Measurement
Pre-Post Analysis
Pre-post analysis compares a metric in a period before a change with the same metric in a period after it, and reads the difference as the change’s effect. It is the simplest way to measure an SEO change and the easiest to get wrong, because seasons, ranking updates and other work can move the metric too.
- Industry term
By QueryRev ResearchPublished Updated5 min read
Why it matters.
Most SEO results are reported this way: clicks in the four weeks after a change against the four weeks before. It is quick and needs no comparison group, which is why it is everywhere.
It assumes that nothing else changed. Google’s guide to traffic drops lists what else can move search traffic in the same weeks: algorithmic updates, technical issues, security and spam issues, seasonality and changing interests, and site moves[1]. Anything of that kind inside the window is counted as the change’s effect.
Statisticians treat the plain comparison as a starting point. The CausalImpact method was set against classical difference-in-differences, and models local trends, seasonality and the influence of outside series over time[2], the movements a plain before-and-after leaves inside its difference.
It still has its place. A large change on a page with steady demand, read against the same weeks a year earlier, gives a fair first reading; what it cannot do alone is tell the change apart from everything else that happened in the same weeks.
The same care applies to any two readings. QueryRev’s study is a dated reading: its October 2026 edition read robots.txt files on 15 October 2026, from a list of the top 10,000 domains made on 14 October 2026[3]. A change between two editions can come from the list as well as from the sites, and is read with that in mind.
How it is measured.
Choose the windows before the change: equal lengths, the same days of the week, long enough to hold ordinary swings. Compare each with the same weeks a year earlier, so that seasons cancel out; Search Console’s Performance report can compare a period year over year over the last 16 months[1].
Add pages that were not changed. The changed pages’ before-and-after minus the unchanged pages’ before-and-after removes much of what moved both, such as a season or an update; this is the logic of difference-in-differences.
Keep the pre-period clean. The CausalImpact documentation uses the period before the change to learn how the measured series moves with its controls, and assumes that relationship holds after it[4].
Write the result as a range, from the least to the most of the difference the comparisons allow, rather than as the raw difference.
A forecast made before the change is another way to say what the after-period should have looked like. How to build one from your own last year, seasons included, is set out in SEO forecasting: how to forecast traffic and revenue.
Measured by
- Before and after windows
- Same weeks last year
- Unchanged pages
A worked example.
Sample data: an invented site, to show the method.
A travel site rewrites its guide to winter breaks in Lisbon on 1 November. In the four weeks after, the guide earns 6,300 clicks, against 4,500 in the four weeks before: a pre-post difference of 1,800 clicks.
In the same four-week windows a year earlier, with no change to the page, its clicks went from 4,200 to 5,600, a seasonal rise of a third, and ten similar city guides that were not touched rose by about the same share this year. Allowing for the season, the rewrite added about 300 clicks, not 1,800.
What to do, step by step.
Write down the change, the pages and the date before the after-period starts.
Choose equal before and after windows, aligned on the same days of the week.
Compare both windows with the same weeks a year earlier.
Add similar pages that were not changed, and compare their before-and-after with that of the changed pages.
Check the update calendar, the release log and other campaigns for anything else inside the window.
Report the result as a range, with the windows and the comparison named.
Mistakes to avoid.
Comparing next-door months in a seasonal business. The month after a change made in November is December. Use the same weeks a year earlier.
Windows of different lengths or weekdays. Four weeks holding five Saturdays against four weeks holding four are not comparable for a shop. Align the days.
Ignoring other changes. A ranking update, a sale or a new campaign in the same weeks is counted as the change’s effect unless it is ruled out.
Reading one page alone. One page’s clicks swing from week to week on their own. A group of changed pages compared with a group left alone says more than any single page.
Moving the windows after the result. Windows chosen once the numbers are visible can be made to show almost any effect. Fix them first.
How QueryRev treats it.
QueryRev does not report a plain before-and-after as an outcome. A change’s window is set before it is applied, the result is compared with unchanged pages of the same kind and with the same weeks a year earlier, and verification reports it as a range.
Questions people ask.
What is the difference between pre-post analysis and an A/B test?
A pre-post analysis compares the same pages at two times. An A/B test compares two groups at the same time, one changed and one not, so a season or an update moves both groups and drops out of the difference.
How long should the before and after periods be?
Equal, aligned on the same days of the week, and long enough for ordinary swings to even out. For most page changes that means several weeks each, chosen before the change is made.
Is a pre-post analysis ever enough?
For a large change on a site with steady demand and nothing else happening, it can be a fair first reading. With seasons, updates or other work in the window, it needs a comparison with last year or with unchanged pages.
Sources and citation.
Primary sources only: the search engines, the AI vendors, standards bodies and QueryRev’s own research. Each is cited in the text by its number.
Debugging drops in Google Search trafficGoogle Search Central ·
State of AI crawler access, October 2026QueryRev Research ·
CausalImpact: an R package for causal inference using Bayesian structural time-series modelsGoogle ·
Cite this definition
A suggested citation. Its link goes to the definition, whose anchor does not change.
QueryRev Research. “Pre-Post Analysis.” QueryRev AI glossary, 19 October 2026. https://queryrev.com/glossary/pre-post-analysis#definition
Related terms