Learning path · Level 7 · Tutorial 5

Diagnose Paid Campaign Performance and Plan Changes

Analyze simulated campaign reports presented in the tutorial and recommend a justified next action. Produce a controlled optimization note covering search terms, audiences, placements, creative, or budget without accessing a live advertising account.

  • Level: Intermediate
  • Duration: 25 min
  • 8 steps
Start tutorial Back to the path
Diagnose Paid Campaign Performance and Plan Changes

What you will go through

  1. Check Whether the Reports Are Comparable Learn to validate reporting periods, attribution, timing, and measurement before making a paid-campaign decision. 3 min
  2. Locate the Most Likely Bottleneck Use delivery, response, conversion, and cost evidence to identify the first performance stage worth investigating. 3 min
  3. Turn Search-Term Evidence into a Focused Change Use a partial, simulated Google Ads search-term report to distinguish irrelevant traffic from relevant queries that need more evidence, then propose one narrowly scoped change. 3 min
  4. Interpret Meta Breakdowns Without Overreacting Interpret simulated Meta audience and placement breakdowns cautiously, then select a narrow action supported by evidence. 3 min
  5. Separate Fatigue from Destination and Tracking Problems Use frequency, response, and downstream evidence to distinguish a plausible creative issue from destination friction or a measurement problem. 3 min
  6. Justify a Limited Budget Decision Use credible purchase economics, attribution limitations, and fulfillment capacity to choose a bounded budget transfer or a justified hold. 4 min
  7. Specify the Change and Its Review Rules Create a focused, reversible campaign change record with safeguards and lag-aware review conditions. 3 min
  8. Write and Defend Your Optimization Note Synthesize simulated Northstar Bags reporting into a concise, controlled optimization note with one focused route, an explicit budget decision, and lag-aware review rules. 4 min

What you will learn

  • Locate the most likely performance bottleneck using delivery, click, conversion, and cost evidence.
  • Identify reporting-window, attribution, or tracking issues that should be checked before changing a campaign.
  • Recommend targeted search-term, audience, placement, or creative changes while acknowledging alternative explanations.
  • Justify a limited budget adjustment and document its expected outcome, safeguards, review timing, and reasons to wait or reverse course.

Before you start

  • Draft a Google Ads Search Campaign
  • Draft a Meta Ads Campaign
  • Specify Search and Meta Ad Creative Variants
  • Choose Ad Bids and Budgets with Realistic Forecasts
  • Create Consistent UTM Campaign Links
  • Read GA4 Acquisition, Engagement, and Key-Event Reports
  • Design and Interpret a Simple A/B Test

Step 1 of 8

Check Whether the Reports Are Comparable

Learn to validate reporting periods, attribution, timing, and measurement before making a paid-campaign decision.

1. Align the reporting frame

Start with comparable report conditions

Before comparing performance, make the reports describe the same thing. For Northstar Bags, use the complete seven-day baseline period, Aug 3–9, 2026, against the complete seven-day comparison period, Jul 27–Aug 2, 2026. All case reporting uses UTC.

Check these conditions before interpreting a rise or fall:

  • Equal, complete date ranges
  • Same time zone
  • Same reporting snapshot date
  • Compatible counting units, such as clicks versus outbound clicks
  • Same conversion definition and attribution basis

The baseline snapshot is dated Aug 18, 2026. A newer incomplete range must be labeled Provisional, not compared as if it were mature evidence.

A comparison is only as sound as its alignment

Use this comparison pattern: align time, range, snapshot, unit, and conversion rule before judging a performance difference.

Two report cards are aligned by matching calendar ranges, time-zone clocks, snapshot icons, and metric columns.

A valid comparison begins by aligning the reports, not by reacting to the largest number.

2. Separate attribution from timing

Credit rules and date rules can produce different totals

An attributed purchase is credit assigned by a reporting rule; it is not proof that advertising created an incremental sale, and it does not establish a new customer.

For this case, Google Ads reports a 7-day click window. Meta reports 7-day click / 1-day view. Baseline comparison tables isolate click-attributed results using interaction-date reporting. Keep Meta view-attributed results separate.

GA4 or store evidence can legitimately differ from platform totals. Their counting and date basis may differ; GA4 may use purchase-event dates, and consent-related collection gaps can reduce observed events. Different attribution methods can also assign credit differently. Check the labels before treating a difference as an error.

Reporting-rule reference

Read each result with its credit and date basis. Do not add platform purchase totals across campaigns and call the result unique orders: one order can be credited by more than one platform.

Simulated data — Source: Northstar Bags reporting rules; Period: Aug 3–9, 2026 UTC; Level: report-definition reference.
SourcePurchase credit shownWindowDate basis for baseline comparison
Google AdsClick-attributed purchase7-day clickInteraction date
Meta click reportClick-attributed purchase7-day clickInteraction date
Meta view reportView-attributed purchase1-day viewShow separately
GA4 or store recordOwn counting ruleOwn ruleState event or order date

Tip

Timing check

A recent report can change as delayed conversions arrive or platform processing finishes. When a reporting range is still incomplete or its attribution window has not had time to mature, label it Provisional. Do not use an immature CPA difference as the reason for a campaign change.

