs scheduled jobs against snapshots of the scored signal base, computing co-occurrence statistics across 88 dimensions, archetypes, and outcomes. That means the pipeline scales sub-linearly with customer count. Doubling customers does not double calibration cost; it makes the calibration cheaper per discovered pattern, because the same job amortizes across a wider base.

The structural cost insight is that the inference price ceiling is bounded, not the inference price floor. We pay model providers; we do not own the model. The fine-tuning step at the 2,000-customer inflection is the answer to that bound. At 2,000 customers, we have enough trace data (estimated 500K+ scored prompt-response pairs) to fine-tune a model that produces structurally similar output at meaningfully lower per-call cost. The expected one-time cost is $50K-$100K. The expected payback is 3-6 months. Post-tune, blended gross margin moves to 92%+, which is the inflection from a healthy SaaS company to a structurally compounding one.

The other escape valve is a redundant inference path. The fallback model is already wired and tested. If the primary model price moves materially (a hike, a deprecation, a contract change), we route to the fallback inside 24 hours without product impact. This is not theoretical; it is the reason every call site in this codebase passes through a single capture wrapper.

Adapter compute is the second line, at approximately $0.04 per score. Every score requires pulling normalized signals from the operational adapter base (currently 54 providers, 178 signal types, with GitHub and Linear live and OAuth-registered for Slack, Jira, PostHog, Figma, Attio). The adapter compute cost rises with the number of providers a customer has connected, but it is bounded above because additional providers produce diminishing marginal signal once the first three are in place. Storage and indexing rounds out at approximately $0.02 per score, dominated by the trace base itself plus the indexed signal store. Storage is a sunk cost on the platform side, not a per-score variable cost; we attribute it per score for unit-economics honesty.

One way to read this $0.17 figure is as a ceiling, not a floor. As volume grows, three of the four input costs go down. Inference goes down with volume discounts, fine-tuning, and the proprietary path. Adapter compute goes down with shared-cache hits across customers using the same provider stack. Storage goes down per-score because the indexed signal store amortizes across more scores. The fourth input (provider API rate-limit overhead, a fixed cost per provider) is structurally bounded and falls per-customer as the customer base grows.

03 Per-tier monthly economics

Per-tier economics roll up from the per-score unit. The blended monthly COGS estimate per tier assumes the customer is using the product, not just paying for it. We score conservatively: peak-usage COGS is the right input for a margin floor, not average-usage COGS. The blended target is 78-82% gross margin at full Pro usage, with Business and Enterprise cleanly above 90%.

Pro is the only tier where peak-usage margin dips below 80%. That is intentional. Pro customers run heavy scoring frequency in the first two weeks past graduation as the team feeds the diagnostic ritual into Day 60 and Day 90 milestones. Average usage across a Pro cohort settles closer to 89% gross margin once the workload normalizes. The blended 78-82% target captures the worst-case Pro mix.

Business and Enterprise margins do not dip the same way. Their COGS scales with the number of products scored, not with seat count, and large customers concentrate scoring on fewer high-value products rather than running every team flat-out. The result is that the more a customer expands, the better the gross margin on the marginal dollar.

04 Revenue per FTE modeled

The headline efficiency number is $800K+ revenue per FTE, modeled on a 6-FTE founding team. The team composition implied by the funded plan is the founder, a founding engineer at month 3, a head of customer success at month 12, an account executive at month 15, and two implied later hires (a second engineer and a customer-success generalist) inside the 18-month window or shortly past it.

Industry benchmark is approximately $200K revenue per FTE at Series B for healthy SaaS. Dacard targets 4x that benchmark on a smaller team. The argument is not heroism; it is leverage. Every employee runs an agent fleet. The diagnostic engine runs against snapshots, not human input. The integration adapter base is finite and well-bounded. The product surface is one diagnostic with three layers of moat underneath, not a sprawling platform.

The shape of this curve matters more than the absolute number. The trajectory follows the agent-leverage curve, not the headcount curve. If we hired the eight people implied by current product-velocity readings (the team-size that a traditional SaaS plan at this stage would call for), the gross margin would invert. Salary is the largest non-inference cost. Each additional human hire in the agentic-era plan is a high-bar decision because the alternative is not "do less" but "deploy another agent skill."

This is also why there is no CRO line item in the use of funds. Distribution is built first through agent-skill mechanics, MCP, and founder-led content. The first AE is a month-15 inflection-cycle hire. The CRO is not a pre-close commit and will not be hired pre-Series A. The committee consensus on this is unanimous: chief revenue officers work for revenue engines, not category creators.

