Check KPIs on the dashboard

On this page you can check your daily, weekly, and monthly ad performance. Compare your main KPIs against a prior period, slice KPIs by label classification, or measure the Before/After impact of a measure you ran. It is all read-only analysis — no ad settings are changed.

About this page

Item	Details
What you can do	Compare your main KPIs (TACoS, ACoS, ROAS, ad spend, sales, orders) against a prior period, plus by-label-classification aggregation and Before/After impact measurement — all in one place
Applies to	The account-wide dashboard (Prompts 1–3) and Sponsored Products (SP)-focused flash reports (Prompts 4–5)
Data & connection needed	Performance data from a connected account (no extra setup). For Before/After comparison (Prompt 3), registering a measure log lets the AI derive the periods automatically
Scope	Read-only (analysis only). The prompts on this page only aggregate and display KPIs; they do not change ad settings

Terms used on this page

• TACoS (Total Advertising Cost of Sales) … ad spend ÷ total sales (ad-attributed + organic). A measure of your brand’s overall reliance on ads.
• ACoS (Advertising Cost of Sales) … ad spend ÷ ad-attributed sales. Lower is more efficient.
• ROAS (Return on Ad Spend) … ad-attributed sales ÷ ad spend. Higher is more efficient.
• CTR (click-through rate) … the share of ad impressions that resulted in a click.
• CVR (conversion rate) … the share of clicks that resulted in a purchase.
• label classification … a system for classifying campaigns and products by axes such as brand, use case, or competitor. Design it in advance to aggregate KPIs along these axes.
• search terms … the words shoppers actually type. These are distinct from the “keywords” that advertisers register.

Watch-outs on sources and parameters

• ROAS is not returned by the tools. State explicitly that it is derived as the inverse of ACoS (ad-attributed sales ÷ ad spend).
• Always state the aggregation target explicitly when aggregating by label classification. If you leave it unspecified, it defaults to search-term labels and will not be the brand axis. In an aggregation that includes DSP, CTR, CVR, and clicks are structurally unavailable (treat them as not retrievable).
• Fix the numeric source for a Before/After measure comparison to one. Mixing it with values from the dashboard or the placement tools puts you a few percent off, because the aggregation source and the attribution window differ.
• Do not treat values that could not be retrieved as 0. State “not retrieved” and “no data” as distinct cases.

When to use this

• You want to compare your main KPIs against the same prior period every day or week to see what changed
• You want to slice and compare KPIs by label classification such as brand or use case
• You want to measure how ad spend, sales, ACoS, and ROAS moved before and after a measure you ran
• You want to know which search terms drove sales the most in a measure
• You want to check this week’s Sponsored Products (SP) performance week-over-week
• You want a flash report of today’s Sponsored Products (SP) sales

Before you run

• Connected account … if you handle multiple accounts, confirm the target account is correct first.
• Target period … each prompt has a default. Because same-day and previous-day data may not be finalized, a recent finalized period is safest.
• Label classification design (prerequisite for Prompt 2) … if you haven’t designed it yet, set it up first in Classify ads with labels and naming rules.
• Measure log registration (if you use a measure ID in Prompt 3) … register it in Log initiatives and external events and verify their impact so the period can be derived automatically from the measure ID.
• If you register targets in Set targets for sales, ad spend, and ACoS, it becomes easier to compare progress against the current month’s targets.

Quick reference

Goal	What you get	Changes	Prompt
Compare the recent and prior periods to check main KPIs	A comparison table of current / prior / delta	No change	Prompt 1
Aggregate KPIs by label classification	A by-label-classification metrics table with prior-period comparison	No change	Prompt 2
Measure the Before/After impact of a measure	A before/after delta and verdict table, plus a ranking of the search terms that drove it	No change	Prompt 3
Check this week’s SP week-over-week	A per-campaign week-over-week table with change flags	No change	Prompt 4
Get a flash report of today’s SP sales	A flash table of today’s sales, orders, and top campaigns	No change	Prompt 5

Everything on this page is “check only”

As the “Changes” column shows — every entry is “No change” — the prompts on this page only aggregate and display KPIs; they do not change ad settings. Once you spot an anomaly, make bid, budget, and other changes by following the “Next steps” of each prompt below to the related page.

---

Prompt 1: Compare the recent and prior periods to check main KPIs

Check TACoS / ACoS / ROAS / ad spend / sales / orders for the recent period, consolidated as a comparison against the prior period of the same length.

When to use — when you want to compare your main KPIs against the same prior period every day or week to see what changed.

What you need — the target period (default is the last 7 days).

Fetch the ad dashboard for {{ACCOUNT}} and compare the last {{DAYS (e.g., 7)}} days against the
immediately preceding period of the same length.

