"We lost 30% of our traffic after the update." Statements like that often turn out to be wrong, not because the traffic did not fall, but because the comparison behind them was flawed. Comparing the wrong dates, mixing weekdays with weekends, or ignoring a seasonal peak can make a small dip look like a disaster, or hide a real loss.
This guide is about the measurement itself. You will learn how to set "before" and "after" windows around a core update, how to handle weekday and holiday effects, which metrics to compare, and how to sanity-check your conclusion before it reaches a client report or a board slide.
If you need the Search Console click-path first, start with how to diagnose a core update drop in Search Console; this article goes deeper on the analysis design.
Key Takeaways
- Anchor the "before" window to the rollout start date and the "after" window to at least one week after the completion date.
- Use equal-length windows that contain the same days of the week; whole weeks are the simplest way to do this.
- Check the same windows a year earlier to separate update effects from seasonal demand.
- Compare clicks, impressions and position at page and query level, not just the site total.
- Write down your hypothesis before you look at the data, and look for evidence that would disprove it.
Why the timing rule matters
Core updates roll out over days or weeks. The Search Status Dashboard lists the March 2025 core update as running from 13 to 27 March 2025 and the June 2025 core update from 30 June to 17 July 2025. During a rollout, rankings can move up and down more than once. If you measure halfway through, you are measuring a moving target.
“After a week, try comparing this week with a week before the core update started rolling out”
— Google Search Central, Google Search's core updates and your website
Google's documentation frames the process simply: wait at least a full week after the update completes, then compare that week with a week before the rollout began. The gap between the two windows (the rollout itself plus the waiting week) is deliberately excluded.
Choosing your comparison windows
| Window | How to set it | Why |
|---|---|---|
| Baseline ("before") | The 7, 14 or 28 days ending the day before the rollout start date | Captures normal performance before any update effect |
| Excluded gap | The rollout period plus 7 days after completion | Rankings are still settling; data here is noisy |
| Test ("after") | Same length as the baseline, starting at least 7 days after completion | Reflects the settled post-update state |
| Year-over-year control | The same two windows, shifted back 52 weeks (364 days) | Shows what seasonal change you would expect without any update |
Shifting by 364 days rather than a calendar year keeps weekdays aligned: a Monday last year lines up with a Monday this year. Moving holidays such as Easter will still shift, so note them.
Handling weekday effects
Most sites have a weekly rhythm. B2B sites are often busiest Tuesday to Thursday; recipe and leisure sites may peak at weekends. If your baseline window has five weekdays and two weekend days, your test window must have the same mix. The easiest rule is to always compare whole Monday-to-Sunday weeks. If you need daily detail, compare each weekday with the same weekday, not with the day before.
Handling low volume
For small sites, a single week may contain only a handful of clicks per page. Random variation then swamps any real effect. Use 28-day windows, group pages by folder or topic, and focus on impressions (which have more volume) alongside clicks.
A step-by-step comparison process
- Record the rollout start and completion dates from the Search Status Dashboard, plus any other updates or site changes in the same period.
- Write a one-sentence hypothesis, for example "Our product comparison pages lost visibility in the March 2026 core update."
- Export baseline and test windows from Search Console (Performance report, compare mode) for pages and queries.
- Export the same windows from the previous year as a seasonal control.
- Calculate the change for each page and query in both years. The "update effect" is roughly this year's change minus last year's change over the same dates.
- Separate brand and non-brand queries, and check whether position, impressions or CTR drove the change.
- Look for evidence against your hypothesis: did comparable pages that you did not expect to move also fall? Did a tracking or site change happen at the same time?
- Summarise findings in a short table with the numbers, dates and caveats, and only then plan actions.
Metrics to compare, and what each tells you
| Metric | What a drop suggests | Watch out for |
|---|---|---|
| Clicks | Fewer visits from Google Search | Seasonality and demand changes affect clicks directly |
| Impressions | You appear less often, or for fewer queries | Changes in how often Google shows certain SERP features |
| Average position (per query) | You rank lower for that query | Site-wide averages are distorted by query mix |
| CTR (per query) | Fewer searchers click when you appear | AI Overviews or new SERP features can cut CTR without a ranking loss |
| Number of ranking queries / pages | Breadth of visibility changed | Long-tail query anonymisation and reporting changes |
For more on the gap between ranking and traffic changes, see rankings vs traffic impact and AI Overviews CTR vs ranking loss.
Avoiding false conclusions
Several situations regularly produce misleading before-and-after numbers:
- Seasonal peaks and troughs. The December 2025 core update (11 to 29 December 2025, per the dashboard) overlapped the holiday season, when demand shifts sharply in many industries. Always check the year-over-year control. Our guide to seasonality vs algorithm updates covers this in detail.
- Overlapping updates. When a spam update and a core update run back to back, a single before/after comparison cannot separate them.
- Your own changes. A migration, redesign, template change or large content release during the window can explain the movement on its own. See site migration or algorithm update?
- Reporting issues. Search Console occasionally has logging problems, which Google lists on its data anomalies page. Rule these out with our tracking-error checklist.
- Survivorship in page lists. If you only analyse pages that still get clicks, you miss pages that dropped to zero.
Common mistakes
- Using "last 7 days vs previous 7 days" on the day the update is announced.
- Comparing unequal windows, for example 10 days against 14.
- Reporting a percentage change without the absolute numbers, which exaggerates movements on small pages.
- Treating correlation with an update date as proof; always look for competing explanations.
- Mixing Web, Image and Discover traffic in one number.
Present the result honestly
A good before-and-after summary includes the dates used, the windows compared, the year-over-year control, the absolute and percentage change, and a confidence note. Our SEO reporting guide and the series article on reporting an algorithm update to clients show how to structure it. Then decide next steps using the core updates hub, keeping in mind that Google says improvements can take months to be recognised.
Related Guides
- Seasonality or Algorithm Update? How to Tell the Difference
- Is It an Update or a Tracking Bug? Ruling Out Analytics Problems
- How to Recover From a Google Spam Update
Frequently Asked Questions
What is the best window length for a before-and-after analysis?
Seven days works for busy sites; 14 or 28 days is safer for smaller sites. Whatever you choose, keep both windows the same length and aligned by weekday.
Why compare with the previous year?
It shows how much change you would expect from seasonality alone, so you can estimate the part that may be linked to the update.
Can I measure impact during the rollout?
You can monitor it, but Google recommends waiting at least a week after completion before analysing. Mid-rollout numbers are still moving.
Should I use third-party visibility tools?
They are useful for competitor context, but Search Console shows your actual clicks and impressions from Google and should anchor the analysis.
How do I report uncertainty?
State the dates, windows and controls used, show absolute numbers, and say clearly when another event in the same period could explain the change.
Measure first, then act
Careful measurement saves you from fixing problems you do not have. If you want a second pair of eyes on your update analysis, our SEO team can run the comparison and turn it into a clear action plan. Get a free quote to start.