The hiring waterfall is deliberately ordered around moments where the existing team has produced enough signal for the next hire to make their first 90 days productive. Founding engineer at month 3 lands when the design-partner cohort has produced four to six weeks of usage data and a real bug surface. Head of customer success at month 12 lands when the cohort has produced expansion candidates the new hire can convert in their first quarter. AE at month 15 lands when the customer base has produced enough referral pattern that an outbound motion can convert against a warm pipeline rather than building cold-call muscle from scratch. Each hire arrives with infrastructure, not into a vacuum.

05 Use of funds ($1.5M pre-seed)

The round is $1.5M, sized to deliver 18 months of runway with the right hires at the right inflection points. The shape of the table below is the round design, not the round narrative. Every line item exists because removing it lengthens runway at the cost of a milestone, and every line item is sized at the lower bound of the plausible range.

The founder line is 40% of the round at a capped salary; the equity is the compounding instrument, not cash. The founding engineer is the highest-leverage hire on the plan: senior, AI-augmented, hired at month 3 once the design-partner cohort is producing usage data the engineer can build against. Head of customer success arrives at month 12 to drive the expansion motion ahead of the first NRR cohort numbers. The infrastructure line carries inference, the integration adapter stack, Vanta for SOC 2, Vercel, Turso, Clerk, Stripe, and the supporting vendor stack. The design-partner program funds incentives, two conferences, and advisor fees for the active and target advisors. The reserve is 5%, sized to absorb either a 6-month overage on the engineer hire or to pull forward the AE-M15 inflection by a quarter without re-opening the round.

There is no chief revenue officer line. There is no second engineer hire inside the round either; the second engineer is funded by the pre-A or seed-extension at Q4 2026, after the cohort numbers prove the unit economics in production. The plan is deliberately under-capitalized on headcount because the leverage thesis depends on it.

06 Cohort economics

Cohort economics are how this category compounds. The model is a reverse-trial Free funnel into Pro, with Pro to Business expansion as the primary value-capture motion. The numbers below are targets calibrated against benchmarks from the PLG committee (Wes Bush, Kyle Poyar, Elena Verna, Hila Qu) and the assessment-as-a-service comparable set (Vanta).

The reverse-trial is the conversion engine. Free is not a marketing tier. It is the diagnostic ritual itself, executed in 30 days against the customer's real Linear and GitHub data. By Day 30, the customer has either internalized the framework and converted to Pro, or the framework did not match their team. Both outcomes are useful. The Day 60 pushback artifact (a stage-validation lookback that compares the projected stage to measured stage) is the second conversion gate for customers who paid Pro and are evaluating renewal.

The 6-month CAC payback target is conservative for the agent-skill distribution thesis. If 30% of customers arrive through agent-originated channels (Battery's projection benchmark, calibrated against Neon's 80% in 24 months), the effective CAC drops, payback compresses, and the model upgrades. We are modeling on the 6-month assumption to keep the floor honest.

Cohort LTV under the modeled assumptions works out as follows. A Pro customer at $299/mo with 92% logo retention and 115% net retention has a modeled LTV of approximately $9,000 over a 30-month effective lifetime. A Business customer at $1,200/mo, with stickier integration footprint and lower churn (95% logo retention modeled), runs to roughly $40,000 LTV. Enterprise depends on contract structure but is materially above the Business line. The blended LTV-to-CAC ratio comes in at approximately 5x at $1,800 blended CAC, which is the ratio venture investors look for at this stage.

The cohort signals to watch in the first six months are not the conversion rate; they are the leading indicators that conversion is about to happen. Activation (sources connected at 2+ within 7 days) is the North Star. Day 14 ritual completion is the second gate. Day 21 sharing behavior (the user shows the diagnostic to a teammate) is the third. If the leading indicators move, the conversion follows. If the leading indicators stall, the conversion will too, regardless of how aggressive the email cadence becomes.

07 Revenue bridge to Series A

The bridge below is modeled, not actual. Q3 2026 is the design-partner quarter. Q4 2026 is the pre-A milestone. Q3 2027 is the Series A milestone. Every line in between is the path under reasonable assumptions about Free-to-Pro conversion (10% blended), Pro-to-Business expansion (15% of Pro within 12 months), and net retention (115%).

The bridge moves on three vectors: customer count grows roughly linearly through the funnel, average ARPU rises as the Pro/Business/Enterprise mix matures, and ARR compounds because retention plus expansion outpaces logo churn. The model deliberately uses conservative ARPU (Pro at $299, Business at $1,200, Enterprise starting at $2,500) and does not model upside from custom enterprise contracts above the $5,000 ceiling.

