Skip to content
Ecomwith

Meta Ads Structure: Roles of Campaign, Ad Set, and Ad

  • Beginner
  • 50 minutes

Commute, gym, gifts, outdoor: one ad set per interest quickly fills eight rows in Ads Manager. With only $80 a day, how many results can each group collect? Which differences deserve a separate share of the budget?

In this classroom example, a Shopify store sells a 20oz tumbler: $49 for a cup and $89 for a gift bundle. It is testing US website purchases first, with Canada deferred. The launch budget is $80 a day; the original sketch divides it across 8 interest ad sets. This is not a record of actual campaign performance. The estimates below depend on their stated assumptions.

By the end, you should be able to place settings at the right level, decide whether existing groups need combining, separating or more observation, and build a draft you can inspect. You can learn the structure without access to an account. To perform the steps, you need editing permission, a working product page and accurate purchase reporting.

A purchase event reports a completed website purchase to Meta. The browser Pixel or server-side Conversions API (CAPI) can send it. Missing events, duplicates and wrong values need fixing first; changing the structure cannot repair the signal. The Pixel and CAPI lesson explains those checks, but it is not required reading for the three levels below.

Separate the three jobs: objective, audience, creative

The campaign sets the direction for this round

Choose the overall objective and budget strategy at campaign level. This case wants website orders, so it uses Sales. A campaign can contain one or more ad sets.

The ad set defines the delivery conditions

The ad set controls who can see the ads, where they appear, the schedule and the result the system should pursue. If you choose ad set budgets, their amounts live here too.

Separate the two budget terms: CBO puts the budget at campaign level; ABO sets a budget for each ad set. This changes how money is allocated. Audience settings and Custom Audience exclusions still belong to the ad set. With Advantage+ audience, distinguish suggestions from controls: use a Custom Audience exclusion to leave purchasers out, rather than entering that list as an audience suggestion.

Choosing Sales does not finish the Purchase setup. The ad set's performance goal specifies the result to pursue; this case uses website conversions and the Purchase event so delivery learns from website purchases.

Attribution also lives at ad set level. Its settings determine which ad interactions and later conversions can receive credit, affecting counting and delivery. Credited orders alone do not prove how many additional orders the ads caused.

The ad contains what the shopper sees

An ad holds the image or video, copy, link, format and the Facebook Page or Instagram identity representing the store. The opening message intended to catch attention is often called the hook. URL tracking parameters are added at this level too.

Ads in one ad set share its settings, including audience, placements, budget and schedule. Three videos changing only the opening can be managed as three ads in that group without a separate ad set for each.

The three layers of the tumbler campaign
  1. Campaign1

    Why this money is spent

    In the UI: Ads Manager › Campaigns

    • Example: Objective (Sales this round) · budget strategy · the budget itself, with campaign budget
    1. New-customer ad set1

      Who sees it and which result to pursue

      In the UI: Ads Manager › Ad setsThe system learns here

      • Example: Audience and exclusions · placements and schedule · performance goal and conversion event · attribution setting · budget amount, with ad set budgets
      1. Ad3

        What people see

        In the UI: Ads Manager › Ads

        • Example: Format · image or video · text and link · Page and Instagram identity · URL parameters
    2. Existing-customer ad set1Optional

      Past buyers only, with a spend cap

      In the UI: Ads Manager › Ad sets

      • Example: Custom audience of past buyers · ad set spending limit

The past-buyer group in the map is a conditional extension. It is not a required second group for a launch. Use the two questions below to check the distinction between audience settings and creative.

Self-check · 2 questions0/2 answered

Tell the layers apart

Question 1 / 2A checked Custom Audience of purchasers in the last 30 days already exists. With CBO or ABO, where do you apply it as an exclusion?
Question 2 / 2Three videos differ only in the opening hook. How should they be set up?

Self-check not answered

This question concerns applying an exclusion to the current ad set, after the purchaser Custom Audience has been created and checked. Meta's new- and existing-customer setup guidance also sets inclusions and exclusions within ad sets. Account-level audience controls are separate general limits that may apply automatically; they are not the CBO/ABO budget choice.

