# Adform Stats Performance > >- - **Type:** Skill - **Install:** `agentstack add skill-adform-agentic-skills-adform-stats-performance` - **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior - **Seller:** [adform](https://agentstack.voostack.com/s/adform) - **Installs:** 0 - **Category:** [Agent Skills](https://agentstack.voostack.com/c/agent-skills) - **Latest version:** 0.1.0 - **License:** MIT - **Upstream author:** [adform](https://github.com/adform) - **Source:** https://github.com/adform/agentic-skills/tree/main/dist/claude/adform-agentic-skills/skills/adform-stats-performance - **Website:** https://site.adform.com ## Install ```sh agentstack add skill-adform-agentic-skills-adform-stats-performance ``` Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent. ## About # Adform stats performance reporting (mcpStats) Dimensional ad-serving metrics via the `mcpStats` GraphQL query. Returns a paginated rows result with any combination of dimensions and metrics. Read-only. ## Connection & tooling Runs on the Adform GraphQL MCP. Use `graphql_validate` to check every query before `graphql_execute`. Use `graphql_introspect` on `McpStatsFilterInput`, `McpStatsDimensions`, or `McpStatsMetrics` to discover fields. Keep calls sequential (~1–2s apart). --- ## Query structure ```graphql { mcpStats { totalRowCount # total rows matching the filter (before paging) totals # aggregate totals across all rows (JSONType) columns { dimensions { ... } # declare which dimension columns to include metrics { ... } # declare which metric columns to include } rows( filter: McpStatsFilterInput! # date is required paging: { offset: Int!, limit: Int! } sort: [{ column: Int!, direction: asc|desc }] timeZoneOffset: Int # optional; minutes offset from UTC ) } } ``` **Key rules:** - `date` inside `filter` is always required. Use `from`/`to` (ISO date strings). `DatePreset` enum values exist but are deprecated — always use `from`/`to`. - `DateFilterKind` can be `utc` or `campaign` (default is campaign time). - `sort.column` is a zero-based index into the columns declared in `columns {}`. - `rows` returns `[[JSONType]]` — a 2D array; each inner array is one row, values align positionally with the declared columns (dimensions first, then metrics). - `totals` returns `JSONType` — a flat object with the same column keys summed. - `paging.limit` is typed as `Limit` scalar — use integers up to a reasonable page size (50–200); paginate with `offset` for large result sets. --- ## Dimension reference Declare only the sub-fields you need. Each sub-field corresponds to one column in the result rows. ```graphql columns { dimensions { date { date } # transaction date (YYYY-MM-DD) date { utc } # date in UTC timezone date { hour } # hour of day (0–23) date { weekday } # weekday name advertiser { id } campaign { id } campaign { currencyName } # campaign currency order { id } lineItem { id } lineItem { name } banner { id } banner { name } tag { id } tag { name } page { id } page { name } rtbDomain { name } # 2nd-level domain (e.g. cnn.com) rtbDeal { id } bidReason { name } # DSP no-bid reason or successful bid reason mobileApp { id } mobileApp { name } mobileAppStore { name } } } ``` --- ## Commonly used metric combinations ### Core delivery + cost ```graphql metrics { impressions clicks ctr cost ecpm ecpc rtbMediaCost rtbCostInAgencyCurrency } ``` ### Viewability ```graphql metrics { impressions viewImpressionsIab # viewable impressions (IAB standard) viewImpressionsPercentIAB # viewability rate % measurableImpressions measurableImpressionsPercent undeterminedImpressions avgViewabilityTime } ``` ### Video ```graphql metrics { impressions videoPlayStartCount videoCompleteCount videoCompletionRate # videoCompleted / videoPlayStarted * 100% videoStartRate # videoPlayStarted / impressions * 100% avgVideoPlayTime } ``` ### RTB bidding ```graphql metrics { rtbBids lostBids bidReasonCount impressions rtbWinRate # impressions / bids % rtbMediaCost } ``` ### Conversions & sales ```graphql metrics { impressions clicks conversions cov # conversion rate (conversions / clicks %) ecpa # cost per conversion sales roi } ``` --- ## Example queries (all validated) ### 1. Campaign daily trend — impressions, CTR, eCPM, viewability ```graphql { mcpStats { totalRowCount totals columns { dimensions { date { date } campaign { id } } metrics { impressions clicks ctr cost ecpm viewImpressionsIab viewImpressionsPercentIAB videoCompleteCount videoCompletionRate conversions } } rows( filter: { date: { from: "2026-06-01", to: "2026-06-30" } campaign: { ids: ["4221341"] } } paging: { offset: 0, limit: 100 } sort: [{ column: 0, direction: asc }] ) } } ``` ### 2. Domain breakdown — where are impressions and cost going? ```graphql { mcpStats { totalRowCount totals columns { dimensions { rtbDomain { name } } metrics { impressions clicks cost ecpm rtbBids rtbWinRate } } rows( filter: { date: { from: "2026-06-01", to: "2026-06-30" } advertiser: { ids: ["2133936"] } } paging: { offset: 0, limit: 100 } sort: [{ column: 1, direction: desc }] ) } } ``` ### 3. Bid reason analysis — why