Metrics and analysis
First 30 days with coexistence: what to switch on, in what order
A four-week plan from connection to the first useful automation, without burning the number or scaring the team.
Also available in: Português · Español
Connecting is the fast part. What separates an operation that uses coexistence from one with a connected, forgotten number is what happens over the following four weeks.
This plan is deliberately conservative. It delays the payoff by two weeks and avoids the two expensive outcomes: a number with sunk quality, and a team that has learned to distrust the project.
Week 1 — connect and do nothing else
Goal: the connection exists and is stable.
- Connect, respecting the sync window
- Run the round-trip test: send from the system, reply from the number, confirm it appears on both sides
- Warn the team first, using the script in does the app keep working
- Automate nothing
It feels wasteful and it is not. Week one exists to find out whether anything broke — and, if it did, to fix it with no automation running on top.
Week 2 — observe
Goal: know what happens in this channel before touching it.
Watch three numbers, acting on none:
| Number | Why it matters |
|---|---|
| Conversations per day | decides whether automating is worth it |
| Peak hours | decides when automation gets in the way |
| What repeats most | decides what to automate first |
The third is the most valuable and the only one requiring reading rather than counting: skim fifty conversations and note what the team types every single time. That is where the first automation comes from — not from a list of ideas.
Week 3 — one automation only
Goal: take the most repeated sentence off the team's keyboard.
Criteria for choosing:
- repetitive — happens several times a day
- predictable — the message is nearly identical every time
- low risk — if it goes out wrong, no damage is done
- expected — the customer does not find it odd to receive
Appointment confirmations, receipt acknowledgements, answering "do you serve this area?". Never start with negotiation, billing or anything involving a number with money attached.
A coexistence rule, defined before switching on: when a human is already in the conversation, automation stays out. Pick the criterion — outside business hours, or no human reply for X minutes — and write it down.
Week 4 — measure and choose the next one
Goal: know whether it worked, with numbers rather than impressions.
Three questions:
Did the team type less? Ask them. If the answer is "sort of", the automation you picked was not the most repetitive one.
Did any customer react badly? Look at blocks in the period and read the conversations where automation stepped in.
Did anyone have to correct it? If an agent had to write "please ignore the previous message", the coexistence rule is wrong.
With all three answered, choose the second automation — or go back and adjust the first, which is the most common and most productive outcome.
What not to do in these 30 days
Do not blast a list. Not even a small one, not even "just to let them know". A new number on the platform has low limits and a reputation to build; the piece on number quality explains what that costs.
Do not build an options menu. It is the automation that most feels like productivity and most pushes customers away. If the channel felt personal, it stops being so on the first message.
Do not connect the other numbers yet. One at a time, and the second only after the first is running.
Do not promise the client a report in week 1. Promise it in week 5, when you know what the channel generates.
The schedule, summarized
Week 1 connect · test · warn the team · zero automation
Week 2 observe · count conversations · note what repeats
Week 3 one automation, the most repetitive and lowest risk
Week 4 measure · adjust · pick the next
After second automation, or second number
Four weeks sounds slow to someone who connected in a hurry. But the fast path — switching everything on in week one — produces, with predictable regularity, a low-quality number and a team convinced the system is a nuisance. Recovering both takes longer than four weeks.
After the 30 days
With one automation running, the channel measured and the team comfortable, the next decisions are easy: the second automation comes off the same week-2 list, and the second number repeats the process in half the time.
For anyone running several clients, the arrangement changes and coexistence for agencies is worth reading before repeating this ten times.
Frequently asked questions
What should I automate first?
The most repetitive, lowest-risk message: a receipt or appointment confirmation. It is what the team types every day and what no customer finds odd to receive automatically.
When can I start sending at volume?
After the number has real conversation history and a raised limit, which takes weeks. Volume in week one is the fastest way to sink quality.
Do I need reporting from day one?
Measuring, yes. Sending a report to someone, no. Two weeks of watching show what is signal and what is noise before you create an alert nobody reads.
What if nothing goes wrong in week one?
Good, and still do not accelerate. Most quality problems appear between weeks two and four, when volume starts rising.