Rank4AI · Palladium Competitions · 16 September 2026

The whole thing, stage by stage

Everything we are doing on Facebook and Messenger, in the order it happens. Ten stages, five phases. Each one says what happens, what it is waiting on, and where the detail is written up.

Facebook Page and Messenger ManyChat
Phase A

Before anything goes out

Roughly week one. Nothing is posted until these three are done.

1

Check the account

Half a day of looking, before touching anything.

  • Does the Page carry a country or age restriction? If it does, comment automation does not work at all, and everything else is pointless until it is lifted.
  • Any warnings on the Page or in ManyChat.
  • How many contacts are still reachable, and what delivery looks like now.
  • What is still switched on and quietly running.
  • Which ManyChat plan the account is on and what it is costing.
Needs account accessRank4AIDetail: the plan, Part one
2

Clean up and rebuild the automation

Two or three days. This is the work that makes everything after it reliable.

  • Screenshot the old numbers, then pause the old comment triggers so nothing fires by accident.
  • Set up the new reply: anyone who comments gets a private Messenger message with buttons. When they tap a button, we can keep messaging them for the next 24 hours.
  • In that same message, ask if they want updates. That yes is the only way to message them after the 24 hours, for example about the draw they entered.
  • Tag people by the button they tap, so the answers are usable later.
Ready to buildRank4AIDetail: the plan, Part seven
3

Settle the freebie question

A decision, not a build. Meta prohibits offering cash or cash equivalents in exchange for engagement, and "comment to win £100" is that.

  • Move to physical prizes, filmed on the desk, rather than cash.
  • Or take the entry off the comment: enter by messaging the Page, on the site, or live on the draw.
  • Or keep cash and accept a known risk, with the lowest-risk version of everything else.
Skye's decisionDetail: the week page
Phase B

Start posting

Weeks one and two. The Page starts working again.

4

Post the week

Seven days with a shape, instead of the same prompt every day.

  • Ask questions people would answer even if there was no prize attached.
  • Film the business: cars arriving, prizes being prepped, friendly office debates, winner calls.
  • Two freebies a week at most, never on consecutive days, never the same style.
  • Somebody works the comments for the first hour. Replies are what turn a wall of one-word answers into a conversation, and Facebook measures that.
Ready to runSkye and the teamThe 7-day planThe playbook
5

Make draw night a programme

The draws already run live. Right now, nobody watching ends up in Messenger.

  • A running order: prizes, the physical prize on camera, audience question, comments read out, the draw, the winner, the reaction, a tease for next week.
  • The trigger has to be set up before going live, because it cannot pick up earlier comments.
  • Cut the best moments into next week's posts.
Ready to runSkyeDetail: the playbook
Phase C

Prove it works

Weeks two to six. Replace every assumption with a number from this Page.

6

Build the baseline

Everything gets judged against this Page's own numbers, so those come first.

  • Run the same few post styles more than once, at the same times, so the numbers can be compared.
  • Judge on opt-ins per 100 comments, not comment count.
  • Separate first-time commenters from the regulars, or later posts look worse than they are.
  • Watch delivery rate weekly. It is the early warning for everything.
Ready to runRank4AIDetail: the plan, Part four
7

Connect the purchase data

The one piece of building work worth doing, and the thing that makes the rest possible.

  • Push every purchase back from the website into the messaging system: which draw, how many tickets, when, and whether a basket was abandoned.
  • Without it there is no "never played", "lapsed" or "almost bought", so none of the personal messages can run.
  • If it all stopped tomorrow, Palladium still owns that database.
Needs their developerDetail: the plan, Part three
8

Switch on the personal messages

Once purchase data is flowing, the messages stop being the same message to everybody.

  • Never played: how it works, is it legitimate, previous winners.
  • Played and lapsed: the win-back, aimed at a segment rather than blasted at everyone.
  • Almost bought: they clicked, landed on the site and did not finish.
  • A draw that is underselling, pushed to the right people before it crosses break-even.
After stage 7To be planned
Phase D

Spend money

Only once the free version is proven to work.

9

Boost, then buy conversations

Same posts, same flows, bigger audience. Two steps, in this order.

  • Boosted posts first. The comment automation works on boosted posts with no changes, so this proves it copes with more people.
  • Then ads that open Messenger. These skip the comment step entirely and are optimised for conversations started rather than engagement.
  • Aim first at people who have already visited or engaged, then at similar audiences, and only then at strangers, who cost the most.
  • Keep paid and free contacts separate, and judge each on what people went on to spend, not how many signed up.
Blocked: three things firstDetail: the plan, Part three

Blocked on: gambling and gaming authorisation for the ad account (Meta Business Suite, Authorisations and verifications), unique or trackable discount codes on the site, and tracking that connects a message click to a completed purchase. Without the codes and tracking, there is no way to tell what the money earned.

Phase E

Own the audience

Weeks five and six onwards.

10

Move retention to WhatsApp

Messenger brings people in. WhatsApp is where you keep them.

  • WhatsApp can send promotional messages at any time, using templates approved in advance. Messenger mostly cannot.
  • The sends a competition business needs are predictable: closing tonight, tickets running low, live draw starting, you have won.
  • It should be cheaper per message than the SMS they send now, but that gets costed against their real invoice, not an estimate.
  • The consent wording has to be right from the first opt-in in week one. A yes on Messenger does not carry over to WhatsApp, and going back for consent later is what usually kills the move.
Needs their SMS invoiceDetail: the plan, Part nine
Reference

The four documents

Everything lives here. Each one has a job.

This page

The stages, in order, and what each is waiting on.

The plan

The account as it stands, the two angles, what changed on both platforms, how the flows are built, and what we need from the client.

Open the plan →
The posting playbook

The editorial system: why people comment, the formats, what to film, the comment desk, fifty ready-made posts and the testing programme.

Open the playbook →
The 7-day plan

The client-facing week: day by day, what to film, example captions, the Messenger reply for each, and how to run freebies with less risk.

Open the week →
Open

What we need from Skye

The short list that unblocks the rest.

  1. Cash or non-cash freebies? Stage 3. This is the one with a policy attached.
  2. Which night is the real draw night? The week is built around Thursday as an assumption.
  3. Access to the Facebook Page and the ManyChat account, so stage 1 can happen.
  4. Can the site issue unique or trackable discount codes? Stage 9 depends on it.
  5. What tracking is on the site now, and who the developer is for stage 7.
  6. The SMS invoice and monthly volume, for the WhatsApp case in stage 10.
  7. The draw calendar: closing times, ticket prices, live draw times.
If only one thing happens this week

Stage 1. Check whether the Page carries an age or country restriction, because if it does, none of the automation works and everyone would spend a fortnight wondering why the posts are not producing anything.