best_practices.md was empty and several core beliefs were operating on stale or unverified assumptions. This report surfaces every gap, updates what can be updated from first principles, and clearly flags what requires input from Mink or external source material to resolve.
| File | Status | Notes |
|---|---|---|
sources.md |
PARTIAL | Good practitioner list. Missing Taylor Holiday (Common Thread Collective) and Andrew Faris (AJF Growth) — both highly relevant to LB's MER-first situation in 2026. |
best_practices.md |
EMPTY | Critical gap. This is the file that governs daily decision-making. All sections blank. Must be populated immediately. |
queue.md |
STALE | Topics listed but none actioned. Several items are actively relevant to LB right now (ASC, iOS attribution, competitive creative). |
semantic.md |
PARTIAL | Account facts loaded. Cross-channel patterns have one entry (email ROAS inflation). Audience, creative, seasonal, and bidding pattern sections are blank. |
procedural.md |
PARTIAL | Pre-loaded decision rules exist. No learned rules from experience yet — expected at this stage, but should be populated rapidly. |
study_log.md |
MISSING | File doesn't exist. No audit trail of what was studied, when, or what changed. Needs to be created. |
These are beliefs I was operating on that may be wrong or outdated in March 2026. Each one is evaluated for what I currently believe, why it might be wrong, the updated belief, and confidence level.
What I currently believe: The primary cause of Meta ROAS degradation is iOS 14.5 signal loss (2021). Citing CAPI as a fix for an active, ongoing problem.
Why it might be wrong: iOS 14.5 was 5 years ago. The platform has rebuilt significantly. Meta's Conversions API has matured, EMQ scoring exists, and brands running CAPI properly have largely adapted. Citing "iOS 14.5" as a current diagnosis signals outdated thinking. The real 2026 tracking challenges are different: browser-level privacy changes, first-party data quality, EMQ scores, and the gap between Meta-reported and Shopify-confirmed revenue.
Updated belief: Stop citing iOS 14.5 as a current diagnosis. The modern tracking problem is EMQ quality, server-side CAPI implementation maturity, and attribution window calibration. Always cross-reference Meta ROAS vs. Shopify revenue (MER) — that's the real signal. Meta's platform reporting can still undercount or overcount depending on setup quality, but the framing needs to be current.
Confidence in update: HIGH — this is directionally correct regardless of what specific Meta changes have occurred since 2021.
What I currently believe: Interest stacking and manual audience targeting are standard practice. Broad targeting is an advanced or risky move.
Why it might be wrong: Meta's AI-driven targeting has matured significantly. Advantage+ audience tools, the expansion of broad matching, and Meta's own guidance all suggest that in 2025-2026, broad targeting (with strong creative) frequently outperforms manually-built interest audiences. The creative IS the targeting signal now. Practitioners like Taylor Holiday and Andrew Faris have been vocal about this shift.
Updated belief: Broad targeting with strong creative is likely the primary prospecting approach that outperforms in 2026 for established brands with purchase data (like LB with 320K+ customers). Interest targeting still has a role in creative testing and initial audience seeding, but it is not the default. Meta's algorithm needs conversion signal to optimize — give it that via CAPI + pixel and let it find the audience.
Confidence in update: MEDIUM — directionally confident, but need Mink's real-world observation and practitioner content to confirm for LB specifically.
What I currently believe: CBO (Campaign Budget Optimization) vs. ABO (Ad Set Budget Optimization) debate. I have no current, informed position.
Why it might be wrong: Meta's own guidance has evolved here. In 2026, Advantage+ Shopping Campaigns (ASC) may have made the CBO/ABO debate partially irrelevant for DTC brands — ASC consolidates everything at the campaign level and lets Meta optimize fully. The "right" structure depends heavily on account maturity, budget level, and whether ASC is being used.
Updated belief: Cannot be updated with confidence without current practitioner input. This is a research queue item. I should not be making structural recommendations until I understand whether ASC is the right move for LB and how the current LB campaigns are structured.
Confidence in update: LOW — flagged for research. Need input from Mink on current campaign structure and practitioner content on 2026 structure best practices.
What I currently believe: Defaulting to "7-day click, 1-day view" as the standard attribution window.