3. Escalate measurement faults first

Look for faults, not just differences

Some discrepancies need investigation before optimization. Warning signs include duplicate purchase events, a sudden unexplained disappearance of purchases while paid orders continue, or campaign tags that stop appearing in traffic records.

Use order and event evidence to decide what to check first. A platform-versus-GA4 difference can be legitimate when rules differ. Repeated events for the same completed order, however, are evidence of a possible measurement fault—not a signal to change targeting or spend.

Purchase-event audit

This small audit is not a campaign performance report. It tests whether the purchase count can be trusted.

Simulated data — Source: Northstar Bags diagnostic audit; Period: Aug 10, 2026 UTC; Level: purchase-event sample.
RecordObserved countOrder-ID evidence
Paid-order ledger1010 distinct paid order IDs
GA4 purchase-event export20Each of the 10 IDs appears twice
Campaign-tagged sessionsPresentTagging is still observed

Warning

Possible tracking fault

Do not interpret the 20 events as 20 purchases. First investigate why a purchase event appears twice per paid order, then use corrected measurement before evaluating campaign outcomes.

4. Decide what must be checked first

Choose the first check

Match each evidence pattern to the first check that should happen before changing a campaign.

Tap an item and then its matching pair.

Step 2 of 8

Locate the Most Likely Bottleneck

Use delivery, response, conversion, and cost evidence to identify the first performance stage worth investigating.

Follow the evidence from delivery to economics

Use a four-stage diagnostic flow

Start at the earliest stage supported by evidence. A report pattern tells you where to investigate first; it does not prove why the problem exists.

  1. Delivery: Was the campaign eligible and able to gain exposure?
  2. Response: Did exposed people click?
  3. Destination conversion: Did clicks become purchases?
  4. Acquisition economics: Are those purchases affordable under the case guardrail?

Move left to right. A downstream result can be weak even when an earlier stage looks healthy. Conversely, low impressions alone do not prove that more budget is the answer.

From exposure to viable acquisition

A four-part flow from ad exposure to click, purchase, and cost evaluation.

Use the sequence to locate the first stage that needs investigation; do not assume a later metric identifies the cause.

Tip

Guard against a premature budget increase

Treat low delivery as a question, not a budget instruction. Check eligibility, settings, audience size, auction conditions, and the reporting period before proposing more spend. Budget can only help if there is credible additional opportunity to buy.

Locate the first concern in the snapshot

Baseline campaign snapshot

The comparison checks from the prior step have been completed for this exercise. Read each platform's results on its own terms, then use the case guardrail to locate the first actionable concern. Do not add platform purchase rows and call them unique orders; neither count establishes a new customer.

Simulated data: campaign-level results

Simulated data — platform campaign reports, Aug 3–9, 2026, UTC interaction-date reporting, campaign level. Google Ads uses 7-day click attribution. Meta uses 7-day click / 1-day view attribution; this table isolates click-attributed results only. Baseline snapshot dated Aug 18, 2026.

Source: Google Ads and Meta Ads simulated campaign reports. Click denominators remain platform-specific.
CampaignImpressionsPlatform click denominatorSpendClick-attributed purchasesClick-attributed purchase CPA
Search | Daypack20,000800 clicks$400.00 USD25$16.00
Meta | Prospecting40,000600 outbound clicks$300.00 USD10$30.00

Example

Example: clicks can look affordable while outcomes do not

For Search | Daypack, CTR is 4.0%, CPC is $0.50, and click-attributed purchase CPA is $16.00—below the $24.00 guardrail.

For Meta | Prospecting, Outbound CTR is 1.5%, outbound CPC is $0.50, but click-attributed purchase CPA is $30.00—above the guardrail. Ten purchases also represent $320.00 of contribution before advertising and fixed costs, using the $32.00 case assumption.

The first investigation is therefore acquisition economics for Meta, not an automatic budget increase. A working hypothesis is that the acquired traffic is less likely to become a viable purchase. An alternative is a destination, order-record, or attribution-counting issue. Corroborate before changing: compare purchase and checkout evidence with appropriately dated store-order records.

Practice: select a stage and corroborating evidence

Diagnose the pattern before acting

Match each short report pattern to the most likely first diagnostic stage and one corroborating evidence check.

Tap an item and then its matching pair.

Step 3 of 8

Turn Search-Term Evidence into a Focused Change

Use a partial, simulated Google Ads search-term report to distinguish irrelevant traffic from relevant queries that need more evidence, then propose one narrowly scoped change.

