Why visitor times deserve their own analysis
An online shop does not close. For retailers, that is one of the most important benefits of digitalization: 83 percent of surveyed retail companies name the removal of time limits on their offers, meaning the end of fixed opening hours, as a positive effect (Bitkom). The basis is the company survey in the study report "Digitaler Handel in Deutschland": 505 wholesale, retail and motor vehicle trade companies with at least ten employees (Bitkom). The figure describes how companies assess the situation, not a measurement of visitor behavior. It only says that the clock no longer sets a limit for retail.
For customers, too, buying online is part of everyday life. In the same study's population survey of 1,120 people aged 16 and over (Bitkom), 96 percent of respondents had bought something online in the last twelve months, and more than a third of online shoppers (36 percent) even buy online at least once a week (Bitkom). The device used most often is the smartphone, at 73 percent well ahead of laptop, desktop computer and tablet (Bitkom). A purchase that no longer happens at a desk is not tied to office hours. When it actually happens is something the survey does not answer, and it cannot answer it for your own website either.
Your own measurement, on the other hand, answers the question precisely. It shows in which hour the website is in demand, and very practical decisions follow from that:
- Maintenance: updates, migrations and restarts belong in the hours when the fewest people use the site.
- Newsletter: a mailing should arrive shortly before visitors are active anyway, not when hardly anyone reads.
- Support and on-call: phone, chat and technical on-call duty are needed where the curve rises.
- Campaign start: a sale is better launched at a time when the team is still around and the peak is still ahead.
- Deploys: even small code changes carry a risk. Rolling them out shortly before the evening peak means finding errors only under full load.
What XICTRAQ® measures: 24 bars and a grid of 7 × 24 cells
In XICTRAQ® cookieless analytics there are two views of visitor times. The Traffic by hour card condenses the selected period into 24 bars, one per hour of the day, and highlights the peak hour. The card header also shows it as a hint, for example "Peak 20:00". The card can be placed on the dashboard and in the analytics overview, for a single website, a group or all websites in the account. On the dashboard, the period is chosen per card from 24 hours, 7, 30 or 90 days; the default is 30 days. In the analytics overview, the page's period applies, from today up to one year or as a custom range.
The second view is the activity heatmap in the Performance tab of the analytics view. It splits the same data by weekday as well: seven rows from Monday to Sunday, 24 columns for the hours. The darker a cell, the more pageviews fell into that hour, measured against the strongest cell in the period. A cell shows its exact value as soon as the mouse pointer rests on it. This view has nothing to do with the click heatmap, which shows where visitors click on a page: it measures time, not space.
The hour with the most pageviews in the period, shown as a hint in the header of the Traffic by hour card and as the darkest bar.
The heatmap shows whether Monday morning looks different from Sunday evening. The total curve across all days blurs exactly this difference.
Country, device, channel, page or campaign from the filter bar also apply to the heatmap. That way smartphone and desktop times can be viewed separately.
On the Starter plan, analytics history goes back 90 days; from the Business plan on, it is unlimited. A year-over-year comparison needs data that reaches back far enough.
The analysis counts pageviews, not people. A visitor who reads ten pages in the evening contributes ten times to that hour. Events such as clicks on downloads are not included here, and requests recognized as machines, meaning bots, data centers and known crawlers, stay out as long as the "Incl. bots" toggle is off. That matters for visitor times because automated requests like to run at night, and a quiet hour would otherwise look busier than it is. How events are recorded separately is shown in the article Track downloads and outbound links without cookies.
The analysis needs no identifier in the browser. The time is part of every single record; recognizing the visitor is not required for it. Where the line between anonymous and pseudonymous measurement runs and what that means for consent is covered in the article Anonymous or pseudonymous? Web analytics under the EDPB guidelines.
