Research
Approved source pack, dated references and product facts.
Output: evidence notes + gapsHEAD OF ACADEMYINTERACTIVE BUSINESS CASE / 2026
Learning that
earns its place.
My blueprint for an AI-first Academy.
Built around learner needs. Connected to business results.
THE ACADEMY LOOP01 / 04
01
Understand the learner
Profile, goals and behaviour
02
Give the right next lesson
Two audiences. Individual paths.
03
Check what changes
Knowledge, decisions and habits
04
Measure business value
Retention, acquisition, contribution
Learn. Observe. Adapt. Repeat.↻
90days
Working MVP
Proposed launch horizon
+10%
Trader retention
12-month employer target
+10%
IB trader acquisition
12-month employer target
+20%
Exclusive IB agreements
12-month employer target
A proposal, not a record of delivered trading results. Interactive demos use fictional data and transparent rules. Targets are from the role brief; definitions would be agreed at discovery.
01 / THE LEARNING PRODUCT
One Academy.
Two paths to value.
Start with a clear outcome at each stage. Click a stage to see what I would teach, assess and track.
MVP boundaryBuild the first three stages for traders and IBs. Map the full pathway. Strategy automation and MT5 / TradingView compatibility belong to a later, separately validated phase.
02 / PERSONALISATION LAB
The next lesson
has a reason.
Change a learner profile and see the path change. This demo uses visible rules, not a live AI model or trading feed.
LEARNER SIGNALS
Illustrative signals only. A real system would check data coverage and freshness before making a recommendation.
Broker profile + permitted behaviour dataValidated features + rulesLearning platformLearning events + CRM
Production proposal: use structured rules for eligibility and safety, then AI to explain approved recommendations. Keep user IDs pseudonymous and raw trading records outside public AI tools. No live integrations are connected here.
03 / WORKING CONTENT SAMPLE
Risk before return.
A complete micro-lesson: diagnose, explain, practise, assess and apply. All trade examples use a fictional demo account.
Educational demonstration, not a trade recommendation. The 1% rule is an exercise assumption, not a suitable risk level for every person. Concept reference: CME Group: position sizing. Examples and questions are original.
04 / IB CONTENT THAT CAN BE USED
One lesson.
A useful partner toolkit.
IBs are introducing brokers: partners who refer clients. Give them clear education and practical distribution tools, with a way to follow the result.
Campaign: Before your first trade
Audience: new learners comparing platforms.
Goal: start a demo learning path and complete the first lesson.
Attribution planPartner ID + asset ID + campaign ID → first-party event → CRM lead → funded trader. Agree a 30-day window, deduplicate by customer ID, and check incrementality in an IB-level pilot. Attribution alone does not prove impact.
05 / THE BUSINESS CASE
Make the economics
visible.
Change the assumptions. Every number updates. These are illustrative USD scenarios, not a supplier quote or an earnings forecast.
01 · Retention value
02 · IB acquisition value
03 · Platform, people & operations
MODELLED ANNUAL NET VALUE
Return on Academy cost
Retention margin and new-trader margin must cover the same 12-month value horizon. Year-one benefit factor discounts the steady-state estimate for launch timing and reach.
Open the calculation and measurement assumptions
Retention
Eligible cohort × baseline retention × relative uplift = additional retained traders.
Example: 40% retention with a 10% relative uplift becomes 44%, a gain of 4 percentage points. The calculator caps retention at 100%.
Acquisition
Participating IBs × baseline annual registrations per IB × relative uplift × funded conversion × annual contribution per new funded trader.
Use new customers only, outside the existing retention cohort. Margin is after variable servicing costs and partner commission.
Cost and ROI
12 × recurring monthly costs + initial integration and content costs, then contingency. Net value = adjusted contribution − cost. ROI = net value ÷ cost.
Exclusive-partnership value is tracked separately and is not added, to avoid counting the same trader contribution twice.
Platform cost uses billable provisioned seats, not course completions or active users. The example seat price is an editable assumption, not Sana pricing. The financial cohort must match people reached by the intervention over the year; if seats are reused, validate provisioning rules and the quote.
The model excludes tax, financing and discounting. It is a decision aid, not causal evidence. Agree retention definition, margins, launch exposure and target interpretation with Finance before approval.
06 / HOW I WOULD KNOW IT WORKS
Measure the change.
Then decide.
A weekly operating view for the Academy lead. A monthly contribution review with the CHRO and Finance.
ILLUSTRATIVE DASHBOARD · FICTIONAL DATA · NOT LINKED TO THE ROI MODEL
| Learning asset | Learners | First-attempt mastery | Signal | Next decision |
|---|---|---|---|---|
| Risk before return | 120 | 76% | Many retry the sizing task | Improve example |
| Demo platform basics | 94 | 89% | Strong task completion | Keep monitoring |
| IB content toolkit | 24 | 83% | Few tracked campaign links | Fix distribution |
01 / DEFINE
Agree the baseline
Define eligible traders, retention windows, active status and contribution. Record the pre-launch baseline; separate geography, tenure and experience.
02 / TEST
Compare fair groups
Where feasible, randomise eligible traders to an Academy invitation or delayed invitation. Analyse everyone as assigned. Randomise IBs by partner to reduce content spillover.
03 / INTERPRET
Do not confuse cause
Engaged users may already be more motivated. A matched comparison is a fallback, not proof. Report group sizes, uncertainty and possible bias.