Classify the search-term signal

Read terms as evidence, not verdicts

A search-term row is evidence, not an automatic verdict. Read five things together: the searcher’s intent, fit with the $80.00 daypack offer, clicks, spend, and whether click-attributed purchase outcomes have had time to mature.

Use the evidence to separate three situations:

  • Clearly mismatched intent can support a narrowly scoped negative keyword.
  • A relevant query with few clicks and no purchases usually needs more evidence, not exclusion.
  • Useful, offer-fitting wording can justify closer alignment in an existing or refined ad group, including an accurate ad-to-destination message.

A visible search-term report is only a partial view. It can support a targeted action on a shown query, but it cannot describe every search that triggered the campaign.

Three evidence-based paths

A diagram shows mismatched search intent leading to a narrow stop, limited relevant evidence leading to observation, and strong relevant intent leading to message alignment.

Intent and evidence strength determine whether to exclude, observe, or refine alignment.

Tip

Scope before scale

Protect relevant traffic. A negative keyword should be as narrow in match type and scope as the evidence allows. Do not use one weak-looking row to block a broader family of daypack searches.

Inspect a compact report

A mature but partial view

The full Search | Daypack campaign spent $400.00 USD during Aug 3–9, 2026. The rows below are not campaign totals: they show only terms visible in one ad group.

The baseline snapshot is dated Aug 18, 2026 and uses UTC. Google Ads is using the case setting of 7-day click attribution with interaction-date reporting. Because the latest interaction was Aug 9, its 7-day click window ended by Aug 16; these displayed click-attributed outcomes have had time to mature for this comparison.

Zero purchases still means different things for different queries. A sewing-pattern search has an obvious offer mismatch. A work-use daypack search may be relevant, but 18 clicks are limited evidence; first verify that the product page can honestly support the implied use.

Visible search terms only

Simulated data — Source: Google Ads search terms report; period: Aug 3–9, 2026; level: Search | Daypack campaign > Daypack Core ad group; interaction-date reporting in UTC; baseline snapshot: Aug 18, 2026.

Click-attributed purchases use the stated 7-day click attribution window. These rows are partial and must not be treated as campaign totals.
Search termClicksSpendClick-attributed purchasesInitial reading
Search term: backpack sewing pattern60$36.000Clearly mismatched intent
Search term: daypack for work18$10.800Potentially relevant; limited evidence
Search term: lightweight daypack150$75.006Useful wording; check claim fit
Search term: college daypack40$24.001Relevant category; limited evidence

Tip

Do not infer the unseen rows

A row missing from this report is not necessarily absent from delivery. Keep the decision limited to the visible evidence, then review additional terms before drawing a campaign-wide conclusion.

Practice: make one narrow exclusion

Choose the focused action

Based on the visible report, which first change best protects relevant traffic while reducing clearly irrelevant traffic?

Apply the evidence

Recommend one protected change

Write a 2–4 sentence recommendation from the visible search-term report. Select one exclusion or ad-group refinement, state its scope, and explain the evidence for it. If you choose a negative keyword, include its match type. Then name one other apparently weak query that should remain eligible for now and explain why it should not yet be excluded. Keep this to one primary change.

Write at least 180 characters (0/180).

Step 4 of 8

Interpret Meta Breakdowns Without Overreacting

Interpret simulated Meta audience and placement breakdowns cautiously, then select a narrow action supported by evidence.

Make the audience comparison visible

Read the whole row, not one metric

Read a breakdown row across its full path: impressions, outbound clicks, spend, and click-attributed purchases. A lower CPC or higher Outbound CTR is only an early signal. Click-attributed purchases and click-attributed purchase CPA indicate whether the traffic appears commercially useful.

For Meta | Prospecting, the baseline campaign report for Aug 3–9, 2026 shows 40,000 impressions, 16,000 reach, 600 outbound clicks, $300.00 USD spend, and 10 click-attributed purchases. This comparison uses UTC, interaction-date reporting, and Meta’s 7-day click attribution. View-attributed results are not included.

A click-attributed purchase is not proof of a unique new customer or an incremental sale.

Uneven delivery changes what a comparison can show

A campaign sends uneven impression volume to two audience rows while campaign-level reach remains separate from the audience-row comparison.

Audience-row impressions can be added when the report supports it. Campaign reach is a separate, potentially overlapping measure and should not be summed across audience rows.

Simulated audience breakdown

Only the additive fields in these rows reconcile to the campaign totals: 40,000 impressions, $300.00 spend, 600 outbound clicks, and 10 click-attributed purchases. Campaign reach is 16,000, but it is intentionally not divided into audience rows because people can qualify for or overlap across multiple audience definitions.

Automated delivery is not a randomized comparison. It may favor one row because of auction conditions, predicted response, available inventory, or other delivery signals not visible here.

