How to Compare Time Periods in Analytics Without Being Fooled by Seasonality

Short answer: compare like with like. Use whole weeks so weekday patterns cancel out, compare with the same period a year earlier when your business is seasonal, line up festivals and holidays that move between dates, and keep a log of campaigns and releases so you can explain a change. Most alarming week-on-week moves disappear once the comparison is fair.

Table of Contents

1. Four Cycles Hiding in Your Data
2. Choosing a Fair Comparison
3. Moving Festivals and Holidays
4. Keep an Annotation Log
5. Separating Season From Change
6. Key Takeaways

Four Cycles Hiding in Your Data

Any comparison that cuts across one of these cycles unevenly will show a difference that has nothing to do with your site.

Choosing a Fair Comparison

Practical rules:

• Use complete weeks. Seven, fourteen, or twenty-eight days. Never eight or thirty.
• Start on the same weekday.
• Avoid "month to date" against a full month.
• For daily checks, compare a Tuesday with the last four Tuesdays.
• Prefer rates within segments to raw totals when traffic mix shifts (why).

Check that the tools you are comparing use the same reporting timezone, or the day boundary moves (timezones).

Moving Festivals and Holidays

A naive year-on-year comparison assumes the calendar repeats. Many of the events that matter most do not fall on the same date each year.

• Festivals set by lunar or other calendars, such as Diwali, Eid, Easter, and Lunar New Year.
• Sale events whose dates are set by retailers each year.
• Public holidays falling on a weekday one year and a weekend the next.
• School holidays and exam periods, which vary by region.
• Payday timing.

Align by event. Compare "the week before the festival" this year with "the week before the festival" last year, whatever the dates were. If you sell across regions, remember that their holidays differ.

Keep an Annotation Log

Six months from now, nobody will remember why traffic doubled in the second week of March. Write it down when it happens.

Include campaigns, releases, pricing, outages, press, tracking changes, and competitor events you know about. A shared spreadsheet is enough.

Tracking changes deserve special mention. A new consent banner or a tag moved to a different page can change the numbers without any change in behavior.

Separating Season From Change

You shipped a new checkout in November and conversion rose. Was it the checkout or the season?

1. Compare with last year's lift over the same period. If conversion rose a similar amount then, with no redesign, the season explains it.
2. Look at a segment the change did not touch. If desktop got the new checkout and mobile did not, mobile is your control.
3. Look at the step that changed. A checkout redesign should move checkout-step conversion, not product-page conversion.
4. Watch sessions. Is the friction you set out to remove actually gone?
5. Wait. If the gain persists after the season ends, it was real (novelty and regression).

Key Takeaways

• Data contains daily, weekly, monthly, and yearly cycles.
• Compare complete weeks and matching periods.
• Align moving festivals by event, not date.
• Keep an annotation log of everything that could move the numbers.

Conclusion

A fair comparison answers most "why did this change?" questions before they are asked. Choose the period with care and write down what happened.

Start free with DeepSync.