Proof, in layers
Read the work. Don't take our word for it.
Most consultants have a logo wall and a deck. We have a directory. Below: hard outcome numbers from current Onwards engagements, anonymised case studies, the twenty-plus active builds in flight this month, the way we take POC startups from idea to live, and links you can verify in five minutes. Skip to the section that matters. Or read the whole thing if you're the kind of buyer who reads the whole thing before they pay.
All client names below are anonymised. The patterns and numbers are real. If you need named references, we can provide them under NDA in a scoping call.
- The numbers, outcome metrics from current engagements
- Anonymised case studies, six recent engagements with situation, approach, and result
- Currently shipping, twenty-plus active builds in flight this month
- How we take POC startups from idea to live, the five-phase method we run
- The capability stack, what we use, where it came from, why it works
- Verify it yourself in five minutes, direct links to the work
- The recursive proof, what you're reading right now is one of these
Layer 1 · the numbers
The model already runs at scale.
These are real outcome numbers from current Onwards engagements. They're already public on the company site (linked at the bottom of this page) so you can cross-check. Anonymised by client below.
10x
Output increase against an incumbent vendor at 50% of their cost. Same operator. Different model.
32x
Output increase year-over-year at 20% of original cost. The 65th outcome took a fraction of the time the 5th did.
3
Simultaneous workstreams in year two of the same engagement. Speed higher than year one. Cost per outcome lower.
Layer 2 · case studies
Six recent engagements, anonymised.
Same shape each time: situation, approach, outcome. Some are years deep. Some are last month. All real. All happening alongside the team you'd be hiring.
Cost parity wasn't achievable. It was beatable.
The brief from the procurement team was simple: don't spend more than the incumbent vendor, and get dramatically more from it.
Situation
Incumbent BI consultancy charging full enterprise rates for a fixed catalogue of report-build outcomes. Client wanted same throughput, lower cost.
Approach
Senior architecture in-region, execution distributed offshore. Outcome-based pricing tied to discrete logged deliverables, not hours.
Outcome
20 discrete outcomes at 50% of the incumbent's price. Output 10x. Knowledge captured into the engine for the next engagement.
Budget pressure hit. We delivered 32x the output.
Same client, year two. The procurement environment tightened mid-engagement.
Situation
Renewal conversation under budget squeeze. Same operator, same mandate, lower ceiling.
Approach
Compounding-knowledge model. Each new outcome built on the previous one. Reusable templates and scripts replaced bespoke builds.
Outcome
65 outcomes delivered at 20% of the original cost. Output 32x. Capacity per dollar climbed every quarter.
A founder with a vision. We built the engine around him.
A domain expert had the product idea and the technical skills to build it. What he didn't have was the structural conditions to actually ship it sustainably while also serving paying clients.
Situation
Founder pulled between client work and product work. Both suffering. Personal capacity at the limit.
Approach
The Six-Lever Engine wrapped around his time. Context defined. Work structured. Economics aligned. Senior judgment isolated as the scarce resource.
Outcome
Platform now in production at a global mining major (5 modules live). Founder builds what he wants. The engine handles the rest. Equity stake retained.
Quick Wins workstream inside an Operating Model Program.
An ASX-listed lithium operation running a multi-stream Operating Model Program. We took the workstream that had to ship something real every fortnight while the bigger streams were still in design.
Situation
Three workstreams running concurrently. Two were strategic and slow. The third had to demonstrate value early to keep the program funded.
Approach
Ryan as workstream lead, 3-4 days/week. Business planning framework, monthly priority process design, SharePoint technical solution, all designed and shipped in parallel.
Outcome
Currently active. Drafts shipping every fortnight. Program retained funding. The other two streams now reference our process designs.
A vet-tech marketplace from hand-coded MVP to launch-ready in 5 phases.
A two-co-founder startup (one a domain expert, one operator) with an MVP that worked but was a security and ops liability before going to market.