Simulated data — Source: Meta Ads audience breakdown; period: Aug 3–9, 2026; level: audience within Meta | Prospecting; UTC interaction-date reporting; 7-day click attribution; baseline snapshot dated Aug 18, 2026.
Audience rowImpressionsSpendOutbound clicksClick-attributed purchasesClick-attributed purchase CPA
Broad prospecting30,000$220.004808$27.50
Interest suggestion10,000$80.001202$40.00

Why an apparent loser may not be a loser

Treat breakdowns as clues, not verdicts

An apparent winner or loser may reflect uneven opportunity rather than a stable audience or placement effect. Eight versus two click-attributed purchases is limited evidence, especially when delivery differs greatly.

Use the same discipline for placements. Feed and Reels can differ in creative crop, device use, viewing context, and destination behavior. A weak placement row is a diagnostic question, not an automatic exclusion. Cheap clicks or low CPM do not establish valuable traffic; low delivery alone does not establish poor traffic.

Simulated placement breakdown

Feed received more than twice the impressions and four times the click-attributed purchases. This does not prove Feed is superior or that Reels should be removed. The report does not establish whether audience composition, creative rendering, device experience, conversion timing, or auction opportunities were comparable.

Simulated data — Source: Meta Ads placement breakdown; period: Aug 3–9, 2026; level: placement within Meta | Prospecting; UTC interaction-date reporting; 7-day click attribution; baseline snapshot dated Aug 18, 2026.
PlacementImpressionsSpendOutbound clicks / Outbound CTRClick-attributed purchasesClick-attributed purchase CPA
Feed28,000$220.00460 / 1.6%8$27.50
Reels12,000$80.00140 / 1.2%2$40.00

Warning

Avoid a one-metric decision

Do not exclude a placement solely because its click-attributed purchase CPA is higher in a small, uneven breakdown. First check whether the creative fits that placement’s format, whether the mobile destination experience is usable, and whether purchase outcomes have matured. Also inspect audience overlap and delivery context before interpreting an audience row as a causal result.

Distinguish a correction from a performance verdict

Match the intervention to the evidence

Choose the narrowest action supported by verified evidence. A confirmed control mismatch can justify a focused correction even when the performance comparison remains uncertain. In contrast, an unsupported audience restriction or placement exclusion can discard useful demand and make the next report harder to interpret.

Keep recommendations within controls actually available in the account, applicable consent conditions, and advertising policies. Do not infer sensitive traits or invent targeting segments from a weak breakdown.

Example

Example of a supported narrow correction

A campaign brief states that Northstar Bags should reach people located in Harbor County. An account check finds that the current geographic setting includes people who merely show interest in Harbor County. That is a verified mismatch with the brief—not a conclusion drawn from click-attributed purchase CPA—and supports correcting the available location control to match the approved requirement.

What to do next

Use this guide before editing. Preserve uncertain audience and placement settings while checking their missing context.

A cautious action addresses what is known while preserving evidence about what remains uncertain.
What the report or audit showsDefensible next actionDo not conclude
Verified geographic control differs from the approved briefCorrect the available location control; keep other major settings unchangedThat a placement caused the CPA gap
Few purchases and uneven audience deliveryGather more comparable evidence; inspect overlap and delivery contextThat the cheaper-click row is more valuable
Weak placement rowCheck creative fit and mobile destination experience before an exclusionThat the placement is inherently poor

Practice: reject the premature cut

Choose a controlled next action

For Meta | Prospecting in Aug 3–9, 2026, Feed recorded 8 click-attributed purchases and Reels recorded 2. Delivery was uneven. The report does not show whether Reels creative rendering, mobile destination experience, audience overlap, or auction context were comparable.

Separately, an account audit confirms that the campaign includes people who only show interest in Harbor County, although the approved brief requires people located there.

Which next action is most defensible?

Step 5 of 8

Separate Fatigue from Destination and Tracking Problems

Use frequency, response, and downstream evidence to distinguish a plausible creative issue from destination friction or a measurement problem.

Frequency is an average

Frequency is an average, not an exposure cap

Frequency describes average repetition within one defined scope and period.

Frequency = impressions ÷ reach

For example, 18,000 impressions delivered to 6,000 reached people produces a frequency of 3.0. It does not mean every person saw the ad exactly three times: some may have seen it fewer times and some more.

Always name the campaign, audience scope, and reporting period before interpreting frequency. A change in frequency can reflect delivery changes as well as a changing audience.

A diagram showing that impressions divided by reach produces an average number of ad exposures per reached person.

Frequency is a campaign-level average for a specified period; it is not an identical exposure count for every person.

Tip

Interpret the pattern, not a fixed threshold

There is no universal frequency value at which an ad has “failed.” Higher repetition becomes useful evidence only when it is read alongside response, delivery conditions, and downstream outcomes.

