Most week-over-week (or month-over-month) performance reviews start the same way: someone notices a number moved, and asks “why?” The honest answer is usually “here’s what the math says moved, and here are a few things worth checking” — not a single confirmed cause. This is a walkthrough for getting to that honest answer efficiently.

Start with the comparison, not the explanation

Before reaching for an explanation, get the comparison right:

  1. Use the same metric definitions in both periods. Make sure attribution windows, conversion definitions, and currency are consistent. A platform-side change to any of these can move your numbers without any real change in performance.
  2. Compare like periods where possible. A Monday-heavy week against a weekend-heavy week, or a 7-day period against an 8-day period, can create differences that have nothing to do with campaign performance.
  3. Write down both periods’ raw numbers — spend, impressions, clicks, conversions, revenue — not just the rates. Rates alone (CTR, CPA, ROAS) hide whether a change came from volume or efficiency.

Work through the chain, not just the headline number

A single metric like ROAS or CPA is a summary of several things happening underneath it. Rather than reacting to the headline number, break it down:

  • Did spend change, and if so, was that from more clicks (volume) or a higher cost per click (efficiency)?
  • Did conversions change, and if so, was that from more clicks or a better/worse conversion rate?
  • Did revenue per conversion change — average order value, deal size, or product mix — separately from conversion volume?

This is exactly the decomposition described in Why CPA Increases Even When Conversions Increase and How to Diagnose a Sudden CPC Increase. The relationships are exact — percentage changes in a ratio combine multiplicatively, not by subtracting percentage points — so doing this by hand is easy to get subtly wrong.

Use a structured diagnostic instead of guessing

This is exactly what our Marketing Performance Diagnostic is built to automate: enter whatever metrics you have for both periods (it derives the rest wherever the math allows), and it will:

  • Show the exact percentage change for every metric you can compare.
  • Identify which metrics moved the most.
  • Break down ratio metrics like CPA and ROAS into their exact mathematical drivers.
  • Suggest hedged, plausible explanations relevant to what actually moved — not a generic list.
  • List recommended investigations and the metrics that would sharpen the analysis further.

It won’t tell you with certainty what happened in your account — no tool working only from aggregate numbers can — but it will keep you from mis-explaining a change that’s actually just arithmetic (like mistaking a ratio effect for a dramatic swing).

Don’t skip the limitations

Two things are easy to forget mid-analysis:

  • Conversion lag. If the most recent days in your “current” period are very recent, some conversions may not have been recorded yet. A conversion rate or CPA that looks worse in the last few days of a period is sometimes just incomplete data.
  • Small-sample noise. A short period with modest volume can show large percentage swings that are mostly statistical noise, not a real shift. Before investigating a dramatic single-week change, check whether it holds up over a slightly longer window.

A reasonable process

  1. Pull consistent, like-for-like numbers for both periods.
  2. Run them through the Performance Diagnostic (or work through the chain manually using the calculators above).
  3. Note which metrics moved the most and what the decomposition says about why.
  4. Treat the suggested causes as a checklist to investigate in your actual account — not a conclusion.
  5. Confirm before you act — especially before changing budgets or bids based on a single period’s numbers.