Ad Budget and ROAS Tracking

You keep paid media honest: every week the owner knows what was spent, what it really returned according to the business's own sales data, whether spend is on pace, and what to change. Platform dashboards each claim credit for the same sales, so you reconcile them with the store or CRM and judge by blended numbers as well as platform numbers. You recommend budget moves; the owner approves them.

When to use

• "How are the ads doing this week?"
• "Are our ads actually profitable?"
• "Where should we move budget?"
• "ROAS dropped. Why?"
• "Set up a weekly ads report."
• Month end, when spend must be reconciled with the budget.

What you need

• Spend and results by platform, campaign and week: from the ad platforms through the owner's signed-in browser (Google Ads, Meta Ads Manager, Microsoft Advertising, LinkedIn Campaign Manager, TikTok Ads) or exported CSVs.
• The business's own outcomes: orders and revenue (Stripe, the store backend, QuickBooks or Xero through connected apps), or leads and deals (HubSpot, Salesforce, Pipedrive). Google Analytics for sessions and conversions by source.
• Unit economics: gross margin, fulfilment costs, refund rate, customer value or subscription revenue per month.
• The approved budget per platform and month, and the target CPA or ROAS.
• A tracking sheet (Google Sheets connected app, or a CSV you maintain with write_file).

Method

1. Fix definitions once. Write them at the top of the tracker (full list in references/tracker-template.md): spend; platform-reported conversions and revenue with each platform's attribution setting; CPA = spend / conversions; ROAS = revenue / spend; MER (marketing efficiency ratio) = total revenue from the business's own data / total paid spend; break-even ROAS = 1 / contribution margin rate; new-customer CPA from the business's own data.
2. Pull the week's data. Same days and time zone for every source. Record each platform's attribution window (for example Meta's default has been 7-day click and 1-day view; Google Ads uses its own model); results under different windows are not comparable.
3. Reconcile. Sum platform-claimed conversions and compare with real orders or qualified leads in the same period. A ratio above 1 shows double counting. Track the ratio weekly; a sudden change usually means tracking broke or attribution settings changed.
4. Calculate profitability. Compare each campaign's CPA with the target and break-even; compare blended MER with break-even ROAS. For subscriptions, compute payback months per plan or customer segment: CAC / (monthly revenue per customer x gross margin). Blended averages can hide a cheap plan whose customers never pay back.
5. Check pacing. Spend to date / budget for the month versus days elapsed / days in month. Flag over 110 percent or under 90 percent of pace. Project month-end spend at the current daily rate.
6. Flag anomalies. Week-on-week changes beyond normal variation (for example CPA up more than 30 percent, CTR down more than 25 percent, spend up with flat conversions), zero-conversion days on campaigns that normally convert, frequency rising on a small audience (creative fatigue), and conversion rate drops on the landing page. Look at the cause before recommending: tracking, budget change, creative fatigue, seasonality, competition, site change.
7. Separate brand from non-brand and new from returning. Brand search and retargeting look efficient because many of those buyers were coming anyway. Judge growth spending on non-brand prospecting and on new-customer CPA.
8. Test incrementality when budgets justify it. Platform conversion lift studies where available and eligible, or a simple geographic or time holdout (pause a campaign in comparable regions and compare sales). Plan these with the owner; they cost results while running.
9. Recommend. Per campaign: scale (CPA comfortably below target with stable volume: raise budget in modest steps), hold (near target or still learning), fix (good volume, worsening efficiency: refresh creative, check landing page), or cut (well above break-even after a fair test). Give the expected effect in money.
10. Report and wait for approval. Weekly report on a show_card plus the updated tracker. Budget changes, pausing or launching campaigns are made only after the owner's explicit go-ahead. Set a cronjob to produce the report every Monday and a pacing check mid-week.

Output

Weekly report (template in references/tracker-template.md):

• Headline: spend, new customers or qualified leads, blended MER or CPA against target, one sentence on what changed.
• Table by platform and campaign: spend, results, CPA, ROAS (platform), share of real outcomes, status (scale / hold / fix / cut).
• Pacing: month to date, projected month end, versus budget.
• Reconciliation ratio and what it implies.
• Anomalies with likely causes.
• Recommendations with expected effect, waiting for approval.
• The updated tracker sheet.

Checks before you finish

• All sources cover the same dates and time zone.
• Revenue and orders come from the business's own system, not only platform claims.
• Brand and non-brand reported separately.
• Break-even ROAS and target CPA are based on current margins.
• Every recommendation names the campaign, the action, the amount and the reason.
• No budgets or campaigns were changed without approval.

Pitfalls

• Adding up platform ROAS. Each platform counts the same sale. Use blended MER and reconciliation.
• Judging on revenue, not margin. A 3x ROAS loses money if the contribution margin is 25 percent (break-even ROAS is 4).
• Reacting to one day. Daily numbers are noisy. Judge on weeks unless something is broken.
• Ignoring refunds and cancellations. Subtract them, or results look better than they are.
• Cutting what is still learning. New campaigns need enough events before judgement.
• Big budget jumps. Large changes can reset learning and spike costs. Step changes up gradually.
• Moving money without a go-ahead. You recommend; the owner decides.