Situation
MVP shipped on Lovable. Critical security holes (self-assigned roles in user_roles table). No SMS lifecycle, no email lifecycle, no payment retry, no E2E tests.
Approach
Five-phase plan: Security → Integrations → Notifications → UX Polish → Launch Prep. 19 polished tickets across 5 phases. Server-side role enforcement via PostgreSQL trigger. Stripe authorise-then-capture, signature-verified webhooks, GA4 funnel tracking.
Outcome
All 19 tickets ✅. 22+ Playwright E2E tests. RLS audited and patched. 387-line deployment guide. Production-ready, currently in user acceptance testing.
A swimming AI from spec to launch-ready in three months.
A consumer AI app for swimmers. Upload a video, get pose estimation, biomechanical metrics, AI coach feedback.
Situation
Solo founder, ambitious scope. Pose estimation, 11 metrics, multi-coach prompt customisation, premium-tier gating, video transcoding pipeline. No clear MVP boundary.
Approach
v1 plan stripped scope to core: pose pipeline, AI coach, Stripe tiers, premium gating. Killed wearables, killed gamification, killed RevenueCat. Replaced with Lovable AI gateway + Gemini 2.5 Flash + Firebase Cloud Functions for pose.
Outcome
Live web app. Auth flow, 8-step onboarding, dashboard with goal progress, video pipeline, premium gating, Stripe checkout + billing portal. 14 DB migrations across 4 months. 3-tier pricing live.
Layer 3 · currently shipping
Twenty-plus active builds, this month.
Below is a partial list of what's live or in flight on the team's drive right now. The breadth is the proof. This is what becomes possible when senior judgment is the bottleneck and AI handles the hands-on work alongside us.
The funnel layer
Built ground-up in the last 30 days. Most of this is the same machinery you're touching right now.
- This Clarity Library funnel. Cold-traffic written-memo product line. Built end-to-end in 4 days: 8 paid offers, 3 tripwires, 24 sample memos, inline Stripe checkout, post-purchase intake, automated email confirmation, referral codes, Hormozi 4-pillar money model, 5-agent CMO oversight system. You're reading one of the surfaces.
- Automated Meta ad creative engine. Generates 50-200 scored ad variations from one offer brief. Runs them through a multi-armed bandit, identifies winners, kills losers automatically. 158-ad cold campaigns built and deployed in a single tick.
- Autonomous micro-offer generator. Input a customer problem or theme. 1.5-3 hours later: 3-5 ready-to-test offer specs with full ICE scoring, ad creative clusters, landing-page outlines. Used to generate the offer ladder you're looking at.
- Automated landing-page bandit. Multi-armed bandit serving 5+ page variants per traffic source. Cloudflare Worker edge serving + KV state. Tested 12 copy variants × 3 designs × 3 prices × 6 CTAs in a week.
- Marketing Foundations system. A complete SLO funnel architecture, 30-day cold-traffic playbook, ICP language banks (3 ICPs deeply documented), Stage Guides, Meta creative research libraries.
The autonomous business layer
The systems that run the company while we focus on the work.
- Paperclip: a 16-agent autonomous back office. Project management, 7 directors per product line, 5 specialists, 3 support roles. Telegram bot for command + control. Evening batch jobs. Daily morning briefings at 8am. Replaces roughly 3 ops hires.
- BI ticket automation pipeline. Inbound BI requests (Slack, email, Jira) auto-classified, auto-routed to the right team or agent, status auto-tracked, SLA auto-monitored. The mining-major engagement runs on this.
- Hindsight: vector-indexed memory across networks. World-facts, observations, experiences, mental-models. Lets every new project draw on every prior one. Compounding knowledge made structural.
- ClickUp automated data export pipeline. Workspace-level export across all spaces, folders, lists. Used for ClickUp → Jira migration and historical reporting.
- Microsoft 365 MCP server. Built and shipped a Model Context Protocol server for Microsoft 365 integration. Published to npm. Used internally for inbox + calendar workflows.