Delivery gives different ads different exposure. Keeping them together avoids unnecessary fragmentation, but normal delivery results are not a randomized A/B test. They can suggest the next creative test.

What remains for each ad set when the budget is split?

Start with the daily allocation

An equal split of $80 across 8 groups leaves $10 each. Two groups get $40 each. One group keeps the original total.

Equal allocation makes this relationship easy to see. Actual campaign budget allocates dynamically: most may go to one group, and the full budget may not spend.

Estimate weekly purchases per ad set

Cost per result is spend divided by the corresponding result count. Here we assume a fixed $20 per purchase. That cost is a classroom assumption.

Real delivery costs change. In Tinuiti's published Q2 2026 client benchmark, Facebook cost per thousand impressions (CPM) was 13% higher than a year earlier. This is an observation from its client sample; the public summary does not disclose the Facebook account count. It is neither a market-wide average nor a purchase cost. Use your own purchase CPA to estimate purchases; CPM cannot stand in for cost per order. [1]

Assume full spend, equal allocation and an unchanged purchase cost. Daily allocation × 7 ÷ purchase cost estimates a group's weekly purchases. Eight groups give $10 × 7 ÷ $20, or 3.5 each. One group gives $80 × 7 ÷ $20, or 28.

Weekly purchase estimates under equal allocation
  • Split into 8 ad sets, each3.5 purchases/weekPer ad set: daily budget × 7 ÷ cost per purchase
  • Concentrated in 1 ad set28 purchases/weekWhole daily budget × 7 ÷ cost per purchase

Reference line: Meta learning reference: about 50 results in the week after the last significant edit · ~50 purchases/week

Classroom arithmetic: full spend, equal allocation and a fixed $20 purchase cost. These are not measured results.

The eight smaller groups still add up to the same modeled total. Consolidation concentrates budget and learning opportunities without creating extra orders in this calculation. It does not establish an unchanged purchase cost after merging. The fraction is an average estimate, not part of a real order.

Read the learning status in context

After creating an ad or ad set, or making a significant edit, the system explores how to deliver it. Learning in the Delivery column means performance is typically less stable and cost per result can be higher. Learning is normal; it should not be avoided altogether.

Meta's learning-phase guidance says ad sets exit when they can deliver stably. This usually occurs after about 50 results within the week following the ad set's last significant edit. This figure applies to one ad set, not the campaign total. This example uses website purchases; other conversion locations can have different conditions.

If the system expects too few optimization events, it can show Learning limited. That is not a penalty. Inspect audience size, budget, cost controls and event frequency. The status does not by itself mean the store is losing money or needs more spend.

Key point A delivery status cannot set your loss limit

About 50 is a learning reference, not a profitability test, universal permission to split or instruction to increase spending. Check actual orders, margin and affordable loss first.

Self-check · 1 question0/1 answered

Check the budget judgment

A new account spends $50 a day across 8 interest ad sets, with no different product or market tasks. What comes first?

Self-check not answered

Optional practice with your own budget

The worked example already explains the relationship. If you have a budget and historical purchase cost, try them below; otherwise continue to the budget choice.

Calculator · change the numbersExample data

Optional practice: use your own numbers

Try changing these numbers.

Apply a set of numbers

USD

The campaign's budget per day, in dollars

Drag or enter a count to see how many purchases each ad set gets

USD

The $20 default is a classroom assumption; use your own account's number if you have one

Bar unit: purchases/week

Total purchases per week

28.0

Full spend at a fixed cost gives about 28 results a week in this model. It is an estimate, not a forecast.

Purchases per ad set, weekly

3.5

This assumes equal allocation. Compare with Meta's learning reference, then check the actual Delivery status in your account.

Groups at the reference count

0

Estimated total divided by about 50, rounded down. Zero means one group falls short in this model; it does not mean you cannot advertise or must increase spending.

10x budget guide

$200

Meta's general rule of thumb: a daily budget of at least 10 times the average cost of the performance goal.

The green bar shows estimated total weekly purchases; the clay bar shows the estimate per ad set under equal allocation. They use the same unit and scale. Change the ad set count and watch the per-set bar. The colors identify the two quantities; they do not mean profit or loss.

