Media buying · Agency selection

What should your paid media agency report each week?

By Mirow · September 27, 2026 · 7 min read

A useful weekly paid media report should let you decide what to do with the budget, see the work delivered, and identify the next test or blocker. Ask for comparable performance trends, a creative and campaign work log, and a short decision list with owners. Each recommendation should point to evidence you can inspect.

Before hiring or renewing an agency, ask it to complete that report using a fictional example. You can evaluate how it thinks without seeing another client's account. The test is whether you can follow a recommendation from the reported result through the explanation to the next action.

Start with a decision and define the reporting window

The opening paragraph should state the proposed action, the evidence behind it, and what remains uncertain. “Hold the prospecting budget while the current purchase cohort matures; review the latest creative test on Thursday” gives the reader a decision to discuss. “Performance remains strong” leaves the decision unstated.

Above the numbers, require the account, channel, currency, account time zone, reporting dates, extraction time, and attribution settings. Use the same completed weekday range for comparisons. Identify partial weeks and changes to purchase events or attribution settings.

Recent results can change after the reporting week closes. Google Ads documents that conversion delay can make recent cost per conversion look higher because spend is already reported while conversions are still arriving. Show when the report was extracted and which periods remain provisional; preserve the earlier snapshot when figures are revised. Google Ads: conversion delay.

Show the trend and name what the numbers measure

For ecommerce media buying, a compact trend can include spend, platform-attributed purchases, attributed purchase value, cost per attributed purchase, and attributed return on ad spend. Use consistent definitions and include a simple chart where it helps expose the pattern.

The following four-week example is entirely hypothetical, in USD. Every number is invented to demonstrate a report format. It represents one ad platform using unchanged purchase-event and attribution settings. All weeks are complete delivery weeks, but the newest week's conversions may still arrive. These are platform-attributed outcomes, not a store order ledger or a measure of incremental sales.

Scroll horizontally to see all columns.

Cost per attributed purchase is media spend divided by attributed purchases. Attributed ROAS is attributed purchase value divided by media spend. Neither calculation establishes new-customer acquisition cost or profit.

At this extraction, Week 4's cost per attributed purchase is 20% above Week 3's, while spend is unchanged. That earns a diagnostic and a review date. It does not establish creative fatigue or justify automatically replacing every ad. The agency should check conversion maturity, purchase-event changes, delivery mix, and the actual creative evidence before assigning a cause. Our ad-fatigue guide separates those decisions.

An attribution model assigns conversion credit among ad interactions. The assigned value is a measurement lens; it does not demonstrate that every credited order was caused by the advertising. If store results are included, label them separately and reconcile their dates and definitions. Do not add overlapping platform purchase totals together and present the sum as unique store orders. Google Ads: attribution models.

Give creative production and testing their own evidence

A performance chart cannot tell you whether the agency delivered the promised creative work. Ask for a compact log linking each concept to its variants, approval, first delivery, spend, and current decision. A concept is the underlying message or hypothesis; a variant is one execution of it. These are useful operating definitions to agree with the agency, not universal platform labels. For each campaign change, record the date, budget or setting changed, reason, expected effect, and next review date. That makes the account work inspectable alongside the creative assets.

In a fictional production week, two concepts might produce six variants. Four variants are approved, three of those receive their first impressions, and one approved variant is still waiting to launch. The other two await approval. Report each stage separately. “Six new ads” conceals how much work shoppers actually saw.

The test log needs a result status as well as an asset count:

Scroll horizontally to see all columns.

Ordinary delivery can give different ads different audiences and amounts of spend. A cheaper purchase cost on one ad is useful evidence to investigate, but it is not automatically a controlled comparison. TikTok's split-testing documentation, for example, describes equal, mutually exclusive audience groups while other variables are held constant. Its winner selection requires a statistically significant result. If an agency calls an ordinary campaign breakdown an A/B test, ask how allocation was controlled. TikTok: split testing.

“Inconclusive” can be an accurate result. Judge whether the report explains the uncertainty and the next decision; a weekly deadline does not make an unfinished experiment conclusive.

Keep commercial conclusions inside the available cost evidence

When actual costs are complete and reconciled for a consistent period or cohort, show contribution after media in dollars and as a percentage of net sales. Define the calculation: net sales after discounts and refunds, less the specified variable costs and full media spend. State the treatment of shipping, returns, fees, and tax, and identify any production costs, agency fees, or overhead left outside the measure. Contribution after media is not net profit. Our break-even ROAS calculation explains the cost definitions behind a media target.

If that cost coverage is missing, report contribution and profit as unavailable, with the missing input and its owner. A ROAS multiple cannot fill the gap. Even a store's profit report needs a coverage check: Shopify says its product and variant profit reports include only items with a recorded cost when sold. Shopify: profit reports.

Keep the agency's agreed service scope visible. For a media-and-creative engagement, account performance, campaign changes, creative delivery, and tests are core reporting. Store-wide finance, email, website development, or fulfillment may be context or dependencies owned by someone else. Missing accounting data should prevent an unsupported profit claim; it should not erase the agency's available media and creative report.

End with actions you can approve and revisit

Use an action list that names the evidence, owner, boundary, and next check. This fictional example shows the level of specificity to ask for:

Scroll horizontally to see all columns.

Those dates are agreed operating deadlines, not promises that conversions or test results will mature by then. Record what happened at the next review. If the evidence is still insufficient, make the continuation or stop decision explicit.

For your next agency meeting, select one recommendation and ask to see its supporting numbers, the work behind it, and the person responsible for the next step. Then agree on the one decision the following report must help you make. That gives you a repeatable way to evaluate the relationship as well as the results.