- Teams bot for client integration. Dockerised Microsoft Teams bot for client-side notifications and task triggers.
The product layer
The bets that turn services into IP.
- An organisational assessment platform deployed at a global mining major. 5 modules live in production sandbox. Working with the client's engineering lead on rollout. Replaces a $150k/yr legacy compliance tool.
- A pre-launch SaaS for compliance teams. Currently in security-hardening sprint before public demo. Three-sprint launch plan: Make it safe to show → Make it observable + onboardable → Make it sellable.
- A consolidated reporting POC for the resources sector. Demonstrating how 3 legacy systems (gap assessment, improvement tracking, BI reporting) consolidate into one managed platform.
- A $9 self-serve Power BI health-report tool. Upload a .pbix file. Get a 5-dimension health score and risk-tiered issue list in minutes. Designed for the BI manager, not the BI builder.
- A dynamic SLO management system. Cloudflare Workers + KV bandit infrastructure. Used for landing-page testing and offer testing.
The startup builder layer
Currently shipping for two startups, with capacity for one more in late 2026.
- Vet-tech marketplace startup (Perth). Two-co-founder team. 19 polished-MVP tickets shipped across 5 phases. RLS audit + server-side role enforcement. Stripe authorise-then-capture. Resend lifecycle email (5 templates), Twilio SMS lifecycle (9 templates), 22+ Playwright E2E tests, 387-line deployment guide.
- Sport-tech AI startup (consumer). Web app, pose estimation pipeline (Firebase Cloud Functions), 11+ biomechanical metrics, AI coach via Gemini 2.5 Flash + Lovable AI gateway. 14 DB migrations across 4 months. 3-tier Stripe pricing live.
The knowledge layer
The IP that compounds outside any single engagement.
- A book in progress (~90% complete). "They Don't Get It." On the gap between business and technical leaders. The translation layer that should have existed fifteen years ago. Used as the voice anchor for every memo and ad we ship.
- A 21-skill autonomous book-writing pipeline. Genesis system used to build the book itself. Skill-by-skill orchestration. Same architecture used for the memo generator.
- An automated business goal plan generator. April 2026 plan was Claude-generated, then human-edited. Used for monthly Onwards strategic reviews.
- A "30+ founder" content series + lead-magnet system. Content infrastructure, Notion prompt DB, automated batching, quality scoring against a 96/100 bar.
The craft layer
The small but compounding tools that close the gap between intent and shipped.
- Figma rebuild after dev-to-design drift. Once a sprint ships, we rebuild the design from the live code, diff it against the original Figma, and ticket the gaps automatically. Stops the "the build doesn't match the design" loop most teams just live with.
- Screenshot-to-Figma reverse engineering. Take a screenshot of a competitor or reference design. Get back a structured Figma file ready to fork.
- Voice-DNA matching for content. Manuscript-derived voice fingerprint. Used to score every piece of generated copy against Ryan's actual voice patterns. Anti-AI 20-pattern density scan with hard caps.
Layer 4 · the POC startup method
How we take a POC startup from idea to live.
Most fractional CTO engagements either ship slow because they're trying to be perfect, or ship fast because they're skipping the parts that bite later. The model we run does neither. Five phases. Each one has a gate. Each gate is verifiable. Tested across two live startups in the last quarter.
Phase 1 Spec + scope freeze
Single-source-of-truth spec doc. What we're building, why, who buys it, what the pricing is, what's in MVP and what's deliberately out. Every decision documented with the reasoning. If it's not in the spec, it doesn't get built. Two-page maximum. We rewrite it until both founders nod.
Phase 2 Make it safe
Before any feature work. Audit the auth flow. Audit RLS policies on every user-facing table. Server-side role enforcement (never client-side). Bearer-required edge functions. Stripe webhook signature verification. Manual security tests scripted into the CI pipeline. The product can be demoed safely after this phase.