The defaults give zero groups at the reference count. Change the group count to one: its weekly estimate rises from 3.5 to 28 without increasing the total. Zero means the model falls short of that reference, not that advertising is prohibited.

Budget presets keep the current group count and restore the classroom purchase-cost assumption. If available, enter a historical cost for the same event and reporting window. Actual allocation, costs and conversion delays can still differ.

What needs maintaining as the budget grows?

The calculator's three budget presets also illustrate different maintenance demands. More budget does not call for proportionally more ad sets. Keep the same tumbler store in mind: these are classroom situations, not spending thresholds that determine every store's structure.

What someone needs to maintain at three budget scales

SituationExample daily budget (USD)Work to maintainBefore adding structure
Small-scale test50Maintain one market, one advertising task and one review path. Reconcile spend, Purchase events and orders.Check that the existing group can collect useful evidence and that someone can review it. Do not copy a larger account's group count.
Starting to expand500Each additional market, product line or customer task needs readable spend, order-quality evidence and an observation window.Name the real business difference and the person who will keep reviewing it. A larger budget alone is no reason for more interest groups.
Several tasks running5,000Assign responsibility for creative versions, stock, support and fulfillment as well as delivery. Respond to supply or page-promise changes.Give each task a clear scope, evidence, owner and review time. Budget cannot replace those conditions.

At the same budget, different prices, margins and purchase costs support different workloads. Confirm that someone can keep checking a proposed structure before adding reports with no reviewer.

Which layer allocates the budget?

With campaign budget, you set one total and the system looks for opportunities across ad sets without promising equal spend. Start by assessing total campaign results and average cost per result. An ad set spending less is not necessarily worse.

With ad set budget, you control amounts and schedules separately. Consider it when customer value, audience size, performance goals or bid strategies differ substantially, or a test needs its own budget control. The level holding the budget is a separate setting from an ad set's minimum or maximum spending limit.

Choose who allocates the budget

  • Who allocates it

    Campaign budget
    The system, dynamically across groups
    Ad set budget
    You set each group's amount
  • Where to start reading

    Campaign budget
    Campaign results and average cost
    Ad set budget
    The corresponding ad set's results and spend
  • What matters more

    Campaign budget
    Groups matter equally; judge the whole and allow flexible allocation
    Ad set budget
    Separate control for different tasks or funding requirements

The tumbler store starts with $80 in campaign budget and one US new-customer group. Meta says budget behavior with one group is the same as ad set budget. Canada waits until its page and fulfillment promises are ready.

Daily budget is an average, not a strict maximum for every day. Monitor actual spending and establish an affordable loss before choosing the test's scale.

What do CBO, ABO and Advantage+ mean in the budget settings?

Campaign budget is Advantage+ campaign budget, formerly campaign budget optimization, often shortened to CBO. Ad set budget is commonly called ABO. Sales, Leads and App promotion select Budget strategy at campaign level; Sales campaigns have campaign budget on by default.

For the budget section's Advantage+ state, selecting campaign budget keeps it on with one or more groups. Multiple groups without campaign budget turn that section off. This alone does not establish that the whole campaign meets every Advantage+ condition.

Advantage+ sales campaigns are the renamed Advantage+ shopping campaigns, called ASC in older material. Automation for audiences, placements and budget still operates within the three levels. It does not define your business task, naming or review standard.

Ad set budget sharing is another option: groups can share up to 20% of daily budget. Inspect the options actually available in the account and the amount of control you need.

What eligibility and spend limits matter with several ad sets?

Campaign budget requires the same budget type and bid strategy across groups, with standard delivery. The optimization event must also match under the highest-volume bid strategy. Ad set budgets remain subject to their own eligibility, event and budget requirements.

Campaign budget supports ad set minimum and maximum spend limits. Meta recommends using as few as possible to avoid restricting allocation. If an established and a new market share budget, existing opportunities may take more of it. Ad set budget or a minimum can reserve test space, but inspect actual spend afterward: neither full delivery nor profit is guaranteed.

Why can actual daily spending exceed the configured amount?