Steps:
1. First check the ingestion status and exclude days that are not finalized (today, and days still being
   ingested). Take the last finalized day as the end point, treat the {{DAYS (e.g., 7)}} days up to it as
   the current period, and the same number of days immediately before that as the prior period. Always
   state the finalized date range you actually used in the body of your answer.
2. Fetch the dashboard at daily granularity with prior-period comparison enabled, and return TACoS / ACoS /
   ad spend / total sales / ad sales / orders. Use the value / previous_value / delta the tool returns as-is
   for the figures and the prior-period comparison; do not recalculate them yourself. Also confirm whether
   the prior period was actually retrieved.
   * This tool returns the period total. When you need a day-by-day breakdown, retrieve one day at a time.
3. ROAS is not returned by the tool, so calculate it as the inverse of the returned ACoS
   (= ad sales ÷ ad spend) and state explicitly that it is a "derived value."
4. Use the currency field the server returns as-is (do not infer it from the account name).
5. Do not treat metrics you could not retrieve as 0 — write them as "not retrieved" or "no data," kept
   distinct from each other.

Criteria:
- ACoS worsened by more than {{DEGRADATION_THRESHOLD (e.g., +2pt)}} versus the prior period → verdict
  "needs investigation"
- Sales or orders fell by more than {{DEGRADATION_THRESHOLD (e.g., -10%)}} versus the prior period → state
  explicitly that a guardrail was breached
- If neither applies, the verdict is "keep monitoring"

Finally, output the following three points.
1. Recommended actions — if the verdict is "needs investigation," show whether N-gram analysis (search-term
   driver) or SQP analysis (visibility driver) should come first, with target, reason, and priority. If the
   verdict is "keep monitoring," say so and name the metric to look at first the next time things worsen
2. How to execute — which prompt to hand this result to (identify it by its number or by the name of a new
   prompt). Include the arguments to pass as well (the finalized date range, the ACoS delta, and the gap
   between the growth rates of ad spend and ad sales)
3. Measurement — on {{EVALUATION_DATE (e.g., 7 days later)}}, retrieve again under the same conditions (the
   same number of days, ending on the last finalized day) and judge improved / flat / worsened against a
   band of {{VERDICT_BAND (e.g., ±0.5pt)}} on ACoS. Re-check the guardrail KPIs at the same time

As a note, add one line stating that for ACoS and TACoS the tax treatment of the ad spend in the numerator
is undetermined and so they are not identical to the values shown in the SaaS screens, and that even
finalized periods can change value on re-ingest.

What you can change — the target period (any length you like, such as “last 7 days,” “this week,” or “last 28 days.” If you don’t specify one, it uses the last 7 days).

What you get back — a three-column layout of current / prior / delta (both absolute and %) for ad spend, sales, orders, ACoS, TACoS, and ROAS. The currency is whatever the server returns in its currency field (it is not inferred from the account name). ROAS is a derived value calculated as the inverse of ACoS.

Changes — read-only (analysis only). No ad settings are changed.

Next steps — if you spot an anomaly, go to Adjust bids. To compare progress against the current month’s targets, register targets in Set targets for sales, ad spend, and ACoS.

Current limitations

• If you request a monthly aggregation over a roughly 30-day range, the totals may come back as 0. In that case, rerun with a daily aggregation
• The current and prior periods are always aligned to the same number of days. If you provide mismatched lengths, the AI asks for confirmation
• If the period includes data that hasn’t finalized yet (the last 1–2 days), the AI notes that at the top

---

Prompt 2: Aggregate KPIs by label classification

Use a pre-designed label classification system to slice and compare KPIs by classification axes such as brand, use case, or competitor.

When to use — when you want to slice and compare KPIs by label classification such as brand or use case.

What you need — a label classification that is already designed. The target period (default is the last 28 days) and the classification axis (default is all registered ones).

Aggregate KPIs by label classification for the last {{WEEKS (e.g., 4)}} weeks along the brand axis (use
analyze_label_kpi; compare_measure does not support by-label aggregation).

Prerequisites (must be followed):
- Always state apply_to explicitly. If you leave it unspecified it defaults to SEARCH_TERM (search-term
  labels) and will not be the brand axis.
  To line up the nine metrics impressions / clicks / ad spend / sales / orders / CTR / CVR / ACoS / ROAS,
  use apply_to=SA (Sponsored Ads only; DSP not included).
  To include DSP, use apply_to=ALL — but clicks / CTR / CVR are structurally unavailable there, so state
  them explicitly as "not retrievable (null)."
- Label tree (root label) names differ by account. First present the list of root labels in the response,
  declare in one line which root label you adopted as the "brand axis" and on what grounds, and only then
  aggregate. If there is no root that corresponds to a brand axis, do not aggregate — report that "no
  corresponding axis has been set up" and stop.
- analyze_label_kpi has no prior-period comparison argument, so call it again for a prior period of the same
  number of days and compare.
- Keep labels with zero actuals, not retrievable (null), and unassigned distinct in your notation. Do not
  drop a 0 as if it were missing data.
- Do not total rows across different roots (the same entity gets double-counted). If you need an
  account-wide figure, use get_dashboard.