Phase 3 Build the core flow
The single user journey that defines whether the product works. For the vet marketplace: book → pay → vet shows up → review. For the swim AI: upload → analyse → coach feedback → upgrade prompt. We build that loop end-to-end before anything else. If that flow doesn't convert, no other feature matters.
Phase 4 Make it observable + onboardable
Sentry on frontend + edge functions. PostHog or GA4 with named events covering the full funnel. New-user dashboard never empty (sample data on first visit). Loading skeletons + empty states everywhere. Account deletion that actually purges. The product can survive 5-10 beta users without us watching.
Phase 5 Launch readiness
E2E tests for every critical path. Stripe live mode + production webhooks. Comprehensive deployment guide written. Lifecycle email + SMS templates wired. Mobile responsiveness audit. Payment retry flow. Refund SOP. Then, and only then, ship to live traffic.
Each phase has a written gate that must be true before the next phase starts. The gates are public to both founders. No moving forward on vibes.
Layer 5 · the capability stack
What we use, where it came from, why it works.
A short, honest inventory of the tools we lean on. The point isn't that we use them. The point is that we've used them on enough live builds to know exactly which problems they solve and which ones they don't.
- React 18 + TypeScript + Vite. Default startup stack. Used on every fractional CTO engagement.
- Tailwind + shadcn/ui. Component library that lets a senior dev or designer ship production-grade UI in days.
- Lovable. Rapid scaffolding for POC startups. Get to a working prototype in hours, not weeks. Then layer the security, observability, and ops on top.
- Cloudflare Workers + KV + R2. Edge-served LP infrastructure. Multi-armed bandits at the edge. The Clarity Library you're touching runs on this.
- Supabase (Postgres + Auth + Storage + Edge Functions). Default backend for both startup engagements. RLS as the security model. Triggers for server-side enforcement.
- Power BI + DAX + Azure Synapse + Databricks. Enterprise BI stack. 15 years of accumulated patterns, scripts, templates, and gotchas. Where we beat incumbents on cost.
- Stripe (Embedded + Elements + Payment Links). Payments + subscriptions across every product. Authorise-then-capture, signature-verified webhooks, server-side promo codes.
- Anthropic Claude (API + Code). The hands-on partner for everything we build. Skills + sub-agents + MCP servers. The breadth above is what becomes possible.
- Gemini 2.5 Flash via Lovable AI gateway. Used for in-product AI features (coaching, classification, summarisation) where Claude is overkill on cost.
- OpenAI embeddings. Vector search for the Hindsight memory system + the Clarity chatbot.
- Resend (transactional email) + Twilio (SMS). Lifecycle messaging across every paid product.
- Meta Marketing API + Pixel + Conversions API. Server-side CAPI dedupe. Custom Audiences via API. Lookalike automation.
- Sentry + PostHog or GA4. Error tracking + funnel analytics. Standard for every Phase-4-onwards build.
- Notion + ClickUp + Jira + Telegram. Project management spine. Paperclip routes commands across all four.
Layer 6 · verify it yourself
You don't have to trust any of this.
Five direct links. Five minutes total. Each one is a piece of the work above, live, that you can poke at right now.
Five-minute verification
- Read 24 actual sample memos. Same structure, same voice, same depth as the paid version. Pick the one closest to a decision you're sitting on. Browse all 24 →
- Talk to the chatbot. It runs on the same library that backs every paid memo. Ask it the actual question you've got. Open the chat →
- See the Onwards Analytics company site. The numbers in Section 1 above are public there. onwardsanalytics.com.au/proof →
- See Ryan's personal site. Years, capacity, current commitments, the book. ryanlikes.tech →
- Read the book in progress. "They Don't Get It." 90% complete. See the book →
Layer 7 · the recursive proof
You're reading one of these right now.
Ready to send a decision?
Pick the offer that fits. Or start with a $19 sniff-test on the same decision before committing $297.
See the offers → Or just the $297 →