Day 90 can show activation, learning and early behaviour. It cannot establish a 12-month retention effect. Estimate the sample size from baseline and detectable effect before the pilot; 200 traders / 30 IBs would be a usability pilot, not automatically an impact test.
Event model, quality checks and decision cadence
Event contract
learner_id, event_id, timestamp, content_id, content_version, experiment_group, attempt_number, source, partner_id where relevant.
Events: assigned, started, answered, completed, recommended, content_used. CRM owns funded / active status.
Daily data checks
Deduplicate event IDs. Monitor missing joins, delayed feeds and unexpected score changes. Log recommendation rule versions. Do not infer behaviour from missing data.
Operating rhythm
Daily: integration failures and safety flags. Weekly: learning and content review. Monthly: Finance checks cost and contribution. Quarterly: curriculum and platform decision.
07 / A SOLO LEAD, WITH A SYSTEM
Automate production.
Own the decisions.
I would build a source-based workflow, with clear inputs, outputs and release checks. AI drafts and reviews; a named owner approves publication.
Learning design
One observable outcome, a worked example and assessment plan.
Output: module specificationProduction
Lesson, scenario, quiz and an IB distribution kit.
Output: versioned draftsQuality checks
Check calculations, sources, answer keys, language and claims.
Output: errors + release checklistApproval
Academy owner approves. Broker specialists review high-risk claims and local rules.
Output: approved releaseImprovement
Review first attempts, feedback and behaviour signals.
Output: next change + testExample agent brief and quality gate
Content agent brief
Create a 6-minute lesson for a beginner. Outcome: calculate planned loss in a demo trade. Use only the approved source pack. Return: source IDs, lesson, worked example, 3 questions, answer keys and feedback. Use plain English. Mark unsupported claims. Do not promise returns or recommend a trade.
Release gate
- Each factual claim links to a reviewed source.
- All arithmetic is recalculated outside the LLM.
- Each question tests the stated outcome.
- Incorrect answers receive specific feedback.
- Risk and market-specific claims have a named approver.
- Version, review date and rollback copy are stored.
Capacity assumptionInitial scope: six short modules, one language, two audience paths and one IB kit. Budget shared broker support for data access, security and financial-content approval. AI can reduce production work; it does not remove these dependencies.
08 / TECHNOLOGY CHOICES
Buy the learning layer.
Own the logic.
Sana Learn is a candidate for evaluation. The decision depends on external-user economics, the API proof and the learning experience.
Sana Learn
AI authoring, learning paths and learner support in one platform. Its published model supports external learners and charges for provisioned users, with plans starting at 300 users.
Pricing & external-user modelThe documented REST API and xAPI events to webhooks make it a candidate for a connected learning layer. A ready-made Tradefora integration is not assumed.
Integration documentationBehaviour dataTradefora or broker data exports / APIs. Validate access, fields, consent and update frequency.
OrchestrationEvaluate n8n or Make. Use retries, event IDs, logging and a manual fallback.
AI content layerSelect a provider after source-quality, data-handling and per-lesson cost tests. Model choice remains open.
CRM + analyticsExisting broker CRM as the commercial source of truth. Join learning events in a warehouse; report in Power BI or Looker.
Choose Sana if…
A sandbox proves assignment + event export, external SSO works, and total cost fits the approved margin model.
Choose another LMS if…
Per-seat cost is too high for occasional learners, export is limited, or external access adds friction. Compare active-user and usage-based quotes.
Custom build only if…
Critical journeys cannot be supported by a learning platform, and maintenance capacity is funded. Avoid building a whole LMS inside the 90-day MVP.
Vendor facts checked against the linked official pages on 4 October 2026. Pricing and integration scope require a written quote and proof of concept.
09 / THE FIRST 90 DAYS
A launch plan
with decision gates.
Start small enough to operate alone. Make data access and approval dependencies visible from week one.
01–15 DAYS
Understand
Interview Sales, traders and IBs. Map journeys, agree baseline and financial definitions. Confirm data owners and budget.
Exit gateApproved problem, target cohorts, data access and named reviewers.
16–30 DAYS
Design & prove
Draft curricula and assessments. Test platform provisioning, course assignment and event export with sample records.
Exit gateOne working integration, platform quote and approved sample lesson.
31–60 DAYS
Build the MVP
Create six modules and the IB kit. Set up rules, event tracking, CRM links, a dashboard and support workflow.
Exit gateComplete a learner journey end to end; reconcile learning and CRM events.
61–90 DAYS
Pilot & release
Run a small usability pilot. Fix friction and content errors. Start the agreed impact experiment and launch within scope.
Exit gateLive MVP, cost report, owner runbook and next-quarter test plan.
Day 90
Two working audience paths.
One measurement system. A clear next investment decision.
12+ years
Marketing and audience growth
Experience turning audience needs into content, campaigns and measurable funnels.
2GIS
Analytics built around decisions
Built a recruitment-funnel dashboard with stage conversion, rejection reasons and cost metrics. Power BI, SQL and ATS data.
3 employers
Onboarding collaboration
Participated in working groups creating or improving welcome books and onboarding materials.
LMS + content
Practical learning operations
Course assignments, content uploads and communications at DataArt and 2GIS; own Employer Brand course and AI-assisted content projects.
What I would need to prove early
Broker-domain knowledge, production-grade trading-data integrations and the reliability of the AI workflow. This prototype demonstrates my approach; it does not claim that I have already run a trading Academy.