Business Requirements Brief For review & alignment v0.1 · Jul 2026

Marketing that learns from the procedure, not the lead.

A system that runs our ailment-category acquisition end-to-end — landing pages, Google & Meta campaigns, lead handling — and continuously reallocates money toward what actually produces procedure margin. People stay in charge of medicine, claims, and strategy. The machine handles the daily grind of optimization.

Working title · The Margin Loop
Owner · Growth
Read time · ~8 min
Ask · 6 decisions, p.9
Why now

We buy leads well. We can't yet buy margin.

Our campaigns are steered by cost-per-lead, because that's where the ad platforms' visibility ends. Everything that decides whether a lead was worth buying happens afterwards — in calls, clinics and invoices — and none of it flows back into the buying.

Optimization stops at the form-fill

Two keywords with identical CPL can differ 4–5× in procedures produced. Today they look identical in every dashboard we steer by.

Margin is invisible to marketing

Procedure revenue and cost sit in CRM and finance systems, disconnected from campaign IDs. "What margin did this ad group make?" is a manual project.

Launches take weeks

Each new ailment needs a page, copy, keywords, tracking and compliance review — coordinated by hand, every time.

Learning doesn't compound

What works lives in people's heads and old decks. When a person moves on, the learning goes with them.

The proposal

Close the loop from click to margin — then let the loop steer the spend.

Today — feedback stops early
Ad spend Clicks Leads feedback ends here Appointments Procedures ₹ Margin

Bidding, budgets and creative choices are tuned to whatever makes leads cheap — including leads that never pick up the phone.

Ad spend Clicks Leads Appointments Procedures ₹ Margin

Every completed procedure, no-show and junk lead is traced back to the exact keyword, ad, audience and page that produced it — and the next day's bids, budgets and creatives move accordingly.

How it works

One loop, five stages, running continuously.

This is the whole system on one diagram. Three stages are machine-run, one belongs to people, one is shared bookkeeping. The centre is the only number the machine is allowed to chase.

THE NEW PART OPTIMIZES FOR ₹ margin not cost-per-lead SYSTEM 1 · Create pages + Google/Meta campaigns SYSTEM 2 · Capture every lead, source-tagged PEOPLE 3 · Convert calls → appointments → procedures CRM + FINANCE 4 · Record outcomes, revenue & cost per case SYSTEM 5 · Learn shift money to what earns
Machine-run People-run Shared bookkeeping
The economics

What one funnel looks like when the whole thing is measured.

Ad spend
₹70,000
Clicks
1,000
Leads
90
Qualified
45
Appointments done
14
Procedures
4
Contribution margin
₹1,60,000
Margin-ROAS
2.3× margin ÷ spend
Margin per lead
₹1,778
Allowable CPL at 2.0× target
₹2,050

Illustrative numbers · bar widths compressed for readability · targets set per ailment category

Keyword A · symptom browsing
"knee pain reasons"
Cost per lead₹800
Leads → procedures2%
Margin per lead₹800
Keyword B · treatment intent
"knee replacement cost bangalore"
Cost per lead₹800
Leads → procedures9%
Margin per lead₹3,600

Same CPL, 4.5× difference in what the business earns. Today both keywords look identical, so budget splits evenly. In the loop, money migrates to B automatically — and A gets a cheaper page-and-nurture treatment instead of premium bids.

Division of labour

The machine grinds. People decide.

The system does

Daily, within hard limits
  • Builds & tests landing pages and ad variants from pre-approved content blocks
  • Moves budgets and bids toward the highest marginal margin, across Google, Meta, categories and cities
  • Mines keywords — promotes converters, blocks junk intent
  • Prioritizes leads so agents call the highest-value ones first
  • Feeds margin values back into Google's bidding; runs its own downstream optimization for Meta
  • Watches itself — anomaly alerts, auto-rollback, full action log

People do

Always
  • Approve every patient-facing claim — medical & legal review before anything goes live
  • Talk to patients — qualify, book, remind, care
  • Consult and treat — doctors decide medicine, full stop
  • Set the money rules — budgets, caps, margin targets, category go/no-go
  • Approve big moves — new ailments, new campaign structures, anything outside bounds
  • Hold the kill switch — one button freezes all automation
Never automated: medical claims, medical advice, pricing, and anything a patient reads that hasn't passed human review.
What it adjusts

Seven levers, one objective.

L1

Bidding

Bid targets tuned to margin, not lead count — nudged daily within bounds.

L2

Budget allocation

Money flows to the campaign, channel and city with the best next-rupee return.

L3

Ad creatives

Winning angles scale, tired ones retire, new variants enter through review.