If Free-to-Pro conversion comes in below 8% in Q4 2026, the Q1 and Q2 numbers reset by approximately 25-30%. If the agent-skill distribution channel originates 30% of paid customers (Battery thesis), the pipeline compresses and the Q2-Q3 milestones can pull forward by a quarter. We are not modeling either deviation in the headline numbers; the headline numbers are the conservative baseline.

Three sensitivities are worth flagging. First, the ARPU lift from $299 in Q3 2026 to $500 in Q3 2027 is driven primarily by Business expansion, not by Pro pricing changes. The price ladder is locked. Second, the customer-count growth from 100 to 300 in three quarters assumes the agent-skill distribution channel and founder-led content together produce roughly 60-70 net new Pro signups per quarter, with a Business expansion conversion rate of 15-20% on the Pro base. Third, the milestone-row ARR numbers (Q4 2026 at ~$480K and Q3 2027 at ~$1.8M) are sized at the conservative end of the modeled range; the upside-case carries Q3 2027 closer to $2.4M ARR if Business expansion runs at 20% rather than 15%.

08 Series A milestones (locked)

The Series A pitch is the layered moat backed by published pattern-validity reports. The required milestones are below. They are locked, not aspirational.

The pattern-validity reports are the moat made visible. Three patterns are live today (Translation Gap, Fragility Signal, Compound Ready). The Q4 2026 milestone target is 8 validated patterns. By Series A, the target is 6-8 with empirical predictive-validity support, plus two consecutive quarterly reports comparing pattern firings to outcome data. Archetype calibration is the dimension-weight personalization layer. Without it, scoring stays generic. With it, the engine measurably outperforms a flat-weighted scoring approach.

The 80% gross-margin floor at Series A is the line we will not cross. If gross margin drops to 75-78% under unforeseen inference price pressure, the Series A pitch is the fine-tune step (already de-risked, models trained, infrastructure proven) and the path back to 92%+ blended margin within two quarters.

09 Risks and mitigations (financial framing)

The financial risks below are the ones that move the model. Each has a mitigation already wired or in motion.

10 Compounding margin curve

The agent-leverage argument is the structural reason this business compounds. Most consumer-AI companies have margin curves that compress as usage grows: more usage, more inference cost, lower margin. Dacard's curve is the inverse. More customers means more trace data, which means a more credible fine-tune, which means a lower marginal cost per score, which means higher gross margin on the marginal dollar.

Year 1 margin is bounded by what we pay third-party model providers. Year 2 margin is bounded by what fine-tuned inference costs to run on existing infrastructure, which is materially less than calling a frontier model. Year 3 margin is bounded by ownership: at sufficient scale, hosting a proprietary scoring model on dedicated infrastructure becomes cheaper than calling any provider, and the trace base becomes the competitive moat that makes that step possible.

The margin curve is the inverse of the consumer-AI-app pattern because Dacard does not pay inference per user query. We pay inference per scored team, which is bounded by the diagnostic cadence (one full score per onboarding ritual; lighter scores between). Heavy users are not unprofitable users. They are the highest-margin users, because they generate the most trace data per dollar of revenue, which compounds the fine-tuning advantage.

11 Summary

The financial model is operator-coach simple. Pricing is locked at four tiers with one paid entry point at $299 and clean expansion altitudes at $1,200 and $2,500+. Per-score COGS is $0.17 today, dominated by inference, with a fine-tuning step modeled at the 2,000-customer inflection that pulls blended gross margin to 92%+. Per-tier monthly margins are 78-82% at Pro under peak usage, 91% at Business and Enterprise. Revenue per FTE is modeled at $800K+ on a 6-FTE founding team, which is 4x the SaaS Series-B benchmark.

The $1.5M pre-seed is sized for 18 months. The use of funds breaks 40/20/10/15/10/5 across founder runway, founding engineer, head of customer success, infrastructure, design-partner GTM, and reserve. There is no chief revenue officer line. There is no second engineer line. The plan is deliberately under-capitalized on headcount because the leverage thesis depends on it.

The pre-A milestone (Q4 2026) is ≥100 paying customers and ~$480K ARR. The Series A milestone (Q2-Q3 2027) is ≥300 customers, ~$1.8M ARR, NRR > 110%, gross margin 80%+ blended, 6-8 patterns validated, two public quarterly pattern-validity reports, and archetype calibration shipped. The pitch at Series A is the layered moat backed by published evidence.

Questions on the model? Reach out to darren@dacard.ai. The pitch deck is at investor-deck.html; the memo is at investor-memo.html; the traction view is at investor-traction.html.

All numbers in this document are modeled targets, not actuals. First NRR cohort numbers ship Q4 2026.