Dialogue views, clicks and conversions are each counted in a specific way. When a result looks wrong, it's usually because two numbers measure different things, or because tracking setup is stopping an event from being recorded. This article explains the most common cases and what to do about them:
Your A/B test isn't splitting traffic the way you set it
Clicks are zero or lower than expected
You have clicks but no conversions
Dialogue numbers don't match a segment
BlueConic numbers don't match GA4, Snowflake or your BI tool
"Variants clicked" doesn't match UTM values
Next Best Action after you fix tracking
For definitions of each metric, see Tracking Views, Clicks, and Conversions.
Your A/B test isn't splitting traffic the way you set it
You set a 50/50 split, but one variant gets most of the views. Check the Evaluate at and After evaluation settings on the dialogue's Optimization tab. Once the evaluation point is reached, for example the evaluation date passes, the After evaluation setting takes over. If it's set to Automatic optimization, BlueConic deliberately shows the better-performing variant more often.
To keep a fixed split, move the evaluation point further out, or choose a different After evaluation action.
You changed the settings, but the split still looks uneven. Two things keep older results in the totals:
The Optimization tab shows totals since the dialogue started. They don't reset when you change settings.
With A/B testing, each visitor keeps the variant they saw first.
To check the current split, compare how much each variant's views grow after your change, not the lifetime totals.
The Original (control) group has clicks, even though it shows no dialogue. This is expected. See Tracking Views, Clicks, and Conversions. To compare the control group with your variants, use conversion rate or conversion lift, not clicks.
For all optimization settings, see Dialogue Optimization.
Clicks are zero or lower than expected
Cause | How to fix it |
The button or link is custom HTML. Custom HTML doesn't record a click for the dialogue on its own | Use a button from the Forms Toolbar, or set a Click event selector that matches your element |
Custom JavaScript opens the link (for example to open a new tab), so BlueConic's shortened URL is never requested | Keep the shortened URL as the link destination, and don't open the page with JavaScript |
For cloned dialogues that record no clicks, and for the shortened-URL requirement in Content and Lightbox dialogues, see Tracking Views, Clicks, and Conversions and Troubleshooting Dialogues.
After changing click settings, test in the Simulator and confirm a click is recorded.
You have clicks but no conversions
The click redirects the visitor. If clicking takes the visitor to another page straight away, the browser can leave before the conversion is recorded. Use the page the visitor lands on as the conversion moment, ideally a page on your own domain where BlueConic runs, and review the conversion settings on the Why tab.
If that isn't the cause, check the conversion basics in Troubleshooting Dialogues: whether a click is required first, when the conversion rule was added, and whether the dialogue has a channel.
Dialogue numbers don't match a segment
Dialogue metrics and segments answer different questions, so they rarely match exactly.
Direct conversions count clicks on the conversion element when this dialogue was the visitor's last interaction. They're counted whether or not the visitor finished a form or gave an email address.
A segment counts the profiles that match its rules right now. Some of them may not have an email address.
A listener counts what it's set to watch. If a Behavior listener watches the submit button but your conversion moment watches another element, the two will never match. A listener on a submit button also counts clicks on forms that failed validation.
For why views don't match the size of your target segment, see Tracking Views, Clicks, and Conversions.
To compare like with like, make the listener and the conversion moment watch the same element, or use dialogue form fields so that saving data and counting conversions happen together.
BlueConic numbers don't match GA4, Snowflake or your BI tool
Unique views are higher than the contacts in your export. Unique views measure reach, so they're expected to be higher. The gap usually comes from:
Anonymous visitors. Unique views include visitors without an email address. Exports usually only include identified profiles.
Merged profiles. When two profiles merge, two counted views end up on one remaining contact.
Cleaned-up profiles. Inactive anonymous profiles are removed over time.
New cookies. Cleared cookies and shared devices create extra anonymous profiles, which adds unique views.
To check that your export is complete, compare a short recent period, such as the last 7 days. Compare the dialogue's unique views for that week with a segment of profiles that viewed the dialogue and visited in the same week. These should be much closer than all-time totals.
GA4 or your BI tool shows clicks BlueConic doesn't. BlueConic only counts its own click tracking, such as shortened URLs and click event selectors. Clicks tracked by Google Tag Manager or GA4 are separate. If you report in both places, keep the BlueConic link and add a GTM click trigger for the same link. This gives you BlueConic segments and GA4 reporting from the same click.
"Variants clicked" doesn't match UTM values
These are recorded at different moments:
Variants clicked is recorded when the visitor clicks the dialogue, even if they then close it.
UTM values are recorded only when the visitor arrives on a page whose URL still contains the UTM parameters, and BlueConic runs on that page.
Visitors who close the dialogue, leave before the next page loads, or pass through a redirect that removes the parameters have a click but no UTM value. To find everyone who clicked, use Variants clicked on its own.
Next Best Action after you fix tracking
Next Best Action only learns from clicks and conversions recorded by BlueConic, never from external analytics. If clicks weren't being tracked:
New clicks are used from the model's next scheduled run.
Data recorded while tracking was broken stays in the model until it falls outside the Look back window (model training period). To retrain on clean data sooner, shorten the window.
Consider pausing variants that would otherwise be compared against the old, incomplete data.
Still need help?
Contact BlueConic Support. Include the dialogue name, the date range you're comparing, and a screenshot of each number.