Local time at today's offset
Every pageview is stored in Coordinated Universal Time (UTC). So that the heatmap shows the clock the viewer knows, the dashboard sends the browser's time zone offset with every analytics request. It comes from the JavaScript method getTimezoneOffset(), which returns the difference between UTC and local time in minutes. The sign takes some getting used to: the value is positive if local time is behind UTC and negative if it is ahead; for UTC+10 the method returns -600 (MDN).
In Germany, legal time is Central European Time, which is UTC plus one hour, and during daylight saving time Central European Summer Time at UTC plus two hours (EinhZeitG). A browser in Hanover therefore reports -60 in winter and -120 in summer. XICTRAQ® uses this value to shift every stored timestamp to local time before weekday and hour are determined. In simplified form, the calculation looks like this:
// Browser offset in minutes, in Germany -60 in winter, -120 in summer
const offset = new Date().getTimezoneOffset();
// Stored in UTC, shifted with today's offset
function toLocalHour(utcDate) {
return new Date(utcDate.getTime() - offset * 60000).getUTCHours();
}
// Pageview on 20 Oct 2026 at 20:30 summer time, i.e. 18:30 UTC,
// analyzed on 14 Nov 2026 with the winter offset -60:
toLocalHour(new Date('2026-10-20T18:30:00Z')); // 19, not 20The catch is in the word "today". One offset applies to the whole period, namely the one in effect on the day of the analysis. Summer time ends on the last Sunday in October at 3 a.m., when the hour count is set back from 3 a.m. to 2 a.m. (Summer Time Ordinance). In 2026 that was October 25. Anyone looking at the last 30 days on November 14 is therefore analyzing roughly ten days of summer time with the winter offset. For those days, every pageview appears one hour early: the evening peak at 8 p.m. shows up at 7 p.m. in the older part of the period.
As a result, the peak smears. Instead of one clear bar, two neighboring bars stand high, and the quietest hour seems earlier than it is in winter. XICTRAQ® calculates this way on purpose so that the heatmap and the time series use the same clock; for the analysis, it means choosing the period with care:
- For the hourly analysis, choose a period that lies entirely after the change, so in mid-November the last 7 days or a custom range starting October 26.
- The next change follows on the last Sunday in March (Summer Time Ordinance); in 2027 that is March 28. Until then, the winter offset stays consistent for new data.
- Long periods need distance: a 90-day range is only free of days with the other offset again about three months after the change.
- If a peak shifts by exactly one hour across the change, that is usually a clock effect and not a change in behavior.
- Anyone sharing analyses with colleagues in another time zone should note in which local time the view was created.
The heatmap shows the local time of the browser in which it is opened. If the team that looks after the website sits in a different time zone from the business, the view shifts by exactly that difference. Conversely, visitors from other countries bring their own clock: an evening visit from North America falls into the night in Germany. Sites with many international visitors should therefore filter the hourly analysis by country.
What the clock change triggers elsewhere in monitoring, from the doubled hour in cron to a maintenance window spanning the change, is described in the article Daylight saving time in monitoring.
Reading the curve: peak, plateau and weekly pattern
An hourly curve has three components that are worth looking at separately. The peak is the single hour with the most pageviews. The plateau is the stretch of hours in which the curve stays high. And the trough is the time when hardly anyone comes. For decisions, the plateau often matters more than the peak: if the peak is at 8 p.m. but the curve is already high from 5 p.m., the critical time starts in the late afternoon, not in the evening.
The heatmap also shows whether the pattern shifts over the week. Two shapes are common. Websites that mainly address business customers have a clear weekday peak in the morning and drop off at the weekend. Shops for consumers tend to show flatter weekdays with an evening peak and a strong Sunday. Both are patterns from experience, not rules; what counts is your own heatmap.
