The spend-per-month problem is structural. Four pricing models exist. One of them solves it. Two funding levers worth more than $800K per year are sitting behind a single missing artifact.
That sentence describes a structural problem, not a negotiation problem. It is not that your customers are unreliable. It is that the contract architecture you have built over 14 years contains no committed baseline. Every relationship, including the Amazon anchor and the million-dollar conversation that sounds like a commitment, is a spend-per-month relationship that restarts every 30 days.
The consequence runs deeper than cash flow. You cannot staff forward against uncommitted revenue without taking on the personal financial risk of overhiring. You cannot present a stable revenue story to any potential partner, acquirer, or investor. You cannot plan quarters. On the call you put the other version of this plainly: “we need to have some kind of Anchor a deal that says, okay, you know, it’s not, it’s not, it’s not painful anymore every end of the month.”
The goal of the Revenue Architecture engagement is not to rebuild your pricing from scratch. It is to convert a portion of what you already have from month-to-month tolerance to annual minimum commitment (even partial, even floor-only) so that the business can plan forward instead of reacting backward.
Four models are available for a data services firm at SmartOne’s stage. None of them are novel. The question is which one your buyers will accept at the volume floor that converts your ARR problem from unsolved to solved. The Phase 1 Revenue Architecture role would stress-test each at three scenarios: slow growth (no robotics deal, flat Amazon), expected growth (robotics deal lands, two new accounts), and high growth (robotics deal plus France activation plus two enterprise closings).
| Model | Fit for SmartOne | Why | First-test deal shape |
|---|---|---|---|
| Per-seat | Low | Your value is annotated output quality at volume, not user seats. Per-seat misaligns the value metric and penalizes buyers who scale usage. No buyer of data annotation services thinks in seats. | Not recommended. Would create immediate friction in any renewal conversation. |
| Per-labeled-unit | Medium | Familiar to current buyers and easy to benchmark against competitors. The problem: it locks SmartOne into commodity pricing rails and creates no floor. Volume drops in month two and revenue drops with it. Every month is still re-decided. | Can be used as a pricing layer inside a subscription structure. Unit price becomes the overage rate above a committed floor, not the primary vehicle. |
| Subscription with volume floor | High | Annual or quarterly commitment with a minimum volume floor addresses the spend-per-month problem directly. Buyers retain flexibility above the floor. SmartOne gets predictable revenue to plan against. This is the clearest path to ARR and the only model that produces a revenue-stability story that survives investor or partner due diligence. | Pilot with the robotics deal: propose a 12-month agreement with a floor of 150 annotators and an overage rate for 150-300. Buyer gets cost certainty on the base. SmartOne gets a committed baseline. The floor is the ARR event. |
| Outcome-based | Low (conditional) | Ties fee to model accuracy, annotation velocity, or a buyer-defined metric. High value alignment but requires SmartOne to have clear measurement infrastructure, a defined outcome metric the buyer agrees to in writing, and margin headroom to absorb a miss. Realistic only for well-defined, repeatable use cases with a stable quality baseline already established. Not for the current breadth-first catalog. | Worth piloting on a single healthcare or defense-prime account where the accuracy floor is already demonstrated and contractually documented. Not a first move. |
Your growth has always been expansion-led. On the call you described it directly: no real go-to-market, customer acquisition was luck and ad hoc outreach, and most new volume came from Amazon-alumni referrals. Shahysta came back in January 2026 with three to four warm leads, not a new-logo pipeline.
The instinct in most growth conversations is to treat expansion as a consolation prize and new logos as the real signal of health. That framing is wrong for a company at your stage.
Comparable plays at SmartOne's stage, iMerit being the closest analog, built their early ARR base by deepening existing accounts before scaling outbound. The Series B ($20M, December 2019) followed the revenue, not the other way around. The pattern: lock annual commitments in the verticals where you already have trust, use that revenue stability as the proof point for every subsequent expansion conversation, and build the cold-outreach motion from a position of stability rather than survival.
Your warm accounts and the robotics pilot are not a fallback position. They are the fastest path to a committed revenue base. The deliberate strategy is: convert the warm four to annual minimum commitments at whatever floor the buyer will accept, use those four contracts as the proof of concept for the subscription model, and then build the new-logo motion from a position of stability rather than survival.
Amazon is approximately 60% of your revenue. On the call you put the number at $2M to $3M per year against $6M total, and named it precisely as a structural information filter, not just a revenue risk: “our only view was Amazon when it represents like sixty percent of our revenue, everything comes from Amazon, so this has fed everything we saw about the market for the last two years.”
Before discussing how to unwind that dependency, one comparable is worth naming directly.