Why it might be wrong: Meta has changed attribution defaults and reporting multiple times since 2021. The 7-day click window may still be standard, but I don't know if 1-day view is still recommended or if there have been changes to how Meta handles view-through attribution in 2026. This affects how I evaluate ROAS numbers from the daily data pull.
Updated belief: Cannot confirm without checking current Meta documentation or Mink's account settings. What I do know: whatever window is set in the account is what the data reflects. I need to know LB's current attribution window setting to interpret ROAS numbers accurately. This is a direct question for Mink.
Confidence in update: MEDIUM — process question that can be resolved quickly with account access or Mink input.
What I currently believe: I know ASC exists. I have zero operational knowledge of how it works, when to use it, or whether it's right for LB.
Why it matters right now: LB is running below breakeven (0.82x ROAS yesterday). ASC could be a structural answer, or it could be irrelevant — I genuinely don't know. This is a gap that could be materially affecting recommendations I make.
Updated belief: Cannot update without research. Priority item in the queue. Before recommending any structural changes to LB's campaign architecture, I need to understand what ASC is, whether LB should be running it, and how it differs from current structure.
Confidence in update: LOW — unknown territory. Research required before I can form a belief.
| Action | Who / What | Why | Priority |
|---|---|---|---|
| ADD | Taylor Holiday — Common Thread Collective | Publishes MER-first frameworks and Meta strategy for DTC brands. Directly relevant to LB's situation. One of the most data-forward voices in the space. | HIGH |
| ADD | Andrew Faris — AJF Growth | Former CEO of 4×400 (CTC portfolio brand). Very current on Meta mechanics, MER, and what's actually working in 2026 for DTC brands at LB's scale. | HIGH |
| VERIFY | All existing practitioners | Confirm they're still active and publishing in 2026. Some practitioners go quiet or shift focus. Need to verify the list is still live and relevant. | MEDIUM |
| ADD | Meta Performance Marketing newsletter | Direct from Meta — algorithm updates, new features, policy changes. Should be required reading monthly. | MEDIUM |
| Item | Status | Priority | Why it matters for LB right now |
|---|---|---|---|
| Advantage+ Shopping Campaigns (ASC) — full understanding | NOT STARTED | CRITICAL | LB is below breakeven. If ASC is the right structural move, I need to know. If it's not, I need to know why. I can't make structural recommendations without this. |
| Current Meta attribution windows + settings best practice (2026) | NOT STARTED | CRITICAL | Every ROAS number I analyze depends on attribution window. If LB's window is misconfigured, the numbers are wrong. |
| Broad targeting vs. interest targeting — current 2026 best practice | NOT STARTED | HIGH | LB's prospecting strategy likely needs to be evaluated. Can't make a recommendation without knowing which approach Meta's algorithm favors in 2026. |
| Full audit of LB's current Meta campaign structure | NOT STARTED | HIGH | I can see 3 active campaigns and 50 total in the account. I don't have ad set-level data yet. Can't diagnose root cause of 0.82x ROAS without it. |
| Creative fatigue benchmarks for DTC grooming brands | NOT STARTED | MEDIUM | LB's creative refresh rate is unknown. Need benchmarks specific to the category. |
| Video vs. static performance benchmarks for LB's audience | NOT STARTED | MEDIUM | Informs creative direction recommendations to Iris and Ares. |
| Competitor analysis — beard/grooming brands on Meta | NOT STARTED | MEDIUM | What creative angles are competitors running? Are they running heavy promos? Relevant context for CPM spikes. |
| Retargeting window optimization — 7 / 30 / 90 day | NOT STARTED | LOW | Retargeting is a lever, not the priority right now. Address after prospecting is stabilized. |
These are rules and frameworks I'm adding to best_practices.md based on this self-audit and first-principles reasoning. Items marked UNVERIFIED need Mink's confirmation or external source validation.
Meta-reported ROAS is a platform metric. MER (Total Shopify Revenue ÷ Total Meta Spend) is the truth. When Meta ROAS and MER diverge, trust MER. Platform attribution can over or undercount. The real question is always: is total business revenue growing relative to total ad spend?
Source: Taylor Holiday / Common Thread Collective framework. Consistent with internal guidance in optimization-framework.md.
