The logs tell you a flow is running; only the numbers tell you it is working. The Stats tab answers two questions: how many people entered the flow, and where they gave up. This guide explains every tile on the screen and every column in the table.
💡 All numbers cover the and . Simulator runs and "try on a real channel" sessions never appear here.
last 30 days
exclude tests
Where the tab is, and what it says when empty
Open the automation and switch to the Stats tab. The period is written at the top of the page: "Last 30 days (tests excluded)."
The table stays empty in two cases:
Stats appear here once published. — The automation was never published. Without a published version there is nothing to count per step.
No sessions in this period. — The automation is live, but nobody entered the flow in the last 30 days.
The five number tiles
Five tiles sit at the top of the page:
Sessions — The total number of sessions started in the last 30 days. This is how many people entered the flow.
Completed — Sessions that ran to the end.
Running — Sessions in progress right now. It sums four statuses: running, awaiting reply, waiting, and paused.
Failed — Sessions that ended with an error. If this is above zero, set the filter to Error on the Logs tab and read the reasons.
Button click rate — The share of people who tapped a button in the first message. With nothing sent it shows a dash (—).
Completed, Running and Failed may not add up to Sessions: sessions in the Interrupted state have no tile of their own but are included in the total. To see interrupted sessions, use the Interrupted filter on the Logs tab.
⚠️ Note: The button click rate only measures buttons in the first message. To find out how often a button in the middle of the flow was tapped, read the table below.
The per-step table
The table headed By step carries one row per step in the flow. The Start and Note steps do not appear: one sends no messages, the other never runs.
The columns are:
Step — The step's type (in bold) with its summary underneath: the first words of the message, how many rules a condition has, how long a wait is. So you can tell which step it is without looking at the canvas.
Visits — How many times the flow arrived at this step. This column is filled for every step.
Sent — On message-sending steps, how many messages went out successfully. On steps that send nothing it shows a dash (—).
Skipped — On message-sending steps, how many could not be sent and were skipped. The most common reason is a closed 24-hour window.
Clicks — The step's interaction. Its contents change with the type of step:
On steps with buttons, each button's label and how often it was tapped, one per line. Card and gallery steps show button labels; a list step shows its row titles.
On steps waiting for a written reply, Replies and how many people wrote.
On steps with a time limit where nobody answered, No answer and the count.
With none of the above, a dash (—).
The gap between Visits and Sent is the most useful piece of information here: when the flow reached the step but no message went out, the difference sits in the Skipped column.
How to read the numbers
Reading the table top to bottom, the Visits column naturally falls: some people drop out at every step. Where it falls sharply is the flow's weak point.
Three example readings:
400 visits on the first message, 90 on the second step. Three quarters of people did not continue after the first message. The wording or the button labels may be unclear.
120 visits on a step, 40 sent, 80 skipped. The flow is running but the messages are not going out. Almost always the 24-hour rule: there is a long wait before that step. Shorten the wait or make the message depend on the person writing again.
50 replies and 150 no-answers on a collect-data step. People see the question and do not answer. Shorten the question, say why you are asking, or extend the reply time.
A concrete example
A training centre wants to know why its signup flow collects so few forms:
Open the Stats tab. Sessions: 620, Completed: 74, Button click rate: 31%.
Look at the By step table. First message: Visits 620, Sent 615, and the Clicks column shows "I'd like more information: 190".
Second step (collect data — phone): Visits 190, Replies 96, No answer 94.
Third step (collect data — email): Visits 96, Replies 74.
The conclusion is clear: the biggest loss is the phone question. Half the people will not give a number.
An experiment: add the reason to the phone question ("We'll only use it to tell you when a place opens up") and reorder the steps to ask for the email first.
Come back to the same table a week later.
How stats differ from logs
The two screens answer different questions and are not interchangeable:
Logs show individual sessions: they answer "why did Ayşe not get the message?".
Stats show the totals: they answer "where in the flow do people give up?".
Start from the logs when you are resolving a complaint; from the stats when you are improving the flow.
Common mistakes and how to fix them
Treating an empty table as a fault. The table is empty when the automation was never published or when no sessions ran in 30 days.
Looking for your own test runs in the numbers. Tests are excluded; simulator and real-channel test sessions never appear.
Reading the button click rate as a whole-flow figure. It only measures the first message.
Ignoring the Skipped column. When "Sent" is low while "Visits" is high, that is where the real problem is.
Trying to compare a period older than 30 days. This screen is a fixed 30-day window; the period cannot be changed. For wider comparisons, look at the account's reports.
Expecting a freshly published version's numbers immediately. The numbers cover the full 30 days, including sessions that ran on the old version.