are we losing auctions? ```graphql { mcpStats { totalRowCount totals columns { dimensions { bidReason { name } } metrics { rtbBids lostBids bidReasonCount impressions rtbWinRate } } rows( filter: { date: { from: "2026-06-01", to: "2026-06-30" } advertiser: { ids: ["2133936"] } } paging: { offset: 0, limit: 50 } sort: [{ column: 0, direction: desc }] ) } } ``` ### 4. Deal-level performance stats ```graphql { mcpStats { totalRowCount totals columns { dimensions { rtbDeal { id } } metrics { impressions clicks cost ecpm rtbBids rtbWinRate rtbMediaCost } } rows( filter: { date: { from: "2026-06-01", to: "2026-06-30" } advertiser: { ids: ["2133936"] } } paging: { offset: 0, limit: 50 } sort: [{ column: 0, direction: desc }] ) } } ``` ### 5. Advertiser-level summary — all metrics, no dimension breakdown ```graphql { mcpStats { totalRowCount totals columns { dimensions { advertiser { id } } metrics { impressions clicks ctr cost ecpm viewImpressionsIab viewImpressionsPercentIAB rtbBids rtbWinRate conversions } } rows( filter: { date: { from: "2026-06-01", to: "2026-06-30" } advertiser: { ids: ["2133936"] } } paging: { offset: 0, limit: 10 } sort: [{ column: 1, direction: desc }] ) } } ``` --- ## Filter reference | Filter field | Type | Notes | |---|---|---| | `date` | `DateFilterInput` | **Required.** Use `from`/`to` ISO date strings. `kind` is `utc` or `campaign`. | | `advertiser` | `AdvertiserFilterInput` | Filter by `ids: [ID!]` or `names: [String!]` | | `campaign` | `CampaignFilterInput` | Filter by `ids`, `names`, `types`, `subtypes`, `active` | | `order` | `OrderFilterInput` | Filter by `ids`, `names`, `status` | | `lineItem` | `LineItemFilterInput` | Filter by `ids`, `names`, `buyTypes`, `status` | | `banner` | `BannerFilterInput` | Filter by `ids`, `names`, `types`, `sizes` | | `tag` | `TagFilterInput` | Filter by `ids` | | `rtbDomain` | `RtbDomainFilterInput` | Filter by `names` | | `rtbDeal` | `RtbDealFilterInput` | Filter by `ids` | | `country` | `CountryFilterInput` | Filter by `ids` or `names` | | `continent` | `ContinentFilterInput` | Filter by `ids` or `names` | | `region` | `RegionFilterInput` | Filter by `ids` or `names` | | `mobileApp` | `MobileAppFilterInput` | Filter by `names` | | `bidReason` | `BidReasonFilterInput` | Filter by `names` | | `trackingPoint` | `TrackingPointFilterInput` | Filter by `ids`, `names`, `preset` | | `referrerTypes` | `[ReferrerType!]` | `directTraffic`, `referringSite`, `naturalSearch`, `campaign`, `socialMedia` | | `metrics` | `[FilterInput!]` | Post-aggregate row filter. e.g. `{ fieldName: "impressions", operation: gt, values: [0] }` | --- ## Reading rows results `rows` is a 2D array `[[JSONType]]`. The column order matches exactly the order in which dimensions and metrics were declared inside `columns {}` — dimensions come first in declaration order, then metrics in declaration order. ``` Example columns declared: dimensions: { date { date }, campaign { id } } metrics: { impressions, ctr, ecpm } Row layout: [date_value, campaign_id, impressions_value, ctr_value, ecpm_value] ``` `totals` is a flat JSONType object with the same column keys summed across all rows (not just the current page). Use it for account-level totals without iterating all pages. --- ## Presenting - **Date trends**: time-series table sorted by date ascending; flag days with zero impressions as delivery gaps. - **Domain breakdown**: ranked table by impressions or cost descending; flag domains with anomalous eCPM (very high or very low vs account average). - **Bid reason analysis**: table of reason names, bid count, lost bid count, win rate; frame each reason as an actionable problem (pricing, targeting, creative audit, budget). - **Deal breakdown**: table of deal IDs with impressions, win rate, eCPM; cross-reference with adform-deal-health-check for deals with zero impressions. - For large result sets, paginate using `offset` and use `totalRowCount` to determine how many pages are needed. - Always show `totals` as a summary row at the top or bottom of the table. ## Source & license This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code. - **Author:** [adform](https://github.com/adform) - **Source:** [adform/agentic-skills](https://github.com/adform/agentic-skills) - **License:** MIT - **Homepage:** https://site.adform.com Install and usage instructions live in the source repository linked above. ## Pricing - **Free** — Free ## Security capabilities Automated source analysis of v0.1.0 — what this tool can access: - **Network access:** no - **Filesystem access:** no - **Shell / process execution:** no - **Environment & secrets:** no - **Dynamic code execution:** no *"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.* ## Versions - **0.1.0** — security scan: passed — Imported from the upstream source. ## Links - Listing page: https://agentstack.voostack.com/l/skill-adform-agentic-skills-adform-stats-performance - Seller: https://agentstack.voostack.com/s/adform - Browse the marketplace: https://agentstack.voostack.com/browse --- Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.