Klaviyo email sends to LB's 40K+ subscriber list inflate Meta-reported ROAS by 20-40% on send days. This is an established pattern. Always check Apollo's email calendar before interpreting any single-day ROAS spike. Do not scale based on email-day numbers.
Source: Pre-loaded into semantic.md. Confirmed as operational rule.
In 2026, the working hypothesis is that strong creative acts as the primary audience signal for Meta's algorithm. Broad targeting + excellent creative likely outperforms interest stacking + mediocre creative. The implication: when ROAS is bad, ask about creative health before asking about audience settings.
Needs confirmation from: Mink's real-world observation + Taylor Holiday / Andrew Faris content. HIGH CONFIDENCE in direction, MEDIUM CONFIDENCE in specifics.
Ad sets need approximately 50 purchase events per week to exit the learning phase. At LB's current CPP, that implies a minimum ad set daily budget that needs to be calculated once CPP is confirmed from the deep dive. Under-budgeted ad sets stay in learning forever and perform erratically.
Needs confirmation: what is LB's current CPP? What ad set budgets are currently set? Requires ad set-level data pull.
These are the specific inputs that would most accelerate my calibration. Ranked by impact.
| # | What I Need | Why | Impact |
|---|---|---|---|
| 1 | Attribution window setting in Meta account (check Events Manager or campaign settings) | Every ROAS number I analyze is only valid in the context of the attribution window. If LB is on 7-day click + 1-day view, that's different from 1-day click only. | CRITICAL |
| 2 | Is CAPI (Conversions API) running? When was it set up? What's the EMQ score? | Tracking quality directly affects ROAS accuracy. If EMQ is below 8.0, Meta is making optimization decisions on bad signal. | CRITICAL |
| 3 | MER for the last 30 days (Total Shopify revenue ÷ Total Meta spend) | LB's Meta-reported ROAS is 1.47x (30-day) and 0.82x (yesterday). I need to know if Shopify revenue tells the same story or a different one. | CRITICAL |
| 4 | What changed around the time ROAS started declining? Was there a specific date? | If ROAS fell off a cliff on a specific date, there's a specific cause. If it's been a slow bleed, that's a different problem. The diagnosis depends on the timeline. | HIGH |
| 5 | Screenshot of ad set-level breakdown for the 3 active campaigns | I can see campaign-level data but not ad set level. The bleed is almost certainly concentrated in specific ad sets. I need to see inside the campaigns. | HIGH |
| 6 | What practitioners or newsletters is your team currently following for Meta? | If your team has already been following Taylor Holiday, Andrew Faris, or others — I want to know what they've been saying. Real-world signal from practitioners you trust. | MEDIUM |
| 7 | How old is the current creative running in the active campaigns? When was it last refreshed? | Creative fatigue is the #1 silent killer. If the active ads are 30+ days old, that alone could explain the ROAS collapse. | MEDIUM |
| File | Action | What Goes In |
|---|---|---|
cani/best_practices.md |
WRITE — EMPTY | MER-first framework, email attribution inflation rule, creative-as-targeting hypothesis, learning phase budget rule. Start here and build from Mink's inputs. |
cani/sources.md |
UPDATE | Add Taylor Holiday (Common Thread Collective) and Andrew Faris (AJF Growth) to the must-follow list. Add Meta Performance Marketing newsletter. |
cani/queue.md |
UPDATE | Reprioritize: ASC and attribution window move to critical priority. Add MER calibration as new item. |
cani/study_log.md |
CREATE | Log this session: date, stale beliefs identified, beliefs updated, source used, confidence level. This is the audit trail of how my knowledge base evolves. |
memory/semantic.md |
UPDATE | Add MER as the primary performance metric. Add attribution window caveat. Populate cross-channel patterns with email inflation rule. |
What I know well: LB's business fundamentals, the breakeven math, email-ROAS inflation, and the basic optimization decision tree. These are solid.
What I don't know: Current Meta platform mechanics in 2026, whether ASC is the right structure for LB, what LB's tracking quality actually looks like, and what's actually causing the 0.82x ROAS crash right now.
The fastest path to getting properly calibrated is a 15-minute conversation with Mink — answering the 7 questions above. That conversation would do more for my knowledge base than any amount of documentation reading. The data is in his head and in the account. I need access to both.
While that conversation is happening, the deep dive protocol is the right next move. That's the structured way to get everything I need in one session.