--- name: analyze-auction-insights description: >- Turns the Auction Insights report into a competitive read you can act on: pulls impression share, overlap, outranking and position metrics per competitor over a 3-month window, classifies competitors, flags significant movements against explicit thresholds, and, critically, correlates competitor moves with the account's own KPIs before recommending any response. Almost fully automatable through VigilDog reads; the only human parts are recognizing who the domains are and deciding whether to respond strategically. Run it monthly inside run-a-monthly-performance-review, or ad hoc when impression share drops or CPCs climb without any internal change. It is read-only analysis: it does not change bids or budgets (route follow-ups to handle-budget-limited-campaigns, improve-quality-score, or allocate-budget-across-campaigns), and weekly KPI triage belongs in run-a-weekly-performance-review. --- # Auction Insights Competitive Analysis ## Purpose Assess how the competitive landscape is shifting, who is bidding into your auctions, who is gaining ground, who left, and decide whether it actually matters. The discipline of this procedure is the second half: most competitor movements deserve no response as long as your own CPA/ROAS stays on target, and the analysis exists to prevent reactive bid wars as much as to spot real threats. ## When to run - Monthly, as the competitive section of run-a-monthly-performance-review. - Impression share fell noticeably with no internal budget/bid/structure change. - CPCs rose account-wide without internal cause. - A stakeholder asks "what are competitors doing?" and deserves a data answer. ## When NOT to run - The KPI problem is clearly internal (budget cap, disapproval, tracking break), diagnose via run-a-weekly-performance-review first; competition is the hypothesis of last resort. - You already know it is a rank problem and want to fix quality, improve-quality-score. - Budget-driven impression loss is confirmed, handle-budget-limited-campaigns or allocate-budget-across-campaigns. - Less than ~90 days of account history: trends will be noise. ## Prerequisites - Around 90 days of account data. - The user's knowledge of who their real top 3-5 market competitors are (domains in the report need business interpretation). - Last cycle's analysis output, if one exists, for trend continuity. ## Procedure 1. Set scope. Default: campaign level for the monthly read. Drill to ad-group or keyword level only when diagnosing a specific theme's CPC spike or IS drop. 2. Pull the data with `gads_get_auction_insights` for the chosen scope: your own impression share, top-of-page and absolute-top rates, and per-competitor IS, overlap rate, outranking share, and position-above rate. Repeat via `gads_run_gaql_query` with monthly segmentation across the last 3 months so every metric has a 3-point trend, not a snapshot. For Shopping campaigns pull the Shopping-side insights too, Search-only analysis misses half the picture on retail accounts. 3. Classify each competitor mechanically: - Core: IS above 30% AND overlap rate above 50%, they contest most of your auctions. - Peripheral: below either bar, context, rarely actionable. - New: absent last cycle, present now. - Exited: present last cycle, gone now. 4. Flag significant movements against thresholds (see decision rules): month-over-month IS swings, outranking-share swings, and strong new entrants. Everything unflagged is recorded and ignored. 5. Correlate with your own KPIs, the step that decides whether anything matters. Pull the same 3 months of own-account data via `gads_get_account_performance` and `gads_get_campaign`, plus `gads_run_gaql_query` for impression share lost to rank vs lost to budget. Walk the chain: did IS actually fall → did impressions/clicks fall → did conversions/revenue fall → did CPC/CPA rise? A competitor gaining IS while your KPIs hold is a non-event. 6. HUMAN STEP (outside VigilDog): identify the flagged domains (who they are, whether a new entrant is a real rival or an aggregator) and make the strategic call on responses that involve spend. 7. Produce the written assessment: own position and trend, per-core-competitor trend table, entrants/exits, KPI correlation verdict, and either concrete follow-up actions or an explicit "no action needed". Persist it so next month's run has a baseline. ## Decision rules - Flag thresholds: competitor IS change beyond ±10 percentage points month-over-month; outranking-share change beyond ±15 points; any new entrant debuting above 20% IS. - Two-cycle rule: no bid/budget response on a single month's movement; require the trend to hold a second month unless your own KPIs are already breaching targets. - Response matrix: - Competitor up, your KPIs on target → monitor only. - Competitor up AND your CPA rising → check whether CPC inflation is the transmission mechanism; if yes, route to bidding/quality work. - Your IS down → split by cause first: lost-to-budget routes to budget skills, lost-to-rank routes to improve-quality-score or bid review. - Competitor exited and CPCs eased → opportunity flag: consider expansion or cheaper coverage via allocate-budget-across-campaigns. - Several competitors rising while your position erodes over 2+ months → escalate to a strategic review with stakeholders, not a knee-jerk bid raise. - Never chase absolute-top position for its own sake: being outranked is costless while CPA/ROAS targets are met. Reading the four competitor metrics correctly: - Impression share: the fraction of eligible auctions where that participant showed. It is share-of-presence, not share-of-clicks, a competitor can hold high IS on positions that earn nothing. - Overlap rate: how often you and that competitor appeared in the same auctions, the best single measure of how directly they compete with you. - Outranking share: how often you showed above them (or showed when they did not). Falling outranking share with stable IS means they are winning position, usually via bid or quality. - Position-above rate: of the auctions you both appeared in, how often they sat above you, the head-to-head version of the same question. - Your own top-of-page and absolute-top rates trend how premium your placements are; a slow bleed here often precedes the CPC/CPA symptoms by weeks. ## Common failure modes - Reacting to one month of movement that reverses the next, the two-cycle rule exists for this. - Treating rank as the KPI: winning the position war while losing the margin war. - Campaign-level-only reads that hide keyword-level battles where the CPC pain actually is. - Ignoring the Shopping insights on ecommerce accounts. - No written record, so every month rediscovers the same "new" competitor. Persist the output. - Skipping the KPI correlation and jumping from "competitor grew" straight to "raise bids", the most expensive mistake this procedure prevents. - Comparing IS numbers across different scopes (a campaign-level figure last month vs a keyword-level figure this month) and reading the artifact as movement, keep scope constant across cycles. - Forgetting that Auction Insights only shows participants above a minimum activity bar: small competitors churn in and out of visibility without entering or leaving the market. ## Related skills - Run inside: run-a-monthly-performance-review (parent cadence); quarterly version inside run-a-quarterly-business-review. - Triggered by: run-a-weekly-performance-review IS/CPC flags. - Route follow-ups to: improve-quality-score (rank-driven loss), handle-budget-limited-campaigns (budget-driven loss), allocate-budget-across-campaigns (opportunity capture).