Meta's budget guidance describes a gradual rollout of greater daily flexibility to some accounts: a day can spend up to 75% above daily budget. With the daily budget unchanged, spend in a calendar week (Sunday–Saturday) is capped at 7 times that budget. Do not assume this rollout condition describes every account or judge a whole week from one day. Inspect the actual account settings and spend.

Can you simply apply Meta's 10x budget guideline?

Meta's general performance-goal guidance recommends daily budget of at least 10 times the performance goal's average cost. At the classroom $20 purchase cost, that is $200 a day per ad set; run eight ad sets and the same rule asks for $1,600 a day, not $200.

This general guideline answers a different question from the learning reference of about 50. Neither replaces margin, cash and affordable-loss judgment. Do not automatically spend that amount to exit learning.

For Learning limited, assess options such as combining similar groups, broadening the audience, increasing budget or cost-control room, and choosing a more frequent event. These are alternatives to evaluate, not a sequence that must be followed.

Temporarily choosing AddToCart changes the task to cart additions. Those are not purchase results. Verify the event, define when Purchase optimization returns, and keep checking real orders. Without those conditions, retain Purchase and accept that Learning limited may remain.

Should these ad sets be combined or separated?

Find the business difference first

Market promises include currency, shipping, returns, free-shipping thresholds and pages. Product economics include margin, average order value, stock and fulfillment costs. Use the store's product or variant identifier, its SKU, to connect results to the right item. Establish evidence of the difference before adding independent budgets or delivery boundaries.

New customers have not purchased; existing customers have. Visitors and cart abandoners can still be new customers. Retargeting means advertising again to previously engaged people; it can include nonbuyers as well as past buyers. Those are not automatically the same task.

ROAS divides ad-attributed sales by ad spend. A good blended ROAS does not independently establish effective new-customer acquisition. Inspect new-customer order share, frequency and refunds too.

Seer Interactive published a restructuring case for one client: cost per link click fell from $1.01 in Jan–Aug 2022 to $0.59 in the same months of 2023. The source uses a dollar symbol without specifying the currency. Placements and ad types changed too, and the case reports no purchase CPA or profit. It illustrates a consolidation project, but cannot isolate consolidation's effect on click cost or establish that the tumbler store would make a profit. [2]

Research supports the distinction between attribution and incrementality. Gordon and colleagues' peer-reviewed study in Marketing Science compared U.S. Facebook ad experiments with observational estimates. Those estimates often failed to reproduce the randomized results, even after extensive demographic and behavioral adjustment. Attribution is not a substitute for an incrementality experiment. The study did not compare CBO with ABO or establish that consolidation would improve this store's profit. [3]

A different label alone does not need another group

Interests, creative formats, placements, languages or small locations usually do not justify a split by themselves. Meta recommends combining similar groups, using multiple languages within one group and Advantage+ placements where appropriate. Use supported reporting breakdowns before creating a delivery boundary for another report row.

A split also needs budget, sample and review capacity. Gather evidence or leave an unaffordable line out. Do not combine incompatible currencies, shipping or returns merely to concentrate budget. When separate control is justified, choose a campaign, ad set budget or spending limits.

UTM parameters are source labels in a link. This case uses utm_campaign for the campaign and utm_content for the creative. They distinguish labels actually recorded; they cannot reconstruct unrecorded interests or customer groups.

How can you tell whether ads added orders?

A holdout is a group intentionally kept from seeing the ads being tested, compared with a group that can see them over the same measurement window. Meta's lift-test explanation uses random assignment to help distinguish advertising effects from purchases that would have happened anyway. This differs from an A/B test comparing two creatives.

An unadvertised Canadian market is not automatically a valid control for the US: customers, page promises and fulfillment can differ. Check the group assignment, measurement window and comparison method first. Attributed orders or a difference between markets alone cannot establish incremental orders caused by advertising.

Choose a concrete situation and its first action

The tumbler store's 8 interest groups sell the same product in the same market for the same task. Inspect whether they can be consolidated while keeping the product page, Purchase event, offer and existing creative fixed. Schedule a separate creative round afterward.

The five classroom situations below each need one first action. Choose before reading the reasoning, and distinguish reviewing separately from immediately splitting delivery.