Read paired trends cautiously

Read paired trends cautiously

The two complete seven-day periods below use the same campaign level and UTC. Reach did not grow, while impressions rose and outbound response declined. That pattern can support a fatigue hypothesis, but it does not establish one: audience composition, placements, promotions, and delivery conditions may also have changed.

Delivery and response comparison

Simulated data — Source: Meta Ads delivery report; level: Campaign, Meta | Prospecting; periods: Jul 27–Aug 2, 2026 and Aug 3–9, 2026; UTC; baseline snapshot dated Aug 18, 2026.

The Aug 3–9 baseline row reconciles to the campaign totals for impressions and outbound clicks.
PeriodImpressionsReachOutbound clicksOutbound CTR
Jul 27–Aug 2, 202632,00016,0006081.9%
Aug 3–9, 202640,00016,0006001.5%

Calculate the comparison frequency

For Jul 27–Aug 2, 2026, calculate Meta | Prospecting frequency: 32,000 impressions ÷ 16,000 reach = ____.

Separate response, destination, and measurement signals

Follow the break in the customer path

Use the point where the path changes to choose the first investigation.

  • A falling outbound response, especially with rising repetition and comparable conditions, makes creative wear or placement-format fit plausible.
  • A stable outbound response followed by weaker product, cart, checkout, or purchase progression points first to the destination experience.
  • A sudden disappearance or contradiction in page views, purchase events, or campaign tags points first to measurement.

These are working hypotheses, not proof. Check equivalent full periods and corroborate the signal before changing an ad.

A customer path from ad exposure to click, product page, checkout, and purchase, with branches indicating response, destination, and measurement investigations.

Locate where the evidence first weakens before selecting a creative, destination, or measurement action.

Choose the evidence path

This guide describes useful first investigations, not causal conclusions.

Observed combinationMost useful first investigationDo not conclude
Frequency rises; outbound CTR falls; conditions are comparableCreative wear or placement-format fitFatigue is proven
Outbound CTR is steady; post-click progression falls; events remain presentDestination experienceThe ad caused the decline
Clicks continue; tagged page views or purchase events abruptly disappearMeasurement implementationDemand disappeared

Apply the diagnosis

Decision packet

Simulated data — Sources: Meta Ads campaign delivery and click-attributed purchases, interaction date; GA4 tagged landing_page_view, event date; level: Meta | Prospecting; UTC. Meta purchase results use the case setting of 7-day click attribution. GA4 page views use their own event counting and date basis.

Both periods are complete. The Aug 3–9 baseline row reconciles to the fixed Meta campaign totals.
PeriodImpressionsReachOutbound clicksOutbound CTRMeta click-attributed purchases (7-day click)
Jul 27–Aug 2, 202632,00016,0006081.9%11
Aug 3–9, 202640,00016,0006001.5%10

Apply the diagnosis

Additional corroboration: GA4 recorded 578 tagged landing-page views in the comparison period and 570 in the baseline period. Audience geography, placement mix, promotion, and the ad itself were unchanged. Delivery was complete. No duplicate purchase-event warning, broken campaign tag, or abrupt loss of landing-page views was found.

Write a short diagnostic note with four parts:

  1. Calculate the baseline frequency.
  2. State the paired trend and the strongest working hypothesis.
  3. Explain why destination friction or a measurement fault is less likely from this packet.
  4. Select one focused next action. Do not propose a budget change.

Write at least 220 characters (0/220).

Step 6 of 8

Justify a Limited Budget Decision

Use credible purchase economics, attribution limitations, and fulfillment capacity to choose a bounded budget transfer or a justified hold.

Compare economics without double-counting

Use separate evidence layers

A budget decision needs two separate evidence layers. First, compare platform outcomes on the same basis: Aug 3–9, 2026, UTC, interaction-date reporting, and click-attributed purchases. Google Ads uses a 7-day click window; the comparison portion of Meta also uses 7-day click attribution. Meta view-attributed purchases stay separate.

Second, use a deduplicated business record and fulfillment capacity to constrain the decision. Northstar Bags has a $700.00 USD seven-day planning budget, a $24.00 USD click-attributed purchase-CPA guardrail, and a $32.00 USD contribution-per-order case assumption before advertising and fixed costs.

Matching click-attribution windows makes the displayed CPA comparison more useful. It does not prove the campaigns are causal alternatives, that attributed purchases are unique orders, or that either campaign created a new customer.

Baseline campaign comparison

Compare reported platform acquisition economics without treating either platform row as a count of deduplicated business orders.

Simulated data — Source: Google Ads and Meta Ads campaign reports; period: Aug 3–9, 2026; level: campaign; UTC interaction-date reporting; click-attributed comparison only; baseline snapshot dated Aug 18, 2026.
CampaignSpendPlatform click denominatorClick-attributed purchasesClick-attributed purchase CPAClick-attributed revenue / ROAS
Search | Daypack$400.00800 clicks25$16.00$2,000.00 / 5.00x
Meta | Prospecting$300.00600 outbound clicks10$30.00$800.00 / 2.67x

Keep capacity separate from attribution

The store ledger is a capacity check, not a way to assign campaign credit. It records 32 distinct paid order IDs across all channels for Aug 3–9, 2026, by order-created date in UTC. Against capacity for 40 orders per week, that leaves numerical room for only eight more orders at this snapshot.