Sama is an East Africa-based data annotation firm with the closest structural analog to SmartOne: an emerging-market full-time workforce, ethical-AI positioning, a B-Corp certification, and the loudest possible post-Meta-Scale messaging about independence. In April 2026, Meta terminated Sama’s contract after Kenyan workers reported exposure to sensitive content from Ray-Ban Meta smart glasses without adequate anonymization or support. The result: 1,108 workers laid off in Nairobi. A class-action complaint filed in US federal court. Investigations opened by Kenya’s Data Protection Commissioner and the UK’s Information Commissioner’s Office. One client relationship, one decision, one month. One thousand people. Source: The Next Web, April 2026.
This is not a criticism of Sama. The situation is more complicated than the headline. But the structural lesson is specific and transferable: ethical-AI positioning is not a hedge against single-client concentration. B-Corp status is not a hedge against single-client concentration. The only hedge against single-client concentration is multiple clients with committed revenue, distributed enough that one termination cannot end your workforce.
The Amazon unwind is not a panic move and it is not a betrayal of the relationship that built SmartOne. It is a 24-month account-mix shift, executed quietly, by closing the three to four warm accounts at annual commitments, activating France in month five or six, and ensuring that by month 24 Amazon is 35-40% of revenue rather than 60%. The absolute Amazon revenue can grow during that period. The goal is a smaller share, not a smaller number.
Europe Growth is Phase 2 of the engagement, not Phase 1. On the call you explained why the window is stronger now than two years ago: LLMs fragmented the market, French enterprises are now building their own AI toolboxes instead of buying single-vendor solutions, and the ICP SmartOne sees in France in 2026 looks like the ICP they saw in the US in 2022 to 2023.
The ESSEC alumni path, the Safran framework agreement approach, the Bpifrance buyer pipeline, and the French-language materials are covered in the Europe Growth section of the /team page under Role 2.1. The short version: SmartOne already has a small business and healthcare team in France covering Europe, backed by SMARTONE EU SAS registered in Paris. The entity and the team are real and operating. The growth-stage actions are scaling operations behind them: deepening the Safran relationship into a framework agreement, activating the ESSEC alumni network for enterprise accounts, and producing French-language materials to support the sales motion. The full Europe Growth analysis lives on /next.
The activation trigger for France is clear revenue stability in North America. Once survival mode is off, France receives focused attention. If survival mode extends past month four, France slips to month 18 or beyond, and the window does narrow as new European alternatives scale up.
Canada has built one of the most accessible AI R&D funding stacks in the world. SR&ED is a refundable federal tax credit. Quebec’s CRIC (Research, Innovation and Commercialization Tax Credit) stacks on top of it at the provincial level. NRC IRAP provides direct project grants. All of them are available to Montreal-based firms doing qualifying R&D work. SmartOne qualifies on the work it is already doing or would do under this engagement: novel annotation schemas for temporal and trajectory data, automated QA algorithms for multi-frame 3D point clouds, tooling to reduce annotator cognitive load.
Federal Budget 2025 significantly improved the SR&ED terms. The 35% refundable federal ITC rate now applies to the first $6 million in eligible expenditures (doubled from $3M). For a CCPC with $1.5M in qualifying R&D salary spend, the combined yield looks like this:
| Component | Amount | Notes |
|---|---|---|
| Eligible R&D spend (salaries, equipment, subcontractors) | $1,500,000 | 25% of $6M revenue directed to qualifying technical work. Conservative for a data quality firm with active tooling development. |
| Exclusion threshold (12 employees × $18,751) | $225,000 | Quebec CRIC per-employee exclusion |
| Net eligible base | $1,275,000 | |
| Federal SR&ED (35% on net eligible) | $446,250 | Fully refundable. No taxable income required. CCPC required. Source: Leyton Canada, Budget 2025. |
| Quebec CRIC on first $1M of net eligible (30%) | $300,000 | Refundable. Replaces old Quebec provincial SR&ED credit. Source: Ryan Tax, CRIC overhaul. |
| Quebec CRIC on remaining $275,000 (20%) | $55,000 | Refundable above the $1M threshold |
| Total annual SR&ED + CRIC yield | $801,250 | Effective recovery rate on eligible spend: 53%. Both as cash, zero taxable income required. |
NRC IRAP stacks on top of this as direct project funding. First-time IRAP applicants in the Montreal region typically receive $75,000 to $200,000 per project over 6 to 12 months. IRAP reimburses 80% of eligible salary costs for technical staff directly on the project. The AI Assist sub-program within IRAP, a $100M federal commitment specifically targeting generative AI and deep learning firms, has funded 250+ projects in its first year. SmartOne’s Physical AI data specialization fits directly. Source: NRC IRAP official program page.
Mila is a third lever. SmartOne became a Mila industry partner in June 2023, cited as a direct reason for the Montreal move. The existing partnership is a standard membership. Deepening it, through a funded research project on annotation workflow optimization for temporal data or a research chair co-funding arrangement, connects SmartOne to Mila’s graduate pipeline at a moment when Physical AI researchers coming out of Mila directly need annotated training data at scale. It also adds a research credibility signal that differentiates SmartOne in enterprise sales conversations, particularly with buyers doing EU AI Act compliance due diligence.