Safe tests, run history and retries
Test with a simulation, read the run history and see how retries work.
Read the lesson
Goal: Test a workflow without sending anything. Prove it live on your own contact. Read run history to catch failures.
You need: Lesson 1 of this course. Core, Growth or Franchise.
Why this matters
A workflow that fails quietly is worse than no workflow. The owner thinks reminders go out. They do not. No-shows rise and nobody knows why. Testing before you publish catches most mistakes. Reading run history after you publish catches the rest. Both take a few minutes. Skipping them can cost a client.
Three kinds of test
1. The builder Test button. It runs a simulation with sample data. It sends nothing and costs nothing. Use it to check the path each sample takes through your conditions. It cannot prove that a real text reaches a real phone.
2. A gated live test. Publish the workflow with a test gate in front of every action. The gate is a Condition that lets only your own test contact through. For example, contact.email Equals your test address. Real customers stop at the gate. Your test contact goes through and gets real messages on your phone. This proves the trigger fires, the text arrives and the timing is right. It uses real credits for your test messages, so keep it short.
A gated live test
- YouAdds the test gate and publishes
- Real customersStop at the gate and get nothing
- YouBooks a test appointment for yourself
- ServiceAgentSends the real text to your phone
- YouChecks the timing, then removes the gate
3. Watching the first week. Remove the gate. Then read the run history every day for a week. Real data has surprises. Missing phone numbers, odd names, contacts with two jobs on one day.
Three kinds of test, in order
- 1Builder TestA simulation with sample data. Free, sends nothing
- 2Gated live testOnly your own contact passes the gate
- 3Remove the gatePublish again
- 4Watch run history for 7 daysReal data brings surprises
Reading run history
Open the workflow and go to History → View run history. Each run shows its steps and its result. A failed step retries up to 3 times. You will see "Attempt 1 of 3", then 2, then 3. After 3 failed attempts, the run stops. Nothing more happens for that contact.
Common causes of failure:
- The contact has no mobile number, so
Send SMShas nowhere to go. - The contact opted out. Opt-out is always enforced, so the text does not send.
- A field the
Conditionreads is empty, so the run takes an unexpected path. - An outside system behind an
HTTP Requestis down.
Retries help with short outages. They do not fix missing data. Fix the data, or add a Condition that skips contacts without it.
Reading a failed run
- Run history shows a failed attempt
- Did a later attempt succeed?YesA short outage. No action neededEnds hereNoThe run stops after Attempt 3 of 3
- Is data missing, such as a mobile number?YesFix the data, or add a Condition that skips itNoCheck the next cause
- Did the contact opt out?YesCorrect result. Opt-out is always enforcedNoCheck the outside system behind the HTTP Request
- Note the cause in your weekly log
Check Billing → Usage history in the same week. Compare credits used with what you expected. A big gap means runs are failing, or firing more often than you planned.
After the first week
Move from daily checks to a weekly check. Put it on your calendar for the same day each week. Each week, open View run history for every workflow that sends to customers. Look for three things: new failures, runs that stopped, and a sudden change in run count. A run count that drops to zero usually means the trigger stopped firing. For example, the team stopped marking jobs Completed, so Appointment Completed never fires. A run count that doubles usually means a trigger is firing twice, or a new location joined.
Add the weekly check to the data checks in Prove lesson 2. Then the monthly report always rests on workflows you know are running.
By business size
An owner-operator has few runs. One failure can be a real customer. Read every failed run. A fleet has hundreds of runs. Read the failures and sort them by cause. Fix the biggest cause first. A multi-location client may have the same workflow in each location. Check each one. A fix in one location does not copy to the others.
Build it in ServiceAgent
- Open your draft workflow. Click
Testwith 3 sample contacts. Check each path. - Add a test gate:
Conditionwithcontact.emailEquals your test address, before the first action. - Click
Publish. The workflow is live but only your contact passes the gate. - Trigger it with your own contact. For a reminder, book a test appointment for yourself.
- Confirm each message on your phone and in
Conversations. Check the timing. - Open
History→View run history. Confirm every step shows success. - Remove the test gate. Click
Publishagain. - For 7 days, open
View run historyeach morning. Note every failure and its cause. - Fix data problems at the source. Add a
Conditionto skip contacts with missing data. - Check it works: at the end of the week, compare runs with the number of events. They should match.
Where to read run history
Execution History
Worked example: Fresh Start Cleaning
Fresh Start has 12 cleaners on Core. You install the Appointment Reminder template. It texts 24 hours and 1 hour before each visit. Fresh Start has about 240 visits a month.
Credits: the template is one workflow: Send SMS, a Delay "Wait until 1h before", then a second Send SMS. So each visit makes 1 run. Each run is 1 credit plus 2 texts at 2 credits each. That is 1 + 4 = 5 credits a visit. 240 × 5 = 1,200 credits a month.
Fresh Start reminder credits
Over the first month of checks, View run history showed 9 failed runs. Each one tried 3 times, so 27 attempts failed. All 9 contacts had a landline in the phone field, not a mobile. You asked the office to collect mobile numbers at booking. You also added a Condition that skips contacts without a mobile and creates a call task instead.
9 of 240 is about 4% of visits. Without the check, about 4% of customers would have had no reminder, and nobody would have known.
Common mistakes
- Mistake: Trusting the builder
Testalone. Fix: Run a gated live test on your own contact. - Mistake: Forgetting to remove the test gate. Fix: Remove it and publish again. Then check the first real run.
- Mistake: Never opening run history. Fix: Read it daily for the first week, then weekly.
- Mistake: Expecting retries to fix bad data. Fix: Fix the data or skip those contacts.
Daily run-history check
- Open History, then View run history
- Count failed runs
- Note the cause of each
- Fix the data or add a skip Condition
- Compare credits in Usage history
Do this now
Take one workflow in your practice account. Add a test gate. Publish it. Trigger it with your own contact and confirm every message. Remove the gate and publish again. Deliverable: a 3-line test log with the date, result and any fix. Time: 25 minutes.
Practice the conversation
- Owner How do I know the reminders really go out?
- Operator I test them on my own phone before any customer gets one.
- Owner And after that?
- Operator I read the run history every day for the first week.
- Owner Did you find anything?
- Operator Yes. Nine customers had landlines on file. Texts cannot reach them.
- Owner So they got nothing?
- Operator Not anymore. The office now gets a task to call them instead.
- Owner Good. Can the office ask for mobiles at booking?
- Operator Yes. I added that to their booking checklist.
Quick check
- What does the builder `Test` button do?
- Sends real texts to your test contact
- Runs a simulation with sample data and sends nothing
- Publishes the workflow for one day
- Charges 1 credit per step
- A step fails because a contact has no mobile number. How many times does the run try that step?
- Once
- 2 times
- Up to 3 times
- Until it works
- A reminder workflow sends 2 texts per visit. Each run is 1 credit, and each text is 2. A client has 200 visits a month. How many credits?
- 400
- 600
- 800
- 1,000
ServiceAgent UniversityLearn to run a service business on autopilot. Free courses, Blueprints and certification for Service Operators.