L4

Targeting & keywords

Converting queries promoted; junk intent ("free", jobs, DIY) blocked.

L5

Landing pages

Headlines, proof blocks and forms tested continuously; winners become defaults.

L6

Lead prioritization

High-value leads reach agents in minutes; contact timing learned per segment.

L7

Ailment portfolio

Monthly scale / fix / kill view per category and city — decided by people.

Built for healthcare

Guardrails first, automation second.

This is health advertising in India. The rules aren't a checklist at the end — they're built into what the system is physically able to do.

Human review on every claimNothing a patient reads goes live without medical & legal sign-off. Permanent — not a phase.
Indian health-ad rules encodedRestricted categories and prohibited claims (D&MR, ASCI, medical-ethics rules) are hard blocks, not guidelines.
No health data leaves our systemsAilment or condition information is never sent to Google or Meta, and there is no condition-based retargeting. Patient data stays consented, encrypted, DPDP-compliant.
Money can't run awaySpend caps, daily change limits, auto-rollback on bad moves, and a one-button kill switch that freezes all automation.

One platform reality we've designed around

Google lets advertisers feed real downstream value (our procedure margin) back into its bidding — so on Google, the loop runs end-to-end inside the platform's own optimizer.

Meta now restricts health advertisers from optimizing on downstream events like bookings. So on Meta we capture leads inside the platform's own lead forms, and the downstream optimization — which ads and audiences produce margin — happens in our system, using our CRM truth. Different plumbing, same loop.

What success looks like

Four numbers we'll be judged on.

≥2.0×
Margin-ROAS — contribution margin per rupee of ad spend, per category
<3 days
To launch a new ailment funnel — page, campaigns, tracking, compliant
≥85%
Of procedures traced back to the exact ad, keyword and page
+15%
Margin-ROAS vs. a human-managed control slice that runs alongside, always

Non-negotiable guardrail: zero ad-policy strikes and regulatory complaints. A system that earns more by cutting corners has failed.

Path forward

Autonomy is earned in phases, not switched on.

Phase 0 · Weeks 1–6

Wire it up

Track every rupee from click to procedure margin for 2–3 pilot ailments. Manually-run campaigns; the system watches and reports.

Observe
Phase 1 · Weeks 6–12

Close the loop

Margin values start steering Google bidding. The system recommends budget, bid and keyword moves; people apply them.

Recommend
Phase 2 · Months 3–6

Assisted autopilot

One-click apply with approval trails. Page and ad testing runs at scale from pre-approved blocks. Page factory live.

Approve & apply
Phase 3 · Months 6–12

Autopilot in guardrails

Routine optimization is automated within caps. People steer strategy, medicine, money — and review everything it did.

Auto, bounded
What we need from each team

The loop only closes if every stage feeds it.

TeamWhat we're asking for
MarketingPick pilot categories; agree to keep a 10–15% human-managed control slice so we can measure the machine honestly.
Medical & LegalOwn the approved-claims library and a review turnaround of under one business day.
Lead OpsCoded outcome on every call (no free-text-only), first call within 5 minutes of lead creation.
FinanceSign off one margin formula per procedure type; supply revenue & cost within a week of each procedure. Set caps.
Clinic OpsAppointment capacity signals per city — so we never buy demand a clinic can't serve.
Tech / DataCRM and tracking integration in Phase 0; this is the critical path.
Decisions needed in this review

Six calls to make before we build.

1

First pilot ailments & cities

2–3 categories that are legally clean, have real margin, and short-enough procedure cycles to learn fast.

2

Margin definition

Which cost lines count (doctor payout, consumables, facility)? Finance owns the formula; everything optimizes to it.

3

CRM & calling setup

Which CRM, and in-house vs. outsourced calling — this decides how clean our outcome data is.

4

Targets & caps

Margin-ROAS target per category and the monthly spend ceiling the system operates under.

5

Domain strategy

One site vs. per-ailment microsites — this changes how exposed we are to Meta's health-advertiser restrictions.

6

Automation comfort

Endorse the phase ramp (Observe → Recommend → Approve → Bounded auto) and who holds the kill switch.

Eyes open

The four risks we take seriously.

Ad platforms tighten health rules further
Designed-in: Meta already treated as restricted; Google loop degrades gracefully to lead-stage signals; compliance pipeline keeps accounts clean.
Dirty outcome data poisons the learning
Mandatory coded dispositions, data-quality checks that pause automation when feeds break, junk-lead detection feeding back as negative signal.
Compliance or reputational slip
Human review on all claims, restricted-category hard blocks, audit trail on every action, periodic legal re-review of live content.
Over-automation incident
Bounded daily moves, spend caps, auto-rollback, permanent human control slice, one-button kill switch.