Exercise · pick oneCheckpoint

Five structure problems at the tumbler store

① Choose the situation
② What would you do

Combine ad sets without a business difference to concentrate the budget and purchase events in one ad set.

Separate products whose margin, order value, stock or landing page differ materially, at least in the review.

Inspect audience sources, purchase status, frequency and order quality separately; isolate delivery only when justified and funded.

Compare formats and hooks as different ads in one ad set.

Run markets separately when currency, shipping, returns or the page promise differ.

Current setup
This round uses Sales and website Purchase events. The team created one ad set each for commute, gym, gifts, outdoor, moms, students, coffee and eco interests.
Structure call
Combine groups without a business difference while keeping the product page, Purchase event, offer and existing creative fixed. Plan a separate creative round after reviewing the structure.

Build a campaign draft for the tumbler

Start with one Sales campaign, one US new-customer ad set and three ads. A past-buyer group is a conditional extension. A beginner can finish the draft checks before an authorized person decides to publish. In an existing account, preserve current settings and screenshots rather than deleting while rebuilding.

This draft is for US website purchases, using Sales, the website conversion location and Purchase events. Verify reporting, deliverable page promises and affordable loss first. The objectives lesson explains the choices in more detail, but the task for this round is already defined here.

A new video can restart learning for its ad set

A significant edit sends an ad set back into learning. Adding an ad is on the list, so keeping creative at ad level does not make it harmless to add videos at any time.

Meta's significant-edit list includes changes to targeting, creative, the optimization event or bid strategy; adding an ad; and resuming an ad set paused for 7 days or longer.

Which edits depend on size, and which leave other ad sets alone?

Changes to budget, ad set spend limits, bid controls, cost-per-result goals or ROAS-goal amounts depend on magnitude. Meta's example says $100 to $101 is unlikely to restart learning, while $100 to $1,000 may. This does not establish a universal safe increase.

Campaign-budget redistribution does not itself restart learning. An edit made at one ad set's level does not restart sibling groups, nor does adding a group to a campaign using campaign budget. Changing the campaign budget amount or bid strategy may affect several groups.

On day 3, suppose the tumbler store's team wants two more videos, twice the budget and another interest. Write down the reason for each change and choose this round's action. Adding ads and changing targeting are already significant edits; several moving conditions make later results hard to interpret.

This lesson recommends saving nonurgent changes for the planned review and acting on one clearly defined problem per round. Schedule the first review around one week after the last significant edit. Day 7 does not automatically permit an edit or higher spend: inspect status, spend and orders first. Last significant edit and Activity history show the timing, change and operator.

Playable chart

Creative lives at ad level, so on day 3 can I add two videos, double the budget and switch the interest?

As text: the whole walk-through
  1. Day 3 Three days in, some results have arrived; the team wants two more videos, double the budget and another interest.
  2. Videos added Adding an ad is on the significant-edit list: the whole ad set goes back to learning and both readings restart.
  3. Interest switched Changing targeting is on the list too: a second edit the same day, and later you cannot tell which one did what.
  4. Hands off Stop touching it. Counting restarts from the last edit; the earlier progress does not carry over.
  5. Day 7 review Day 7 is the review point: check status, spend and orders. Results are short of about 50, so it is still learning.
  6. Day 7 review Since the last significant edit, results have reached about 50.
  7. Stable Only with enough results does the status leave learning; day 7 is a day to review, not a pass out of learning.

Illustrative: day 3, day 7 and "about 50" come from the lesson ("usually", "about"); results in between are assumed.

Watch three things: how the two readings, days and results, drop back to the start after each significant edit; why day 7 is a review point while the status still says learning; and what a day-3 pile-up costs compared with one edit at the review point.

Give each layer a recognizable name before publishing

Names appear in Ads Manager, exports and website source labels. Use stable facts such as market, objective, product, customer task, variable and version. Leave customer details and people's names out.

This lesson uses short uppercase codes joined with hyphens, campaign prefixes for ad sets, and creative identifiers for ads. Add a date only when useful, at the end as year and month. This is a team convention rather than a Meta naming requirement.

