Where to find it
- Open Brand → Workflows.
- Click the workflow's name to open it.
- Switch from Editor to Runs at the top.
A date range picker at the top drives everything below it. It is set to the last 30 days when the view opens. Under it you get the Workflow triggers chart with the total for the chosen range, then the Runs table and the Starts held back table.
The Workflows list itself also has a Sessions column, so you can compare volume across every workflow in the brand at a glance without opening each one.
What the number counts
One trigger is one time the workflow actually started in a conversation. It is an entry count, not a completion count — it tells you how many people the script began talking to, not how many reached the end. How each run ended is in the Runs table below.
A brand-new workflow shows nothing until the first visitor sets it off.
Reading the trend
The chart plots those starts over time, so the shape is usually more informative than the total:
- A steady rise means more people are meeting the script — good news, as long as it is doing something useful once they get there.
- A sudden drop on a specific day almost always means something changed on that day. Check whether the workflow was switched off, whether its trigger or conditions were edited, whether the frequency setting was tightened, or whether the brand's chat widget came off a page.
- A flat zero is a configuration problem, not a traffic problem. Look at Starts held back first, then work through the checklist in the article on several workflows competing.
The Runs table
Every time the workflow started in the chosen period, newest first, with these columns: Started, Started by (the trigger, or who started it on purpose), State, Ended as, Steps and Failed steps. Filter by State, Started by or Ended as to narrow the list down.
Open a run with the arrow at the start of its row to see:
- every recorded step, in the order it ran, and what it recorded;
- the branch it took — the button the visitor pressed, or whether an API call worked;
- the values the run collected;
- for a Call an API step, what the API answered — and, when the call had trouble, the first part of its reply.
If you have access to the inbox, Open the conversation on the row takes you to the conversation itself.
A step is recorded when it finishes, and a question, buttons or a form only finish when the visitor answers. So a run still waiting on its first question shows no steps yet.
Starts held back
This table lists every time the trigger matched but a rule kept the workflow quiet — for example a colleague was in the conversation, the visitor did not match Only start if, or another workflow won on priority. Each line says why. A start held back only because the workflow already ran as often as How often allows is not listed: for a greeting that runs once per visitor, that is every returning visitor. Held-back starts are kept for 30 days. Nothing is recorded when the AI simply did not pick a User says workflow.
Failed steps
When a Call an API step fails, or a Start a workflow step cannot start the next workflow, the visitor gets an apology and your team is asked to take over, unless the workflow says what to do instead. Failures from the last 24 hours show up in two places: as a red badge such as 2 failed steps next to the workflow's name on the Workflows list, and as a warning at the top of its Runs view. Click the badge, then open a run to see why.
A call that answers normally but with the wrong content is not counted as failed. To catch that, check the values a run collected.
Comparing workflows
Compare like with like: the same date range, and workflows from the same list. The figures are counted per workflow, so a copy of a script is a separate script with separate figures — the two never add up.
If two scripts run on the same trigger and one gets far fewer starts, that is a targeting difference, not a quality difference — only one workflow runs per event, so the other one is winning those moments. The quieter one's Starts held back table shows it as Another workflow won.
What the Runs view does not show
- Phone calls are not recorded. A workflow that runs on a call leaves no run here and is not in the chart.
- There is no per-channel split. Chat, Facebook, Instagram and email starts are all in the same total.
- Only the newest 1,000 runs and held-back starts of the period are listed. Pick a shorter period to see older ones.
Next step
Low numbers? Work through What happens when several workflows could fire at once. High drop-off around the handover? Review Hand off from a workflow to a human agent.