Time Series
The time series chart on the Results page shows how the conversion rate for each variation has changed over time. While the variation table gives you the cumulative totals, the time series tells you the story behind those totals — revealing patterns, trends, and anomalies that the aggregate numbers hide.
Reading the chart
The chart shows one line per variation, including the control. The horizontal axis is time, in 4-hour buckets. The vertical axis is the conversion rate for the visitors exposed inside that bucket. Each variation has a distinct color that matches the variation table.
When lines are close together, the variations are performing similarly. When they diverge — one line climbing while another stays flat — that is the experiment producing signal.
Customizable date range
Use the date range picker above the chart to zoom into a specific period. You can view the full experiment duration, or narrow down to a specific week or day to investigate a particular event. The date range affects only the chart — the variation table always shows cumulative totals for the full experiment.
The selected date range is preserved in the URL — share or bookmark a link to bring a teammate to the same window. See Bookmarking and sharing for details.
Interactive tooltips
Hover over any point on the chart to see a tooltip with the exact conversion rate for each variation on that date. This is useful for identifying the exact day a variation started pulling ahead or the day a sudden change occurred.
What to look for
Novelty effects
A novelty effect happens when a variation performs unusually well (or poorly) in the first few days simply because it is different from what visitors are used to. Visitors may click on a new button more often just because it caught their eye, not because it is actually better.
On the time series chart, a novelty effect looks like: the variation line spikes high in the first few days, then gradually falls back toward the control line. If the lines converge after the initial spike, the early "winner" may not have been a genuine improvement.
This is one of the main reasons to run experiments for at least a full week — to let novelty effects wear off before reading the results.
Seasonal patterns
Many sites have different traffic and conversion patterns by day of the week. E-commerce sites often see higher conversion rates on weekends. B2B sites often see their best traffic on Tuesday, Wednesday, and Thursday. The time series chart makes these patterns visible.
If you see all variation lines rising and falling in unison on a weekly cycle, that is a seasonal pattern — both variations are equally affected by the traffic change, which is what you would expect. What you watch for is whether one variation's line diverges from the others, especially during high-traffic periods.
Understanding when the experiment started working
Sometimes an experiment is inconclusive for the first several days and then starts showing a clear signal. The time series chart lets you see exactly when the divergence started. This might coincide with a marketing campaign that brought in more relevant traffic, a change to another part of the site that affected user behavior, or simply the point at which enough data was collected to distinguish signal from noise.
Sudden changes
A sharp vertical change in all variation lines on the same day indicates an external event that affected all traffic equally — a promotional email, a viral post, a news mention, or a site outage. This is not a problem with your experiment. A sharp change in only one variation's line, however, may indicate that the variation code broke or changed on that date and deserves investigation.