Google Ads

Pre-Post Analysis

Also called before-and-after analysis, pre/post read

A before-and-after comparison with no control group, easily confused by seasonality and anything else that changed.

Quick facts: Pre-Post Analysis

Category
Google Ads
Also called
before-and-after analysis, pre/post read
Level
Intermediate
Affects
Reporting accuracy, budget decisions, channel credit
Where to see it
GA4 date comparisons, Google Ads reports, your own sales records
In this article4
  1. How pre-post analysis works
  2. Why the weakness matters
  3. Where pre-post analysis goes wrong
  4. How to use it responsibly

How pre-post analysis works

Pre-post analysis compares a period before a change with the period after it. Spend went up in one month and sales went up in the next, so the extra spend worked. It needs no special setup, it is the most natural way to look at a result, and it is the method most marketing reports quietly rely on.

The structure has one thing going for it and one thing badly against it. In its favour, the comparison is against your own business rather than against an assumed benchmark. Against it, there is no control: nothing in the design separates the effect of your change from everything else that moved at the same time.

Why the weakness matters

Almost every business has a season. Demand for trekking, education consultancy, remittance and retail all move with the calendar, so a comparison across two different months is partly a comparison of two different demand levels. Add competitor activity, a price change, a public holiday, a news story or a platform update, and the later period is not the earlier one with a single variable altered — it is a different set of conditions.

The consequence is expensive rather than academic. Budgets get scaled on effects that were really seasonal, and channels get cut for a dip they did not cause. A weak method used confidently is worse than no measurement at all, because it produces decisions.

Where pre-post analysis goes wrong

Two failures show up repeatedly. The first is choosing the comparison period after the fact — sliding the start date around until the story improves. If the window is not fixed before the change goes live, the analysis is a search for a favourable result rather than a test of anything.

The second is stacking changes. Bids, creative, landing page and budget all move in the same week, then the following month is credited to whichever one the team likes best. A before-and-after read can only ever tell you that the whole picture changed. It cannot say which part of it did the work.

How to use it responsibly

Use it where a real experiment is impossible, and be plain about the limits when you report it. Compare like periods rather than adjacent ones, so the same season sits on both sides. Fix the window before the change goes live. Change one thing at a time. Look at parts of the business your change did not touch over the identical period as a rough control — organic traffic, a region you left alone, a product line you did not promote.

When a decision is big enough to matter, replace the method rather than dressing it up. A geo experiment or a holdout test costs more effort and returns an answer you can defend in front of whoever is paying. Pre-post is a reasonable first look and a poor basis for spending more money.

Do and do not

Do

  • Fix the comparison window before the change goes live
  • Compare like seasons rather than adjacent months
  • Check untouched parts of the business as a control

Do not

  • Shift the date range until the result looks better
  • Change several things in the same week
  • Raise budgets on a before-and-after read alone

Questions people ask about this

Is pre-post analysis ever acceptable?

Yes, as a first look and as a description of what happened, provided nobody claims it proves cause. It is also reasonable when a change cannot be split at all — a site migration, a rebrand, a pricing change. In those cases state the limitation openly and support the read with anything that stayed untouched over the same period.

How do I reduce the seasonality problem?

Compare the same period in a previous year rather than the month just gone, so the season appears on both sides of the comparison. Then look at parts of the business your change did not touch over the identical window. If they moved in a similar shape, the market moved, and your campaign is not the explanation you were about to write down.

Why does my platform report disagree with a before-and-after read?

Because they count different things. A platform credits conversions it can attribute to its own clicks and views inside its window, while a before-and-after read sees total business outcomes including everything no channel is tracking. Neither is wrong; they answer different questions. Disagreement between them is normal, and it is a good reason to run a proper test.

Related terms

Found this useful?

Share it, or ask an AI to summarise it

Back to the glossary

Knowing the term is the easy part

Applying it to your own site and budget is the work. Book a call and I will tell you what actually applies to you.