Do not add 25 and 10 platform-attributed purchases and call the result unique orders, new customers, or store orders. Platform credit can overlap. Likewise, do not call the $32.00 contribution assumption multiplied by platform purchases actual business contribution: it is a planning lens applied to attributed outcomes, not a deduplicated profit record.

Do not combine credit with capacity

Use platform outcomes for directional comparison, and use deduplicated all-channel orders plus capacity to limit the risk of scaling.

Two platform report paths overlap before reaching one order ledger and a nearly full fulfillment-capacity gauge.

Overlapping platform attribution is not a deduplicated order total; the order ledger and capacity measure a different business constraint.

Choose a bounded transfer or a reasoned hold

A recent average is not a scaling promise

Search | Daypack has an observed $16.00 click-attributed purchase CPA, below the $24.00 guardrail. Meta | Prospecting has an observed $30.00 click-attributed purchase CPA, above it. This supports reducing exposure to the weaker observed acquisition outcome and cautiously testing whether additional qualified Search demand exists.

It does not mean the next Search dollars will repeat the recent average. Available demand, auction conditions, audience saturation, and delivery can change as allocation changes. The 32 paid all-channel orders also mean capacity can tighten through any channel, not only paid media.

Example

Example: a limited transfer

A bounded option transfers $70.00 from Meta | Prospecting to Search | Daypack. Search changes from $400.00 to $470.00; Meta changes from $300.00 to $230.00. The total remains within the $700.00 planning budget.

The expected direction is less spending in the campaign currently above the CPA guardrail while testing additional Search opportunity. It is not a promise of a particular CPA, purchase count, ROAS, or number of incremental orders. A downside condition is that the transfer should not be extended if mature, appropriately reconciled Search outcomes move above the guardrail or if confirmed all-channel paid orders approach the 40-order weekly capacity.

Warning

When a budget hold is justified

Holding the existing $400.00 / $300.00 allocation can be the stronger decision when purchase tracking is suspect, results are still provisional, the store record cannot be reconciled for the intended use, or fulfillment capacity is uncertain. A hold preserves the fixed planning budget while the uncertainty is checked; it is not an instruction to ignore the reports.

Practice: protect the planning envelope

Select the bounded action

Northstar Bags has a fixed $700.00 seven-day planning budget. Search | Daypack reported a $16.00 click-attributed purchase CPA; Meta | Prospecting reported $30.00. The guardrail is $24.00, and the store recorded 32 distinct paid all-channel orders against capacity for 40. Which action is most defensible?

Apply the decision rule

Make and justify the budget decision

Write a 3–5 sentence recommendation for Northstar Bags. Choose either a bounded reallocation or a hold. State the resulting allocation within the $700.00 planning budget; use the two click-attributed purchase CPA values and the $24.00 guardrail; acknowledge that platform purchases can overlap; and account for 32 distinct paid all-channel orders against 40-order weekly capacity. State an expected direction and one condition that would prevent extending, or would reverse, the decision.

Do not project unchanged returns at higher spend.

Write at least 280 characters (0/280).

Step 7 of 8

Specify the Change and Its Review Rules

Create a focused, reversible campaign change record with safeguards and lag-aware review conditions.

The parts of a usable change log

Build a controlled change record

A controlled change log is not a claim that an edit caused a result. It records a cautious decision so another person can see what was observed, what remains uncertain, and how the decision will be checked.

Use these headings in this order:

  • Evidence — dated observations and their counting basis.
  • Hypothesis — a plausible, unproven explanation.
  • Change — one bounded adjustment.
  • Expected effect — a measurable prediction, not a promise.
  • Guardrails — spend and operational limits.
  • Review — when mature evidence will be checked.
  • Wait or reverse — conditions for holding or undoing the change.

Operational before-and-after monitoring can support a careful next decision. It cannot provide the causal certainty of a randomized A/B test.

From observation to a reversible decision

A seven-part flow showing evidence, hypothesis, one change, expected effect, guardrails, review, and wait-or-reverse conditions.

Each part has a separate job: observation is not proof, and an expected effect is not a guaranteed result.

Protect the interpretation

Change one meaningful variable

Keep the intervention interpretable. A budget transfer is one allocation intervention only when the audience, placements, creative, optimization event, and destination remain unchanged during the review period.

Changing a budget, audience, and creative at once may be operationally tempting, but it creates several explanations for any later movement. Material edits can also disturb delivery or begin a new adjustment period, depending on the platform and setting. Repeated edits make the record harder to interpret.