| Pattern in the heatmap | Common cause | What follows |
|---|---|---|
| Peak on weekday mornings | Business customers research from the office | Staff support in the morning, consider maintenance at the weekend |
| Evening peak and strong Sunday | Consumers shop after work | On-call in the evening and on Sunday, maintenance early in the morning |
| Bright stripe at one fixed hour | Newsletter, ad or social media post | Check with the channel or campaign filter before treating the hour as normal |
| Even pageviews during the night | Unrecognized automation or visitors from other time zones | Filter by country and device, check bot settings |
| Peak shifted by exactly one hour | Clock change within the selected period | Choose a period entirely after the change |
Outliers deserve a second look. A newsletter, a TV mention or an ad with a fixed run time creates a bright stripe that says nothing about normal demand. The channel or campaign filter shows whether the peak comes from regular traffic or from a single mailing. The amount of data counts too: with little traffic, chance decides the individual hour. A longer period helps then, but it should not contain a clock change.
Comparing by device is revealing. The device filter lets you put the smartphone curve next to the desktop curve. If the peaks fall at different times, a fault in the evening mainly hits the mobile view, and that is exactly what should be checked first after a deploy. Whether the evening visitors are the same people as in the morning can only be answered in cohorts without cookies, as the article on returning visitors without cookies shows.
Peak hour, start of the plateau and quietest block: anyone who notes these three values for weekdays and the weekend has the basis for maintenance, mailings and on-call duty in one place.
Put the maintenance window in the quietest hour
The most obvious conclusion from the heatmap is the date for maintenance work. The quietest cell of the week is the first candidate. Better than a single hour is a block of two or three quiet hours, so that an update does not run into the next peak. The visitor curve is only one side, though: nightly imports, till closings or syncs with the merchandise management system can make an empty hour expensive anyway. How to put such processes next to the curve and announce maintenance cleanly with status code 503 is covered in the article Maintenance windows: 503 and Retry-After done right.
Once the slot is found, it is stored in XICTRAQ® as a maintenance window so that uptime monitoring does not alert during the work. Recurring windows are available from the Starter plan and consist of weekday, start and end; up to 20 such windows can be set up per website. Checks keep running inside the window and count toward availability; only the notification is suppressed. A window may span midnight, for example from Sunday 11 p.m. to Monday 1 a.m.
A conversion is needed here. The heatmap shows local time, but XICTRAQ® evaluates recurring maintenance windows against server time in UTC. Central European Time is one hour ahead of UTC, summer time two hours (EinhZeitG). If the heatmap shows the trough on Tuesday between 3 and 5 a.m., the entry in winter reads Tuesday 02:00 to 04:00 UTC.
A window entered in UTC stays in UTC. From the start of summer time on the last Sunday in March at 2 a.m. (Summer Time Ordinance), the same window falls one hour later in local time: 3 to 5 a.m. becomes 4 to 6 a.m. Anyone who has tied maintenance to the visitor curve therefore checks the entry twice a year, ideally with a fresh heatmap that lies entirely in the new time.
How to choose the duration, buffer and scope of a window and where the line between planned maintenance and a real incident runs is covered in the article Maintenance windows in monitoring: avoiding false alerts. The settings themselves are described in the maintenance window documentation.
Align newsletters and campaign starts with the curve
Mailings follow the opposite logic to maintenance: they should arrive when readers are active anyway. A good starting hypothesis is the start of the plateau, not the peak itself. An email that lands in the inbox shortly before the peak reaches readers who have time right then, and it does not steer the extra traffic onto exactly the hour in which the website is already under the most load. Whether the hypothesis holds shows up in the heatmap afterwards: filtering on the mailing's campaign, meaning its UTM parameters, reveals at which hours newsletter visitors actually arrive.
Alternating between two send times gives a reliable answer after a few weeks. The key is to change only the time and keep subject line, content and weekday as similar as possible; otherwise the comparison measures everything at once and in the end nothing in particular.
The time of day carries particular weight in Black Week. Of those who wanted to use Black Friday and Cyber Week offers in 2025, 69 percent planned to buy online (Bitkom). The survey covered 1,120 people aged 16 and over (Bitkom). At which hour people buy cannot be read from the survey. For the campaign start, your own curve is what counts: if a sale starts at midnight, the first error falls into an hour when nobody responds, and the evening peak hits a state that nobody has checked. If it starts in the morning ahead of the plateau, the first orders run under supervision, and there is time for corrections before the peak.