US-SALES-20OZ-NEW-BROAD

  1. USMarket: United States
  2. SALESObjective: Sales, from the campaign
  3. 20OZProduct: 20oz tumbler
  4. NEWCustomer stage: new customers
  5. BROADAudience approach: broad

Classroom naming example: the new-customer ad set

A past-buyer group could be US-SALES-20OZ-EXIST-CAP and videos HOOK24H-VID-V1, LEAKPROOF-VID-V1 and GIFT-VID-V1. Names identify message angles, not established effects. Claims such as 24-hour cooling or leak resistance require actual product evidence before advertising. Use CA when Canada is ready, and check its promises again.

Set names before publishing. Name-based dynamic URL parameters retain the initial published names rather than following later renames. Changing those name values requires a new corresponding campaign, ad set or ad. A team with an existing template can switch on Template in the name field; create the template in advertising settings first.

Build and check in six steps

The paths below come from Meta's help pages. Positions and wording can change; use the actual account screen. A new draft can contain all three ads from the start. When changing existing delivery, perform the chosen structure action while preserving other conditions.

Build the tumbler campaign in Ads Manager

  1. Write the structure sketch

    Write down US website purchases, campaign budget, one new-customer ad set excluding defined past buyers, three videos changing only the opening message, and the first review date. Preserve current settings and screenshots so they can be restored. In an existing account, keep the original structure while preparing the draft.

  2. Create the campaign and pick the budget strategy

    Where in the UI: Ads ManagerCreateSalesBudgetBudget strategy

    Choose the Sales objective. Under Budget, set Budget strategy to campaign budget with a daily budget of $80 (example data). Type a name using the examples above, or switch on Template if your team has already created one.

    How to confirm

    The Budget section shows Advantage+ on.
  3. Create the new-customer ad set

    Where in the UI: Ad setAudienceControlsShow more controls

    Set the location to the United States, select a clearly defined past-buyer custom audience for exclusion, and keep Advantage+ placements. Choose website as the conversion location, the verified event data source, a conversion performance goal and Purchase. Record the actual attribution model and settings. An incomplete buyer list means you cannot call the results purely new-customer results.

    Caution

    Don't rebuild the interests as a stack of separate ad sets.
  4. Optional: create the existing-customer ad set

    Where in the UI: Ad setBudget and scheduleAd set spending limits

    Only do this when control of past-buyer spend is useful and audience size and budget support it. If duplicating the new-customer ad set, remove its exclusion of the same past-buyer audience first. Under Audience, select Further limit the reach of your ads, then Switch setup, include the past-buyer custom audience and uncheck Use as a suggestion. Set a daily or lifetime maximum under Ad set spending limits. Use the same creative for this new-versus-existing-customer task.

    How to confirm

    This is Meta's option for controlling past-buyer spending, not a requirement to run two ad sets. Visitors or cart abandoners do not automatically belong in it. Check the customer split against audience definitions and real orders.
  5. Create three ads with URL parameters

    Where in the UI: AdTrackingURL parameters

    Keep the product page and offer the same; change only the opening hook. Check the Page, Instagram identity, destination and tracking. An example is utm_campaign={{campaign.name}}&utm_term={{adset.name}}&utm_content={{ad.name}}. Using term for the ad set is this lesson's convention. Test whether the labels appear in the relevant Shopify or Google Analytics 4 (GA4, website analytics) reports; not every order will retain them.

    Label the hook, proof, offer, format and landing-page version in the ad or creative record. Proof can be a demonstration, review or product test; an offer can be a discount, free shipping or a bundle. Match all of them to the actual product and page. A new draft can contain three ads from the start. When changing existing delivery, perform the chosen structure change first rather than replacing creative and pages at the same time.

  6. Preview, then decide on publishing and review

    Preview the ad and open its destination to check parameters, product, price and shipping promises. Verify budget, schedule, events, exclusions and payment authority; leave errors in draft. After an authorized person confirms publication, inspect all three delivery statuses, note the last significant edit and use roughly one week for the first planned review. Faulty events, unaffordable spending or fulfillment failures need prompt attention rather than waiting for that date.

    Keep as evidence

    Take a dated screenshot of Activity history.

