Tutorial · 6 min read
Getting Started with PostEngage: Your First Week, Honestly Paced
On this page
Most setup guides are written as though the goal is to finish. This one is written as though the goal is to still be running it in three months, which is a different thing and takes about a week.
The pace matters because the failure mode of comment automation is not technical. It is sending something slightly wrong to a few hundred people before you notice. Everything below is arranged so that you see what a reply would look like before anyone else does.
Day one: connect the account, and read what you granted
Connecting is an OAuth handshake with Meta. You will see a permissions screen listing what the app is asking for, and it is worth actually reading rather than clicking through, because it is the honest summary of what any tool on the official API can and cannot do.
Two things to check once it says connected:
- The account name shown is the one you meant. People with several accounts pick the wrong one surprisingly often, and everything afterwards silently applies to an account you were not thinking about.
- The connection reads healthy, not merely present. A token that has expired still shows an account name. It stops working the moment something needs sending.
If you disconnect later, automations stop entirely rather than queueing. That is deliberate. A queue that drains three days later sends replies to conversations that have moved on.
Day two: build one automation, and only one
The instinct is to build the full set on the first afternoon. Resist it, because you cannot tell which of five automations produced a bad reply, and you can always tell which one did when there is one.
Pick your most repetitive comment. For most accounts it is a single word under a specific post: price, link, guide.
The five decisions, in the order the screen asks for them:
- When this happens. Comments on a post, and whether that means any post, one chosen post or posts you pick.
- Only if. The words that should trigger it, and the words that should stop it —
not interestedis the one everybody needs and nobody thinks of first. Matching ignores case and punctuation, soPrice?andpriceare the same thing. - Then do this. A public reply under the post and a private reply by DM. Keep the public reply to a few words. It is a receipt telling the person to check their DMs, not the answer itself. The private reply is the actual answer, and it has to land inside Instagram's messaging window.
- Ask for something. Optional. After the DM, ask for their email, phone or first name.
- Name it and go live. The checkbox "Treat people who trigger this as leads" is off by default. Tick it: the people who ask are worth having even when the reply itself is unremarkable.
Day three: try it, before anyone else sees it
On the automation's page, press Test it and use Try it. The distinction between trying and going live is the whole point of this guide.
Type what somebody might write. Try it shows whether the reply would send, the public and private replies exactly as they would go out with the variables filled in, and runs every check against the live account. Nothing is sent. You get six tries an hour per automation.
Read the result for tone rather than correctness. Correctness you can check by reading the template. Whether it sounds like a human who runs this account is the thing you can only judge by reading it as the person receiving it would.
Day four: go live on one post, and watch what gets stopped
Turn it on for the single post you built it for. Then, later that day, open Activity and switch to Blocked. This is the screen most people never look at and the one that tells you whether your setup is sane.
A healthy first day looks like a stack of POLICY.WINDOW_CLOSED and a few POLICY.DUPLICATE. Those are not faults. They are the platform's rules and Instagram's redelivery behaviour, both working as intended.
The rows worth acting on are different:
Normal, ignore
Act on this
Day five to seven: widen, one automation at a time
Only now add the second automation, and repeat the same loop: build, try it, go live, read the blocked list. It feels slow on day five and it is the reason the thing is still running in April.
Two settings become worth touching once you have more than one automation:
- Quiet hours. Set on the Sending page for each account, in the account's timezone. The default nobody regrets is roughly 9pm to 8am.
- Only the first time somebody triggers this, a checkbox in step 2, if you post frequently. It stops the same follower being greeted on every reel.
What the first week should have taught you
Not that automation works, which you can read anywhere. That your automation refuses correctly, which you can only learn by watching it refuse.
The sixteen checks run on every reply in a fixed order, and the reason a reply was not sent is recorded against the run rather than swallowed. If you want the full list of what can stop a send and why each one exists, the checks are documented here.
If something in the first week goes sideways, the fastest diagnostic is almost always Activity rather than the automation itself. The automation tells you what you asked for. Activity tells you what actually happened, which is the version worth trusting.
