Did my negative keywords work? A fair before-and-after check

The three numbers

• Spend should fall a little: you stopped paying for searches you didn't want.
• Conversions should hold. If they fall a lot, a negative probably blocked searches that were selling.
• Cost per conversion is the verdict: spend divided by conversions.

A fair comparison

• Same length on both sides. 14 days before against 14 days after, not "last month against this week".
• Leave out the day of the change: it is half one and half the other.
• Full days only, up to yesterday.
• Same days of the week on both sides. 14 or 28 days does that for you.

Traps

• Other changes in the same window. A new budget, a new ad or a bid strategy change mixes its effect into yours. Write down what else changed.
• Seasons and sales. A holiday on one side and not the other beats any negative keyword.
• Too few conversions. With 3 conversions on each side, one more or one less swings cost per conversion by a third. Wait longer, or judge on spend saved instead.
• Conversion delay. Recent clicks may still convert in the coming days. Give the "after" window a few extra days before you judge.

Doing it in one question

In Google Ads you can set two date ranges with the compare option on the Campaigns page. From an AI assistant, Ads MCP has a tool for exactly this (did_it_work): tell it the day of the change and the number of days, and it returns the three numbers on both sides. It counts the day of the change in the "after" window, so with very few conversions keep that day in mind.