Caution Handle failures before the scheduled review

A broken landing page, faulty purchase events, fulfillment failure or spending beyond an affordable limit needs prompt attention or a pause in affected delivery. An observation window is not a reason to keep spending. Restoring old settings can also restart learning; inspect the original screenshots and current state first.

Read the first week's results in context

Start with the whole round, then inspect each layer

Ads Manager has Campaigns, Ad sets and Ads tabs. With campaign budget, start with the whole campaign in the same window: total spend, purchases and average purchase cost. Then check the ad set's Delivery and Last significant edit. Status can differ across all three levels; inspecting one is not sufficient.

Schematic; your account's screen may differ

Five places to locate when reading results by layer.

  1. The three tabs: switch between campaigns, ad sets and ads.
  2. Delivery: check Learning, Learning limited or Active at each layer.
  3. Last significant edit: inspect results from this point rather than mixing them with the previous setup.
  4. Breakdown: inspect supported dimensions such as age or country without creating an ad set for each report row.
  5. Activity history: who changed what, and when.

Delivery, Results, Cost per result and Last significant edit are common columns. Add Optimization events through the column settings. Breakdown supports dimensions such as age, gender and country; Activity history shows who changed what and when.

Exercise · pick one

What to read first at each layer

Read first
Total results, average cost per result and the week's total spend.
Don't conclude
With campaign budget, an ad set that spends less isn't necessarily worse; uneven spend is how it works.
The store's first review
Summarize spend and purchases in the same window. Without comparable history, establish a baseline instead of declaring the structure a winner.

Some numbers cannot be added or directly compared

Unique metrics such as Reach do not add up across levels. If group A reaches Ann and Ben and group B reaches Ben and Cal, the campaign reaches three people, not four.

Results using different attribution models are not directly comparable. Ads within one group also receive nonrandom exposure; comparisons across groups add audience and budget differences. Align dates, market, event and time zone before interpreting results.

Connect the result to real store orders

In Shopify or Google Analytics 4 (GA4), follow retained UTM labels to traffic and orders, then combine product costs, fulfillment, stock and refunds to assess margin. Marketing-source reports do not automatically calculate all margin or prove incremental sales.

Meta and website reports can use different attribution methods and reporting delays. Do not force the order counts to match or invent an ad source for an order without the required labels.

With no comparable history, the tumbler store's first week establishes a baseline. Record what changed, what stayed fixed and which evidence is missing. The modeled 28 purchases are not actual performance and cannot establish a permanently winning structure.

Four first-review results and the next check

Explain a change before changing the structure again. Return on ad spend (ROAS) is revenue attributed to advertising divided by ad spend; it is neither profit nor, by itself, a measure of incremental sales. These four situations are possible review findings, not measured results from this store.

Check the evidence before choosing the next action

What you seeCheck firstNext action
One day's ROAS falls after consolidationAlign the same post-edit window and inspect actual spend, Purchase events and reporting delays.Do not split back into eight groups on one day's movement alone. Keep the agreed review window, while acting promptly on faulty events or unaffordable loss.
The single cup has a lower purchase cost; the bundle has fewer ordersCompare margin after costs, refund reasons, days of stock cover and shipping promises for the two products.Establish which orders matter more to the business before reallocating budget. A cheaper sale is not necessarily more profitable.
Separate prospecting ROAS is below the old blended figureAlign dates, Purchase events and attribution, then inspect linkable new-customer orders. The blend may have hidden customer-group differences.Establish a comparable prospecting baseline. This reporting difference alone does not show that restructuring harmed acquisition.
One video's opening appears to leadKeep the asset version and compare its opening, product-proof footage, offer, format and page version. Identify what actually changed this round.Treat it as a candidate for the next controlled creative test. Normal delivery is not randomized; an apparent lead is no immediate reason for another ad set.

Name who will review which data. For example: this round combines interest groups serving the same task into one US prospecting group, keeping budget, product page, offer and Purchase events fixed. The ads and store reviewers will check actual spend, Purchase events, linkable orders, product economics and new-customer share in the same window at the first review. Screenshots and the cost summary stay in the restricted team record. Continue observing when evidence is insufficient; pause affected delivery promptly on faulty purchase reporting or unaffordable loss.

