H2Tab has 9 personas on paper and $838K of L90 ad spend on the ground. This audit compares what's been built against what the engine would build, names every gap that needs to close, and ends with the actual May roadmap that flows out of it.
The pattern: 9 personas exist in Milanote but only 7 have ever produced an ad in market. 62% of spend lands on one micro. 19% of spend can't be classified against any persona. The engine fixes this with structure, automation, and an explicit refresh cadence. This is what changes.
1
Current State · What's Been Built
Where the 9 personas came from, what's in market today, where the data takes us
Origin of the Current 9 Personas · Strategist to fill in
We have 9 personas in Milanote. We do not have documentation of what data, research, or interview informed them.
Before we can validate or revise the persona set, we have to understand its basis. The audit's recommendations downstream change depending on whether these were built off real customer data or off founder instinct alone.
Questions for the strategist:
· Was customer review data analyzed? Whose, and when?
· Was there a founder interview or category scan that fed the personas?
· Were past creative winners used to back-fit personas, or were personas built first and creative second?
· When were these personas last refreshed?
· Are there personas the founder believes in that aren't on the list?
What the 9 Personas Look Like Against L90 Signal
| Milanote Persona | Status | #Ads | Spend (L90) | % Spend | CPA | ROAS | Read |
|---|
| Trust-Driven Optimizer (Brecka/Rogan) |
Working |
43 | $517,695 | 61.7% | $41 | 1.36 |
The engine of the account. Concentration risk: any fatigue here collapses everything. |
| Committed but Stuck (Founder + Testimonial) |
Working |
6 | $38,696 | 4.6% | $54 | 1.78 |
Underspent. Contains Testimonial-Video04 (3.18 ROAS) and MOFUFounder (2.82 ROAS). |
| Trend-Burned (Tweet + Mechanism) |
Working |
5 | $19,004 | 2.3% | $36 | 1.74 |
Tweet statics + 3D animation. Tiny spend, strong efficiency. Room to scale. |
| Cost-Conscious Hydrogen Curious |
Modest |
4 | $30,052 | 3.6% | $49 | 1.25 |
Sale statics 3.25 ROAS but Kits BOFU only 1.07. Mixed within the persona. |
| Active Recovery Seeker |
Underperforming |
3 | $12,090 | 1.4% | $65 | 0.94 |
Small sample, sub-1.0 ROAS. Wrong creative or wrong persona — strategist call. |
| Aesthetic Optimizer (Bethany/Mazza/Creators) |
Failing |
18 | $63,685 | 7.6% | $87 | 0.66 |
Worst cluster. $64K spent for 0.66 ROAS. Wrong persona or wrong creative — strategist call. |
| Relief Seeker (Culver/Thurlow/Demeri) |
Blocked |
5 | $12,969 | 1.5% | $95 | 0.57 |
Inflammation language banned by client. Persona built on words we can't say. |
| The Athlete |
Never in market |
0 | $0 | 0% | — | — |
Listed in Milanote, no ad ever produced. Net new test candidate. |
| Lapsed Supplement Buyer |
Never in market |
0 | $0 | 0% | — | — |
Listed in Milanote, no ad ever produced. Net new test candidate. |
Plus 19 ads ($156K · 18.6% of spend · 1.41 ROAS) that don't fit any of the 9 personas. Includes "103-Copy" UGC, Hydrogen Gas content, Instagram-organic format. Either retag into existing personas or add personas to the map.
2
Engine Gaps · What's Missing
Why the audit was needed in the first place · what has to be built or fixed
Gap 01 · Critical infrastructure
Ads aren't classified to personas in Motion
Today we have to heuristically map ad names to personas (creator keyword, format hint, hook content). The result is 19% of spend unclassifiable and 62% mapped to one micro by string match alone. The engine can't trust this.
Build: Persona tagging discipline at ship · auto-tagger that classifies new ads from copy + creator + format · Motion AI tagging trial (BODi/H2Tab/MyPrize) lands this.
Gap 02 · Strategy infrastructure
No documented basis for the persona set
9 personas exist with no record of what data or research produced them. We can't audit whether they're real, fresh, or out of date. The engine can't refresh what it can't trace.
Build: Strategist captures persona-build basis in Milanote (which data, which interviews, when refreshed) · quarterly refresh cadence on the calendar.
Gap 03 · Coverage
2 of 9 personas have never been in market
The Athlete and Lapsed Supplement Buyer have zero L90 spend. Real untested territory inside the existing Milanote. The engine should have surfaced and scheduled these long before.
Fix: Schedule both for net new persona tests in coming roadmaps. Athlete first (more obviously adjacent to the working M1 cluster).
Gap 04 · Structural
2 of 9 personas are broken as built
Aesthetic Optimizer burns $64K at 0.66 ROAS. Relief Seeker is built on "inflammation" language the brand has banned. Both either need rebuilds or removal from the map.
Fix: Strategist call on each. Aesthetic Optimizer: wrong persona vs wrong creative. Relief Seeker: rewrite around legal language (joint stiffness, fatigue, gut bloat) or remove.
Gap 05 · Concentration
62% of spend on a single micro
M1 Brecka Loyalist alone is 43 ads and $518K. Any fatigue on this micro takes the whole account with it. The engine should have flagged this risk months ago and grown other macros in parallel.