Return impressions / clicks / ad spend / sales / orders / CTR / CVR / ACoS / ROAS, and compare against the
prior period too.

Criteria:
- Extract brands whose ACoS worsened by more than {{DEGRADATION_THRESHOLD (e.g., +5pt)}} versus the prior
  period as "needs action," and present them prioritized in descending order of their share of sales.
- Guardrail check: even when ACoS improved, present brands whose sales fell by more than
  {{SALES_DECLINE_THRESHOLD (e.g., -10%)}} versus the prior period separately as "needs review (driven by
  contraction)." Always include the change in ad spend and impressions as well, so it is possible to tell
  whether the ACoS improvement came from scaling back.
- If there are zero "needs action" brands, state "0 items" explicitly, then present the top
  {{WATCHLIST_COUNT (e.g., 3)}} brands by worst ACoS as a watchlist.

Finally, output the following three points.
1. Recommended actions — for each "needs action" and "needs review (driven by contraction)" brand, show
   whether to dig deeper with N-gram analysis (search-term level) or with per-campaign analysis, along with
   reason and priority
2. How to execute — which prompt to hand this result to (identify it by its number or by the name of a new
   prompt)
3. Measurement — on {{EVALUATION_DATE (e.g., 4 weeks later)}}, retrieve again under the same conditions
   (same apply_to, same root label, same number of days) and check whether the "needs action" brands' ACoS
   improved and whether the "needs review" brands' sales recovered

What you can change — the target period (if unspecified, the last 28 days) and the classification axis (brand, use case, competitor, etc. If unspecified, all registered classifications).

What you get back — a by-label-classification metrics table (impressions, clicks, ad spend, sales, orders, CTR, CVR, ACoS, ROAS) with prior-period comparison. When you aggregate along the search-term axis, the actual search terms from Auto Campaigns are included as well.

Changes — read-only (analysis only). No ad settings are changed.

Next steps — design or revise your label classification in Classify ads with labels and naming rules.

---

Prompt 3: Measure the Before/After impact of a measure and identify the search terms that drove it

For a completed measure, aggregate the deltas in ad spend, sales, ACoS, and ROAS across the Before and After periods, and classify the result as improved / neutral / degraded. On top of that, use contribution analysis to surface the search terms that drove the change in sales the most before and after the measure — so you dig into what actually worked.

When to use — when you want to measure how a completed measure affected ad spend, sales, ACoS, and ROAS.

What you need — the before and after periods (same number of days). If you specify a measure ID, it must already be registered in the measure log.

Show a Before/After comparison for a measure.
- Before period: {{BEFORE_START (e.g., 2026-05-01)}} – {{BEFORE_END (e.g., 2026-05-14)}}
- After period: {{AFTER_START (e.g., 2026-05-15)}} – {{AFTER_END (e.g., 2026-05-28)}}
- Measure ID: {{MEASURE_ID (e.g., evt_2026_05_18_cpc)}} (optional; the ID registered in the measure log)

[Check these before you run]
1. Number of days — the before and after periods must have the same number of days. If they differ, do not
   aggregate: propose aligned periods and ask for confirmation (a difference in days comes through directly
   as the rate of change in sales and orders, which breaks the verdict).
2. Data finalization — set the end date of the after period at least 7 days before the day you run this. If
   it includes a more recent day, aggregate only after stating explicitly that sales, orders, ACoS, and ROAS
   are preliminary values.
3. Existence of the measure ID — if a measure ID is specified, first fetch the list of measure logs and
   confirm that the ID actually exists. A non-existent ID is silently ignored and returns the same result as
   omitting it, so always confirm before using it. If not a single measure log has been registered, state
   explicitly that this delta cannot be shown to be attributable to any specific measure and is merely the
   difference between two periods (including seasonality, inventory, and external factors).

[Aggregation]
Aggregate the deltas in ad spend, sales, ACoS, ROAS, and orders.
Fix the source of the figures to the Before/After comparison tool, and do not mix them with figures from the
dashboard or the placement tools (even for the same period the aggregation source and the attribution window
differ, which puts recent periods a few percent off). State the source you used at the top of the report.