What each entry must do

Use every heading to make the decision testable.

A change log separates observed facts, uncertainty, action, and decision rules.
EntryWhat it should contain
EvidenceDated observations, source, period, and counting basis.
HypothesisA plausible explanation that remains unproven.
ChangeOne bounded action and what stays unchanged.
Expected effectA directional, measurable prediction; never a promise.
GuardrailsA spending limit, capacity check, or pause condition.
ReviewA date plus mature-data and evidence requirements.
Wait or reverseA reason to hold, stop, or undo the action.

Example

Make the review rule lag-aware

Weak review rule: “Check results next week.”

Usable review rule: “Use a mid-period checkpoint for spend pacing only. Make an outcome decision after a complete business cycle, the stated attribution window, expected reporting latency, and the required purchase evidence are complete.”

A routine checkpoint is not an early declaration of success or failure.

Practice: incomplete evidence

Choose the controlled response

When outcome reporting is not yet mature, the appropriate action may be to __ rather than call a winner or make another edit.

Apply the review rules

Repair the controlled change record

Repair the incomplete record below. Do not rewrite the existing Evidence or Hypothesis. Write five concise lines using exactly these headings: Change, Expected effect, Guardrails, Review, and Wait or reverse.

Evidence: Baseline snapshot dated Aug 18, 2026. For Aug 3–9, 2026 (UTC; campaign level; interaction-date; click-attributed platform results), Search | Daypack spent $400.00 USD for 25 click-attributed purchases, a $16.00 click-attributed purchase CPA. Meta | Prospecting spent $300.00 USD for 10 click-attributed purchases, a $30.00 click-attributed purchase CPA.

Hypothesis: A limited transfer toward Search | Daypack may be more efficient, but additional purchase cost is uncertain.

Decision packet: For Aug 19–25, transfer $70.00 from Meta | Prospecting to Search | Daypack while keeping the $700.00 USD seven-day planning budget fixed. The purchase-CPA guardrail is $24.00. Fulfillment capacity is 40 orders per week across all channels. Use a pacing-only checkpoint during the period; purchase outcomes need the 7-day click window and reporting latency to mature. Platform purchases must not be summed as unique orders; check store orders separately on their own counting and date basis.

Write at least 450 characters (0/450).

Step 8 of 8

Write and Defend Your Optimization Note

Synthesize simulated Northstar Bags reporting into a concise, controlled optimization note with one focused route, an explicit budget decision, and lag-aware review rules.

Read the simulated decision packet

Decision packet

Use this self-contained decision packet for Northstar Bags’ $80.00 daypack. The baseline period is Aug 3–9, 2026, UTC; the comparison period is Jul 27–Aug 2, 2026, UTC. The baseline snapshot is dated Aug 18, 2026, so the baseline 7-day click windows have had time to mature.

Google Ads uses 7-day click attribution. Meta uses 7-day click / 1-day view attribution; the campaign comparison below isolates click-attributed purchases using interaction-date reporting. Keep Meta view-attributed results separate.

The store ledger records 32 distinct paid all-channel order IDs for Aug 3–9, 2026, by order-created date in UTC. GA4 records 64 purchase events by event date. Investigate whether duplicate events or different counting definitions explain that difference before using GA4 to evaluate a change. Platform purchase totals can overlap, so they are not unique orders or new customers.

Case assumptions: $32.00 USD contribution per order before advertising and fixed costs; a $24.00 USD click-attributed purchase-CPA guardrail; a $700.00 USD seven-day planning budget; and capacity for 40 orders per week across all channels. The ledger leaves numerical room for eight more orders at this snapshot.

Baseline campaign report

Simulated data — Sources: Google Ads and Meta Ads campaign reports; period: Aug 3–9, 2026; UTC interaction-date reporting; level: campaign; baseline snapshot dated Aug 18, 2026.

Simulated data. Meta view-attributed results are excluded from this click-attributed comparison.
CampaignDeliveryResponseClick-attributed outcomeSpendCost outcome
Search | Daypack20,000 impressions800 clicks; CTR 4.0%25 click-attributed purchases; $2,000.00 click-attributed revenue$400.00 USD$16.00 click-attributed purchase CPA; 5.00x ROAS
Meta | Prospecting40,000 impressions; 16,000 reach; frequency 2.5600 outbound clicks; Outbound CTR 1.5%10 click-attributed purchases; $800.00 click-attributed revenue$300.00 USD$30.00 click-attributed purchase CPA; 2.67x ROAS

Read the evidence in sequence

A flow from delivery through clicks and purchases to one controlled decision, with a separate caution to reconcile store orders and analytics events.

Use the evidence chain to locate an investigation. It does not prove a cause by itself.

Choose one controlled route

Choose one coherent route

A concise note separates what was observed from what is inferred and predicted.

