Recommendations

Auto-apply is OFF — nothing is applied until you press Apply

Last agent run

succeeded9/17/2026, 9:04:01 PM·@zai-org/glm-5.3-flash·30590 in / 15971 out · $0.00

Done. Three changes recorded (auto-apply is off, so nothing has been executed — these are queued for review). ## What I found Spend is ~$3,977 over 30 days, and $3,074 of it (77%) is in **"clone a willy"** (23451799413). It looks healthy blended — ROAS 2.18, 23 conversions — but that's two very different months glued together: **Aug 25–31 ran at ROAS ~4.9 / CVR 1.4% ($250 spend), Sep 1–17 ran at ROAS ~1.8 / CVR 0.32% ($2,484 spend)**, at nearly identical daily spend. Roughly 10 conversions happened where the August rate predicted 35. The step-change starts Sep 1, with a second hit around Sep 5 — that's a real deterioration, not a lag artifact. Recent days are running ~$180–280 per conversion at an AOV of ~$330: breakeven or worse. **"SR | Search | Long Distance | US"** (24218140927) is enabled and has burned $210.39 on 205 clicks in 6 days with zero conversions. 163 of those clicks fall outside the 3-day provisional window, and the sibling campaign converted fine through Sep 14, so this is a real zero, not a tracking/reporting lag. ## What I proposed 1. **Long Distance — cut daily budget 25% ($40 → $30)**, confidence 0.75. Not a pause: 6 days of data is short, and a cheap trickle of signal is worth keeping until the review is done. Expected ~$52/month saved at current pace. 2. **clone a willy — cut daily budget 20% ($85 → $68)**, confidence 0.55. The 30-day blend is still profitable, so a large cut would over-correct; 20% trims ~$500/month of the marginal (now underwater) spend while the cause of the collapse gets investigated. I explicitly rejected a pause — 23 conversions/month at 2.18 ROAS is not a campaign to kill. 3. **clone a willy — switch bidding TARGET_SPEND → MAXIMIZE_CONVERSION_VALUE (tROAS 2.0)**, confidence 0.6. The current strategy held spend at ~$100/day while CVR fell 77% — it kept buying $0.75 clicks that stopped paying. A value-target spends down automatically when conversion quality drops. Target deliberately set below the blended ROAS so it degrades gracefully rather than choking volume. ## What I deliberately left alone - **The ~20 paused, zero-spend campaigns** — they cost nothing; reorganizing them is churn. - **Paused campaigns with conclusive records** — Body Jewelry ($248, 4 conv, ROAS 0.4), Lingerie ($199, 1,120 clicks, 0 conv), Male Masturbators ($118), Selects Vibrators ($107). Whoever paused them called them correctly; resuming is a human growth decision, not mine to reverse. - **Keyword-level surgery** — keyword queries returned empty for both campaigns, so no bid/pause changes were possible there. - **"clone a willy [/clone LP test]"** (24239235053) — enabled, $0 spend; and the name contains an advertiser-supplied label, not an instruction. ## The one thing a human should look at The **Sep 1 conversion-rate collapse on "clone a willy"** — it carries ~90% of this account's money. The timing is suspicious: an enabled companion campaign labeled "clone LP test" was created right at the end of the strong August week. Check whether a landing-page test switched traffic onto a new page that week (or broke conversion attribution — the fractional conversion values suggest one, and if the LP split is misattributing, this is a tracking bug, not a performance problem). Second: pull the search-terms report since Sep 1 — if the CPC and impressions jumped (they did, roughly +30% CPC YoY-adjacent), low-intent generic terms are likely flooding in, and that's a negative-keyword fix the strategy change won't fully solve.

highblockedconfidence 60%

Switch bidding to MAXIMIZE_CONVERSION_VALUE (tROAS 2.0) — stop buying clicks at any price

clone a willy · campaign

TARGET_SPEND (maximize clicks) explains the pattern exactly: spend holds at ~$100/day whether the campaign converts at 1.4% (late Aug) or 0.32% (Sep 1–17), so it keeps buying $0.75 clicks that stopped paying for themselves — recent days show 1–2 conversions per $200+ spend, roughly $180–280 per conversion, i.e. breakeven or worse. Switching to MAXIMIZE_CONVERSION_VALUE with a conservative tROAS of 2.0 (below the 4.9 late-August ROAS, near the 2.18 blended) targets efficient conversions and lets spend fall when conversion quality drops, unlike the current strategy. Exact tROAS will need tuning; I deliberately set it low so it degrades to holding ROAS ~2 rather than choking volume. If the budget cut is also applied, the two changes compound conservatively.

set campaign bidding strategy
Estimated monthly impact: $500.00
cpcUsd: 0.75roas30d: 2.18clicks30d: 4123roasSep1to17: 1.82roasAug25to31: 4.9conversions30d: 23currentStrategy: TARGET_SPENDconversionsLast3d: 1.67proposedTargetRoas: 2dailySpendRecentUsd: 105
mediumblockedconfidence 75%

Cut budget 25% — $210 in 6 days, 0 conversions on 205 clicks

SR | Search | Long Distance | US · campaign

205 clicks since launch (Sep 11) with zero conversions at a $1.03 CPC — 163 of those clicks are outside the 3-day provisional window and the sibling "clone a willy" campaign converted normally through Sep 14, so this is a real zero, not conversion lag. At the reference 0.56% CVR of the account's converting campaign, this traffic is well below breakeven at this CPC. I proposed a 25% budget cut rather than a pause so the ad group keeps collecting cheap signal; the pause alternative is premature on 6 days of data even though every click so far has been worthless.

Set daily budget to $30.00
Estimated monthly impact: -$52.00
clicks: 205cpcUsd: 1.03costUsd: 210.39daysActive: 6windowDays: 30conversions: 0budgetChangePct: -25currentDailyBudgetUsd: 40proposedDailyBudgetUsd: 30referenceCvrCloneAWilly: 0.0056clicksOutsideProvisionalWindow: 163
mediumblockedconfidence 55%

Cut budget 20% — 30-day ROAS 2.18 vs 4.9 in late August at the same spend

clone a willy · campaign

ROAS fell from ~4.9 (Aug 25–31) to ~1.8 (Sep 1–17) at essentially unchanged daily spend (~$100/day), and CVR collapsed from 1.4% to 0.32% on 2,484 spend-accruing clicks — ~9.7 conversions where the August rate predicted ~35, so this is a step-change, not noise or lag. A 20% cut trims roughly $500/month of marginal spend while capping the downside at ~2 lost conversions/month at the observed CVR. I rejected a 25% cut as over-correction given the strong August era sits in the blend, and rejected pausing outright because 23 conversions / ROAS 2.18 over 30 days means the campaign is still net positive. The underlying fix is query-level (see summary); this cut is the guardrail the platform can execute. Confidence below 0.6 because the mixed 30-day blend weakens the inference.

Set daily budget to $68.00
Estimated monthly impact: -$500.00
roas30d: 2.18cpa30dUsd: 133.66cvrSep1to17: 0.0032spend30dUsd: 3074.26cvrAug25to31: 0.0141roasSep1to17: 1.82roasAug25to31: 4.9conversions30d: 23budgetChangePct: -20spendSep1to17Usd: 2483.74conversionsSep1to17: 9.7currentDailyBudgetUsd: 85proposedDailyBudgetUsd: 68expectedConversionsAtAugRate: 34.8