[Verdict] Classify as improved / neutral / degraded.
- Improved: ACoS fell by at least {{ACOS_IMPROVEMENT_MIN (e.g., 1.0)}} percentage points, and both sales and
  orders stayed within {{TOLERANCE_FLOOR (e.g., -5%)}}
- Degraded: ACoS rose by at least {{ACOS_DEGRADATION_MIN (e.g., 1.0)}} percentage points, or either sales or
  orders fell below {{TOLERANCE_FLOOR (e.g., -5%)}}
- Neutral: neither of the above applies
Judge the ACoS change in percentage points (absolute difference). You may show the relative rate of change
as well, but do not use it for the verdict.
Also show the improved / neutral / degraded verdict the tool returns. If the tool's verdict and the verdict
from these criteria disagree, **give precedence to these criteria**, and state the fact that they disagreed
along with the likely reason.

[Isolating the drivers] If the verdict is "degraded" or "neutral," present a breakdown from these angles.
- Search Term — contribution analysis by search term
- Placement — the change in ACoS and cost share by placement (fetch each of the Before/After periods and
  compare)
- Bid / Budget — there is no tool that returns the Before/After delta directly, so cross-check the change
  history over the periods and present it, stating explicitly that it is "an estimate from the history"
The per-campaign breakdown only returns the top entries by ad spend, so always add what percentage of the
total that breakdown accounts for.

Finally, output the following three points.
1. Recommended actions — show Keep (continue the measure) if "improved," Revert (roll back) if "degraded,"
   and {{ADDITIONAL_OBSERVATION_PERIOD (e.g., 2 weeks)}} of continued observation if "neutral," together
   with the target and the reason
2. How to execute — if you Revert, apply it as a bid change only after the user has reviewed and approved
   the proposal. If you Keep, **present a proposal** to update the measure log status to "complete"
   (updating the status is a write operation, so do not execute it here — do it after approval). If no
   measure log is registered, present the fact that a measure has to be registered first
3. Measurement — on {{EVALUATION_DATE (e.g., 2 weeks after the measure log is updated)}}, rerun the same
   analysis and check whether the verdict changed (the pre-run checks 1–3 above must be satisfied on the
   rerun as well)

What you can change — the before and after periods, and the measure ID (optional). If you specify a measure ID, the AI derives the periods automatically from the measure log, so you can skip entering the dates.

What you get back — a four-column table of before / after / delta / verdict (improved, neutral, degraded) with a per-campaign Top 10, plus a ranking of the search terms that contributed most to the change in sales. Period lengths are always aligned to the same number of days.

Changes — read-only (analysis only). No ad settings are changed.

Next steps — register and log measures in Log initiatives and external events and verify their impact.

Contribution analysis is unique to Picaro

Beyond a plain Before/After delta, contribution analysis decomposes which search terms actually drove the rise or fall in sales — so you grasp the “why” behind an improvement or degradation in one shot. That’s a Picaro strength.

---

Prompt 4: Check this week’s Sponsored Products week-over-week

Show this week’s Sponsored Products (SP) performance with a prior-week comparison flag. Items that moved by 20% or more versus last week are flagged with a warning.

When to use — when you want to check this week’s Sponsored Products performance week-over-week.

What you need — the threshold for flagging a change (default is 20% or more versus last week).

Show this week's Sponsored Products performance.
By campaign, display spend / ACoS / ROAS / CTR, and
flag any item that moved by 20% or more versus last week.

What you can change — the change-flag threshold (20% versus last week in the example). To make it more or less sensitive, change this percentage.

What you get back — a table of campaign × spend / ACoS / ROAS / CTR × week-over-week × change flag. Items that moved by 20% or more versus last week are flagged, making it usable as a weekly operations snapshot.

Changes — read-only (analysis only). No ad settings are changed.

Next steps — if you spot a large change, respond in Adjust bids or Adjust budgets.

---

Prompt 5: Get a flash report of today’s Sponsored Products sales

Show a flash report of the total sales attributed to Sponsored Products (SP) today.

When to use — when you want a flash report of today’s Sponsored Products sales.

What you need — nothing in particular (today’s figures are shown as preliminary values).

What are today's total Sponsored Products attributed sales?
Show sales / orders / top 3 campaigns.

What you can change — (none; this just shows today’s flash report).

What you get back — a table of today’s sales / orders / top 3 campaigns (today’s figures are preliminary).

Changes — read-only (analysis only). No ad settings are changed.

Next steps — to check the weekly performance with finalized figures, go to Prompt 4.

---

Related categories

• Classify ads with labels and naming rules — prerequisite for Prompt 2 (label classification design)
• Log initiatives and external events and verify their impact — pairs with Prompt 3
• Create reports — comprehensive reports including dashboard KPIs
• Set targets for sales, ad spend, and ACoS — compare progress against the current month’s targets
• Adjust bids — move here once you detect an anomaly
• Adjust budgets — move here once you detect a budget-related change