  • Evidence states the source, period, counting basis, and relevant values.
  • Hypothesis names a plausible bottleneck and an alternative explanation that remains possible.
  • Change names one focused action and makes the budget decision explicit.
  • Expected effect predicts a direction; it never promises a result.
  • Guardrails set spending, CPA, capacity, and unchanged-setting limits.
  • Review waits for mature reporting and required corroboration.
  • Wait or reverse names conditions to defer, stop, or undo the decision.

Prioritize one primary campaign intervention. A reporting audit can occur alongside the decision, but do not combine a search-term, audience, placement, or creative edit with a budget-transfer test in the same review cycle.

Visible search-term evidence

Simulated data — Source: Google Ads search terms report; period: Aug 3–9, 2026; UTC interaction-date reporting; level: Search | Daypack campaign > Daypack Core ad group; baseline snapshot dated Aug 18, 2026.

Visible search terms only. These rows are partial and do not describe all Search | Daypack traffic.
Search termClicksSpendClick-attributed purchasesInitial reading
Search term: backpack sewing pattern60$36.000Clearly mismatched intent
Search term: daypack for work18$10.800Potentially relevant; limited evidence
Search term: lightweight daypack150$75.006Useful wording; check claim fit
Search term: college daypack40$24.001Relevant category; limited evidence

Prioritize rather than fix everything

Select one route for the full note.

Route A — focused search action with a budget hold: Add Negative keyword: backpack sewing pattern (exact) at Daypack Core ad group scope. Keep Search | Daypack at $400.00 and Meta | Prospecting at $300.00. Make no other campaign edit during review. Audit the GA4-to-store discrepancy separately before relying on GA4 to evaluate results.

Route B — focused budget test: Transfer $70.00 from Meta | Prospecting to Search | Daypack, resulting in $230.00 and $470.00 respectively while keeping the $700.00 planning budget fixed. This is the sole campaign change: do not also edit search terms, audiences, placements, creative, optimization settings, or the destination during review. Audit the GA4-to-store discrepancy separately.

The observed Meta click-attributed purchase CPA is above the guardrail, but that pattern does not prove its cause or promise that additional Search spend will repeat its past result.

Summary

Final decision checklist

Before writing, verify that your note:

  • distinguishes an observed pattern from an unproven hypothesis;
  • keeps GA4 purchase events, store orders, click-attributed purchases, and any view-attributed results separate;
  • does not add platform purchases as unique orders;
  • chooses one coherent route and makes the budget decision part of that route;
  • predicts an expected direction rather than a guaranteed outcome; and
  • gives a mature-evidence condition for continuing, waiting, or reversing.

A useful note lets a colleague understand both the decision and its uncertainty.

  • Platform click-attributed outcomes support comparison, but they do not establish unique orders or incremental sales.
  • The 64 GA4 purchase events and 32 distinct paid store orders require a counting audit before GA4 is used to evaluate the decision.
  • A narrowly scoped exact negative can protect relevant traffic when the visible query and ad-group scope support it.
  • A budget transfer must remain bounded, preserve the planning total, and be the sole campaign change during its review.
  • Expected effect is a prediction; safeguards and review rules make the decision reversible.

Apply the evidence chain

Response 1: Diagnose and choose one route

Write the first part of one optimization note in 90–140 words. Use these headings exactly: Evidence, Hypothesis, Change.

Use one route from the prior screen. Explain the reporting check needed before relying on GA4, identify the most likely bottleneck from the decision packet, and name one alternative explanation. In Change, state only the focused action and its explicit budget decision. Keep the action limited to the selected route; do not add a second campaign intervention.

Write at least 420 characters (0/420).

Defend the decision

Response 2: Set safeguards and review rules

Write the second part of the same note in 80–130 words. Use these headings exactly: Expected effect, Guardrails, Review, Wait or reverse.

Continue the route selected in Response 1; do not introduce another campaign change. Justify the budget decision, set a directional expected effect, and protect the fixed planning budget, the click-attributed purchase-CPA guardrail, and the weekly fulfillment capacity. State what remains unchanged. Require mature reporting and appropriately dated store-order evidence before continuing. Name conditions that would make you wait, stop, or reverse.

Write at least 360 characters (0/360).

Step complete

Congratulations! You completed: Diagnose Paid Campaign Performance and Plan Changes

You have converted simulated evidence into a controlled optimization note: validate reporting, separate evidence from hypothesis, prioritize one intervention, and make the decision reviewable and reversible.

Download the App now to have access to + 5000 free courses, exercises, certificates and lots of content without paying anything!

  • 100% free online courses from start to finish

    Thousands of online courses in video, ebooks and audiobooks.

  • More than 60 thousand free exercises

    To test your knowledge during online courses

  • Valid free Digital Certificate with QR Code

    Generated directly from your cell phone's photo gallery and sent to your email

Cursa app on the ebook screen, the video course screen and the course exercises screen, plus the course completion certificate