Keep the decision for your next review

Record the current decision, fixed conditions and next check. Keep descriptions short and nonsensitive; saved values stay in this browser. Do not enter credentials, customer details or order records. Make the next move concrete: combine duplicate interest groups, pause for faulty Purchase reporting, and add no new videos before review.

Note

Finish the 8 items above and the conclusion appears here.

A filled record does not approve the setup or establish profit. Compare it with actual status, spend and orders at review. Continue observing if evidence is insufficient; handle event, fulfillment or affordable-loss failures promptly.

What you finished

No checkpoint finished in this lesson yet. Start with “Go to the exercise”.

Self-check not answered

Questions worth clearing up

What do campaigns, ad sets and ads each control?

The campaign sets the objective and budget strategy. The ad set sets the audience, placements, schedule, performance goal and attribution setting; if you choose ad set budget, its amount is set there too. The ad is the format, creative, text and link people see. Delivery learns per ad set, so merge-or-split questions are mostly about ad sets.

Should I use campaign budget or ad set budget (CBO or ABO)?

Consider campaign budget when the ad sets matter equally and you judge the campaign as a whole. It's the default for Sales campaigns, and its results are read at the campaign level. Consider ad set budget when performance goals or bid strategies differ, audience sizes differ a lot, or a market needs separate budget control. Setting a budget does not guarantee full spend or profit.

How many conversions does an ad set need to exit the learning phase?

Meta says ad sets usually exit after about 50 results in the week after the ad set's last significant edit. It's counted per ad set, and when an ad set is unlikely to get there, its delivery status shows Learning limited.

Which edits restart the learning phase?

Changes to targeting, creative, the optimization event or the bid strategy, adding a new ad, and resuming an ad set after it has been paused for 7 days or longer are significant edits. Changes to budget, spending limits and bid or cost goals depend on size: small adjustments usually don't restart learning, large ones may.

Are more ad sets better? When should I consolidate?

No. Dividing the same budget among more similar ad sets makes it harder for each to accumulate enough results; actual allocation and results are not guaranteed to be equal. Meta recommends combining ad sets that differ only by interest, creative format, placement or language, and using Breakdown to see the segments.

When is a split actually justified?

Consider a split for different market promises, product economics or fulfillment conditions, or a customer task that needs separate control. Check funding, samples, exclusions and review capacity first. About 50 is a learning-status reference, not a universal permission threshold for every business split.

With Advantage+ sales campaigns (formerly ASC), do I still need a structure?

Yes. An Advantage+ sales campaign is a Sales campaign with Advantage+ audience, placements and campaign budget switched on. The three layers remain, and splitting, naming and where to read results are still your decisions. The existing customer budget cap is gone, so to limit spend on past buyers you use two ad sets with a spend cap.

How should I name Meta campaigns, ad sets and ads?

Use stable, consistent short codes. An ad set can carry the campaign prefix, while the ad name identifies the message and version. Set names before publishing: name-based dynamic URL parameters retain the original published names rather than following later renames.

Choose the next decision

Use the remaining question in your draft to choose what to read next: the objective and event, the creative comparison, or the budget and observation window.

Research and cases

  1. Tinuiti, Digital Ads Benchmark Report Q2 2026. Facebook CPM figures appear in the ROAS by Channel table of its ROAS article (Aug 19, 2026). A directional client-sample observation.
  2. Seer Interactive, Meta Ad Account Consolidation (Sep 19, 2023), Results table. A single-client before/after case with several simultaneous changes.
  3. Gordon et al., A Comparison of Approaches to Advertising Measurement, Marketing Science 38(2), 2019. Peer-reviewed; this lesson uses only a qualitative finding supported by the abstract.

Edition notes

  • Author Ranfeng Wei
  • Published
  • Updated
  • Reviewed
  • Next review

This update: Retained the tumbler-store case and practice, clarified CBO/ABO and audience exclusions, and kept the Ads Manager reading map. Added a Tinuiti client benchmark, a Seer case and advertising research with compact numbered references; colored bars now follow the calculator inputs.