Module spec

The Viral DNA Engine, specified

The Engine is a closed loop that turns high-volume UGC into a ranked, revenue-attributed format library. It ingests your brand, constraints, tracking setup and content history, tags every asset to a format/angle/hook/CTA/creator, scores everything with the Rabk Rate Score, and publishes exactly 3–5 formats worth scaling. Views are treated as diagnostics; revenue and conversion-chain efficiency decide outcomes.

Inputs & outputs

What goes in, what ships out

Required inputs

A · Brand + Offer
App name, category, pricing model, conversion events (install, signup, trial, purchase, subscription), value props, differentiators, proof points, objections, 2–5 ICP segments with their pains, outcomes and intent lexicon seeds.
B · Constraints
Platforms (TikTok, Reels, Shorts), geos, compliance and brand-safety rules, production budget, turnaround time, creator availability.
C · Tracking setup
MMP or equivalent, UTM conventions, deep links, event schema (install, signup, purchase, trial, subscription, refund), conversion windows D0/D1/D7/D30.
D · Content library (optional)
Past posts with views, watch time and any conversion metrics for warm-start scoring.

Outputs

Viral DNA Report
3–5 winning formats, why they win, and the conversion chain that proves it.
Production Queue
Next 30 days of posts by format, creator, angle and hook.
Experimentation Plan
Hook, angle, offer and landing-flow tests with stop-loss rules.
Dashboard Spec
Rabk Rate leaderboards, conversion-chain views, cohort revenue.
Self-Improving Loop
Weekly and monthly templates that update the format library.

Data model

Every asset links to an outcome

Four linked tables. If a field can't be joined back to revenue, it doesn't belong in a decision.

ContentAsset

id · platform · post_date · creator_id · format_id · angle_id · hook_id · cta_id · length · caption · hashtags · geo · spend

Performance

views · 3s_views · avg_watch_time · completion_rate · shares · saves · comments · profile_clicks · link_clicks

Attribution

installs · signups · trials · purchases · subscriptions · revenue · refunds · D0/D1/D7/D30 cohorts

Derived

CVR_view→click · CVR_click→install · CVR_install→signup · CVR_signup→purchase · revenue_per_1k_views · revenue_per_post · payback_proxy

Rabk Rate calculation

Each component is percentile-normalized within the account, weighted, then mapped onto a 300–850 band so scores stay comparable as revenue scales.

Revenue power — revenue_per_post, revenue_per_1k_views, payback_proxy40%
Conversion-chain efficiency — CVR at each step vs. guardrail floors30%
Volume & stability — N vs. minimum, week-over-week variance20%
Quality & compliance — refund rate, brand safety, geo rules10%

Rabk = 300 + 550 × (0.40·revenue_index + 0.30·chain_index + 0.20·stability_index + 0.10·quality_index)

Decision rules

Codified, not discretionary

These rules run before anyone's opinion enters the room.

Winner definition

Top performers by revenue_per_post AND revenue_per_1k_views, with no conversion step below its guardrail floor, Rabk ≥ 700, minimum N met, and stability across two consecutive weekly windows.

Vanity flag

Views in the top quartile while installs/purchases sit in the bottom quartile → flagged Vanity Format, capped at exploration budget, cut if a second window repeats.

Sample size floor

N = 30–50 posts per format per week as the starting floor, adjusted to traffic. Below floor, the format is 'Insufficient' — never 'losing'.

Volume before micro-optimization

If total volume is under floor, the only prescribed action is raising volume. No edit-level tinkering is allowed to consume the cycle.

Output cardinality

Exactly 3–5 formats are published as Viral DNA. If evidence is insufficient, the Engine emits a Data Gap Plan / 30-Day Signal Plan instead.

Scaling split

70–90% of output to winners, 10–30% to exploration, permanently. Winners are re-verified weekly or lose their allocation.

Workflow

Diagnose → Generate → Launch → Measure → Extract → Scale → Self-improve

01 Diagnose

  • Verify tracking integrity: deep links fire, UTMs consistent, events deduped.
  • Lock event definitions and conversion windows before any posting.
  • Map 2–5 ICP segments to search/intent clusters and objection lists.
  • Build a starter library of 10–20 candidate formats across intent stages.

02 Generate high-intent formats

  • Per ICP, produce angles: problem-aware, solution-aware, comparison, objection, proof, tutorial, story, contrarian, myth-busting.
  • Every format must carry a conversion path and one measurable CTA.
  • Assign IDs at creation: format_id, angle_id, hook_id, cta_id.

03 Launch volume

  • Deploy 600–1,500 posts/month across the creator roster (or proportionally scaled).
  • Reject any asset missing tags — untagged posts are unmeasurable and don't ship.
  • Distribute evenly across formats so no format starves for sample.

04 Measure

  • Daily ingestion of platform performance and MMP conversion events.
  • Rolling 7/14/28-day leaderboards by format, creator, angle, hook and CTA.
  • Cohort revenue at D0/D1/D7/D30 so short-window wins aren't over-credited.

05 Extract Viral DNA

  • Rank by Rabk Rate, filter on minimum N and two-window stability.
  • Publish 3–5 format cards plus creator briefs for replication.
  • Document the failure modes that killed near-miss formats.

06 Scale

  • Shift 70–90% of production to winners with coordinated daily volume.
  • Hold 10–30% for new formats and variants so the library keeps refreshing.
  • Expand winners by variant, not by rewriting the winning structure.

07 Self-improve

  • Weekly: update hypotheses, retire losing variants, expand winning variants.
  • Monthly: refresh ICP insights, landing pages and app-store conversion inputs.
  • Quarterly: rebase Rabk thresholds against the new revenue distribution.

Dashboard spec

Six views, one question: what pays?

Rabk Leaderboards

Format, creator, angle, hook and CTA tables sorted by Rabk with N and stability badges.

Conversion Chain

Funnel per format: view→click→install→signup→purchase with guardrail floors marked.

Revenue Cohorts

D0/D1/D7/D30 revenue and refunds by format and post date.

Vanity Radar

Scatter of views vs. revenue_per_1k_views; the high-view/low-revenue quadrant is the kill list.

Volume Monitor

Posts/day vs. required floor per format, with an alert when sample size is at risk.

Experiment Board

Live tests, window, decision date, and automatic stop-loss status.