Fix: Reallocate. Reduce M1 to maintenance iterations. Invest in M2 (underspent winners). Open M3 (acquisition expansion).
Gap 06 · Spend efficiency
Underspent winners, overspent losers
Trend-Burned (2.3% spend, 1.74 ROAS) and Committed but Stuck (4.6%, 1.78) are starved. Aesthetic Optimizer (7.6%, 0.66) is overfed. Pure allocation problem; no creative change required.
Fix: Roadmap reallocates spend share starting next month. Tracked monthly.
3
Proposed Engine State · What We Want to Build
3 validated macros, 9 micros, infrastructure to keep it honest
Current vs Engine — At a Glance
Current
9 flat personas in Milanote, undocumented origin
Ads not tagged to personas in Motion
62% of spend on one micro · concentration risk
2 personas never produced
2 personas broken as built
19% of spend unclassifiable
No persona refresh cadence
→
Engine State
3 macros · 9 micros · 4 angles each · 4 vehicles each (Coverage Map)
Every ad tagged to a micro at ship (auto-tagger + discipline)
Hero macro at ~40-50% max · supporting macros grown deliberately
Net new persona test every month · forces untested micros into market
Broken personas removed or rebuilt with strategist sign-off
100% of spend mapped · "Unclassified" becomes a flag, not a category
Quarterly persona refresh on the calendar · documented basis in Milanote
The Proposed Coverage Map · 3 Macros from L90 Signal
M1 · The Trust-Driven Optimizer
Hero macro · 62% of L90 spend · 1.36 ROAS
Already-convinced biohackers. Trust transfer from podcast and credentialed voices. Proven engine, concentration risk. Engine state: reduce to ~40-50% of spend, hold as hero through iteration only, no new variants beyond top performers.
M1.1 Brecka Loyalist— working at scale; iterate top performers
M1.2 Active Recovery Seeker— 0.94 ROAS · re-brief or merge into M1.1 as sub-angle
M1.3 The Athlete— untested · net new test candidate (proven vehicle, new persona)
M2 · The Quiet Builder
Underspent · 7% of L90 spend · 1.76 ROAS blended · grow this
People doing the work quietly. Founder voice, testimonial proof, mechanism explainer. Less influencer authority, more real-world result. Currently the highest-efficiency cluster in the account but starved of spend. Engine state: grow toward 25-30% of spend.
M2.1 Committed but Stuck— Testimonial-Video04 + MOFUFounder · 1.78 ROAS · scale this
M2.2 Trend-Burned Re-Evaluator— Tweet statics · 1.97 ROAS · scale this
M2.3 Mechanism Believer— 3D animation · 2.02 ROAS · scale this
M3 · The Value Seeker
Opportunity macro · 4% of L90 spend · acquisition expansion
Cost-aware buyers and lapsed supplement buyers. Sale statics already proving the upside (3.25 ROAS). Engine state: grow toward 15-20% of spend with net new test on Lapsed Buyer.
M3.1 Cost-Conscious Hydrogen Curious— Sale static 3.25 ROAS · push the vehicle
M3.2 Lapsed Supplement Buyer— untested · net new test candidate
M3.3 Skeptic Convert— myth vs reality angle · new micro to develop
Removed from the map: Aesthetic Optimizer (failing creative, 0.66 ROAS, $64K spent — call: wrong persona vs wrong creative) and Relief Seeker (built on banned language). Both pending strategist sign-off. 19% unclassified ads need retagging to existing micros or surfacing missing personas.
Open Decisions · Strategist Calls Before We Lock
1 · Document the persona origin. Strategist captures what data / interviews / research produced the 9 personas. Without this, every future refresh starts blind.
2 · Aesthetic Optimizer · wrong persona or wrong creative? Decides whether to remove or re-brief.
3 · Relief Seeker · rewrite around allowed language, or remove? Words available: joint stiffness, brain fog, fatigue, gut bloat. Inflammation is off the table.
4 · Active Recovery Seeker · invest in a real test, or fold into M1.1? 3 ads is too thin to call.
5 · 19 unclassified ads · tag into existing micros, or add a 10th persona? "103-Copy", Hydrogen Gas, Instagram-organic format need a home.
6 · Macro names. "The Quiet Builder" and "The Value Seeker" are first drafts. Strategist owns final naming for client comms.
7 · Build the ad-to-persona auto-tagger. Engineering scope on Franco's Motion MCP work · should ship before next quarter's audit.
4
The May 2026 Roadmap · What Ships This Month
The audit drives this · pivot from M1 saturated to M2 underspent · first Lapsed Buyer test
Output of this audit
May 2026 H2Tab 4-Week Roadmap · 20 assets · Tier 1
The roadmap pivots from prior months: M1 Trust-Driven Optimizer holds at maintenance (5 ads, Brecka iteration only). M2 The Quiet Builder becomes the hero this month (12 ads across Testimonial, Tweet static, 3D animation — the three underspent winners). M3 The Value Seeker ships the first Lapsed Supplement Buyer net new test (3 ads on Sale static vehicle). No new production for Aesthetic Optimizer or Relief Seeker pending decisions above.