A campaign usually adds to existing demand. The usual evening peak therefore stays and becomes the stress test in campaign week; response times in exactly those hours deserve special attention. For the comparison with the previous year, a second shift comes on top, because Black Friday falls on a different date every year. How Q4 reports still stay comparable is explained in the article Year-over-year in Q4: Black Friday moves in the calendar.
Staff support and on-call according to the curve
For consumer offers, a website's peak hour often lies outside regular office hours. Organizationally, that is no small matter, because regular evening and weekend work is the exception for most employees. In 2025, 14.4 percent of people in employment aged 15 and over worked permanently or regularly in the evening, meaning between 6 p.m. and 11 p.m. (Destatis). Regular night work applied to 4.3 percent (Destatis). On Saturdays, 16.8 percent of people in employment worked permanently or regularly, on Sundays 9.1 percent (Destatis).
For planning, this means: on-call coverage for the evening peak is a deliberate decision, not a side issue. It should follow the measured curve rather than a blanket assumption. If the plateau lies between 6 p.m. and 10 p.m., someone is needed in those hours who recognizes order problems, payment errors and outages and can respond. The hours before and after can be covered by alerting.
Then there is German working time law. After the end of daily working time, employees are entitled to an uninterrupted rest period of at least eleven hours (ArbZG). Someone handling requests until 10 p.m. therefore starts no earlier than 9 a.m. the next day. The law allows exceptions for certain sectors; for the standard case, this calculation applies. Covering both morning and evening therefore takes two people or staggered shifts.
In XICTRAQ®, alerts go out by email, Slack, Microsoft Teams, Discord, Telegram or webhook, depending on the plan; quiet hours can be set per channel, and on-call schedules with handover and escalation are available from the Business plan. How escalation levels and on-call duty work together is shown on the page Alerting and notifications.
Step by step to your own hourly analysis
- Choose a period that lies entirely after the last clock change and covers at least two full weeks, later preferably four.
- Look at the Traffic by hour card and note the peak, the start of the plateau and the trough.
- In the activity heatmap in the Performance tab, check whether weekdays and the weekend look different.
- Explain outliers with the channel or campaign filter and exclude them where needed.
- For maintenance, convert the quietest block to UTC and enter it as a recurring window.
- Place the newsletter send time ahead of the plateau as a hypothesis and verify it with the campaign filter.
- Plan on-call and support for the plateau and factor in rest periods.
- Repeat the analysis after every clock change and after larger campaigns.
If you want to see the analysis first, you will find it in the live demo with sample data. Which analytics features are included in which plan is listed on the pricing page, and how the individual views are used is described in the web analytics documentation.
Measure demand times instead of guessing
Visitor times are among the simplest analyses in web analytics and among the most often skipped. Yet 24 bars and 168 cells hold the answer to questions that come up all the time: when does maintenance disturb least, when does the newsletter pay off, when does someone need to be reachable? The answer applies to your own website and to the selected period. It stays correct as long as the period contains no clock change and outliers are recognized. Taking that into account gives you a curve instead of a guess, one that can be checked again every month, and a plan for Black Week that fits your own customers.
This article is based on: Bitkom, study report "Digitaler Handel in Deutschland" 2026 (population survey of 1,120 people aged 16 and over and survey of 505 retail companies, both 2025); Federal Statistical Office (Destatis), indicators on quality of work, pages "Abend- und Nachtarbeit" and "Wochenendarbeit" (Labour Force Survey 2025); Units and Time Act (EinhZeitG), Section 4; German Summer Time Ordinance, Section 2; Working Hours Act (ArbZG), Section 5; MDN Web Docs, "Date.prototype.getTimezoneOffset()". The notes on working time law are not legal advice.