Read your results without over-reacting

Daily numbers are noisy

An unchanged campaign can show a cost per purchase that swings widely from one day to the next. A single bad day is rarely a signal. Judge on at least 3 to 7 days of data, and compare like with like: this week against last week, not Monday against Saturday.

Recent data is also incomplete. Conversions are often credited back to the day of the click, so yesterday's numbers tend to improve over the next few days.

Why the numbers never match

Meta, Google, Google Analytics and your store each count conversions differently. Each platform uses its own attribution window (how long after a click or view a sale still counts) and each tends to claim credit for sales the others also claim.

Add them up and you will usually get more sales than you actually made. A 3x ROAS in a platform dashboard might be break-even in real life.

1. 1

   Pick one source of truth

   Your store, CRM or bank account decides what really happened.

2. 2

   Measure blended results

   Total revenue divided by total ad spend across every platform, week by week.

3. 3

   Use platform numbers for comparisons inside the platform

   They are good for choosing between two ads, less good for judging total profit.

4. 4

   Keep windows consistent

   Don't change attribution settings mid-comparison.

Three problems worth spotting

• Wasted spend: ad sets, keywords or placements that spend steadily with no results.
• Fatigue: frequency climbing week by week while click-through rate falls.
• Timing: results clustered in some hours of the day while spend runs around the clock. Look at at least 30 days of hourly data before cutting any hours.

Adwise surfaces these in its analytics (spend without results, hour-by-day heatmaps), but the same checks work by hand in any ads manager.