Find scaling Facebook ad creatives.

Use the connected Spytrend MCP to define this scaling-creative research scope.

Niche (category slug): ecommerce_and_retail
Country (ISO-2): US
Media: All media

Keep the selected category, country and media as hard constraints. Use Meta only. Run heavy discovery calls sequentially. Do not broaden the scope to fill the result. Do not run discovery in this step.

Define scaling only as the returned `exploding` and `early_signal` launch-momentum/distribution signals used by this workflow. Creative reuse is not the scaling definition, ranking criterion or verification gate.

Return no more than three concise bullets: confirmed scope, verification gates and evidence limits. Do not show plan, quota, token balance, raw parameters or internal identifiers.

Stop after Step 1.

Cost guard: before each paid Spytrend call, run the free `estimate_cost` with the same tool and arguments, estimate the cost and stay within the user's stated budget; if the user has not stated a budget, tell them the estimate before the first paid call. If pricing consent is required, show the returned terms and stop until the customer explicitly agrees; never accept on the customer's behalf. If a required call does not fit the stated budget, make no paid call and return `Budget limit prevents required research`; omit an optional call that does not fit and disclose reduced coverage. When opening ads with `get_ad`, skip multilang ads (`is_multilang`) and never call `get_media` after `get_ad`. Retry once only when `retryable=true`, after `retry_after_seconds`. Read `params.ignored` after every search; an ignored required filter never counts as applied. If a search reports `pagination.total_status=restricted` (a free-plan archive window; `meta.plan_window_cutoff` names the boundary), tell the user the result covers only ads that started on or before that date, and never report a zero from it as "no such ads exist". Cost notices are not part of the final customer report.
Use the connected Spytrend MCP to find active Meta creative cohorts with current scaling signals for ecommerce_and_retail, US and All media.

Run the exploding and early-signal routes sequentially with limits 6 then 4 and the full response payload; do not request compact output or pass an unsupported verbosity parameter. Inspect the complete geo_distribution for every row. Retain a cohort only when US represents at least 40% of the returned distribution. Preserve internally the returned signal order and absolute launches, active ads and active advertiser-page counts.

If either discovery call returns retryable=true with retry_after_seconds, wait that delay and retry once identically. A final failure makes that route unavailable and the run Partial while the other route continues. A completed zero is valid zero evidence; unavailable is never zero.

Label a row Low-base acceleration when the exact returned comparable baseline field (`previous_ads` or `baseline_ads`) is 0 or 1, or `growth_basis` is `insufficient_base`. Do not invent or translate a missing baseline field. Never rank a low-base row by multiplier alone. Keep raw cohort identifiers and complete distributions in retained evidence only.

Return a concise progress summary of no more than eight bullets: route outcomes, returned count, country-qualified count, low-base count and at most four leading qualified cohort metric summaries using plain product labels when returned. Do not show raw IDs, tables, quotas, token usage or full rejected-country distributions.

Stop after Step 2.

Cost guard: before each paid Spytrend call, run the free `estimate_cost` with the same tool and arguments, estimate the cost and stay within the user's stated budget; if the user has not stated a budget, tell them the estimate before the first paid call. If pricing consent is required, show the returned terms and stop until the customer explicitly agrees; never accept on the customer's behalf. If a required call does not fit the stated budget, make no paid call and return `Budget limit prevents required research`; omit an optional call that does not fit and disclose reduced coverage. When opening ads with `get_ad`, skip multilang ads (`is_multilang`) and never call `get_media` after `get_ad`. Retry once only when `retryable=true`, after `retry_after_seconds`. Read `params.ignored` after every search; an ignored required filter never counts as applied. If a search reports `pagination.total_status=restricted` (a free-plan archive window; `meta.plan_window_cutoff` names the boundary), tell the user the result covers only ads that started on or before that date, and never report a zero from it as "no such ads exist". Cost notices are not part of the final customer report.
Use the connected Spytrend MCP to verify representative ads for the country-qualified scaling cohorts already retained.

Open up to five representatives sequentially, one per cohort, so that 3–5 results can survive verification. Prioritize qualified cohorts with surviving active ads, meaningful absolute activity and a non-low baseline before low-base multipliers. Skip multilang ads (`is_multilang`). A sample qualifies only when it is active, uses All media, includes US, matches ecommerce_and_retail and clearly represents the same creative or the same product or offer as its cohort. Same advertiser, page or landing domain alone is insufficient.

If a representative fails only on country, status or format, run one bounded replacement `search_ads` for that cohort (source=meta, countries=[US], status_today=active, the representative's landing domain, limit=5) and open at most one returned ad; accept it only when it shows the same exact creative or the same product or offer as the cohort. A same-domain ad with another product never counts. If no replacement qualifies, keep a useful unresolved opened cohort under Signals needing verification with the exact reason; unopened cohorts are coverage limits, not verification signals.

If a representative-detail or replacement call returns retryable=true with retry_after_seconds, wait that delay and retry once identically. A final candidate-level failure excludes that candidate and is disclosed; it does not turn two completed discovery routes Partial. A failed required replacement search is unavailable replacement evidence, not zero.

Retain full evidence internally, including each verified and replacement ad ID for the supporting record. Return no more than eight concise bullets with verified count, useful unresolved count, rejected count, replacement outcome and short product/reason summaries. Do not show raw IDs, tables, quotas, token usage, raw calls or unrelated product details.

Stop after Step 3.

Using retained evidence only, write the final customer report without rerunning Spytrend.

Keep the complete answer under 380 words and table-free. Do not show raw IDs, category slugs, tracking domains, calls, route names, quotas, token usage or detailed rejected rows. Start with `# Meta Creative Scaling Watchlist: ` followed by the human-readable niche name and the full country name (for example `Ecommerce and Retail in United States`; never the slug ecommerce_and_retail or the code US), then one summary line with verified count, All media and Complete/Partial.

Complete means every configured discovery and selected verification call completed successfully. A completed replacement lookup with zero matching ads is valid zero replacement evidence and does not make the run Partial. An unresolved cohort affects that cohort only. A bounded page, additional available rows and the representative cap are coverage limitations, not failures. Use Partial only for a failed required call or unresolved required filter.

Show 3–5 verified results when that many qualify (never pad), each as one compact numbered block containing: product or offer; scaling label; Why now with absolute launches, active ads and active advertiser pages; representative status/country/format; advertiser; landing domain as plain text; exactly one Why inspect it. Mark Low-base acceleration visibly when applicable and do not foreground multipliers.

When zero cohorts qualify, omit an empty watchlist and use `## Result`. Add `## Signals needing verification` only for opened representatives with useful unresolved evidence, maximum two one-sentence items; never list unopened cohorts. Add `## Research coverage` with at most two bullets covering returned, country-qualified, opened, verified/unresolved counts and bounded-sample limits. State briefly that off-category representatives were excluded and no arbitrary same-domain replacement was accepted.

If the user later asks `Show supporting record`, list each row's Spytrend ad ID (verified and replacement), the calls and the rejected candidates from retained evidence without rerunning anything. Never put ad IDs in the main answer.

End exactly: `Scaling signals show observed launch momentum and distribution spread, not conversions, revenue, ROAS or profitability.`

Stop after Step 4.