The comment-to-Messenger system went live in summer 2025 and it worked: people who had just commented got 84.6% of messages delivered and opened almost all of them. This plan says what has changed since, what that means for the flows already built, and how to run it on the draw cycle rather than the contact list.
The whole thing in about two minutes.
Messages to people who had just commented delivered at 84.6%, with 100% opened and 99.6% clicked. Of the cold contacts messaged, 71% opened the first message. No other channel you run gets close.
Broadcasting to the stored list delivered at 49.2%, and the account carried a Meta warning that 100% of recent messages sent outside the 24-hour window were blocked. More than half of that broadcast never arrived.
You can only message someone freely for 24 hours after they contact you. A competition site has more reasons for people to get in touch than almost any business: freebie posts, draws opening, draws closing, live draws, winners. Every one of those reopens the window.
The 2025 plan used Sponsored Messages to reach people once the window closed. Meta withdrew them in 2024. The route still open to UK Pages is a Messenger List: people opt in to follow-up messages while their window is open, and can then be messaged again later. Every flow should ask for that opt-in.
It sends six messages over eight days after someone opts in. Everything after the first lands outside the 24-hour window. The copy is good; the timing has to be tied to things people do rather than days on a calendar.
Margin on a draw is ticket revenue minus the prize. The prize cost locks when the draw is announced but revenue does not, so every ticket above break-even is close to pure margin. That makes the draw-closing flow worth more than any other.
ManyChat already does the flows. The part worth building is the connection pushing every purchase from the website back into it. Without that there is no "never played", "lapsed" or "almost bought", so the personal messages from the 2025 plan cannot run.
| Area | 2025 plan | Now | What it means |
|---|---|---|---|
| Reaching people after 24 hours | Sponsored Messages at £1.50 to £3 per 1,000 | Gone Meta withdrew Sponsored Messages in 2024 | Messenger Lists instead: ask for the opt-in inside the window, then message later within Meta's frequency limits |
| Recurring notifications (Messenger Lists) | Not used | Still open in the UK Switched off in most countries on 10 February 2026, kept for the UK, EU, Australia, Japan and Korea | Now the main way to reach a contact whose window has closed. Needs a live test in the account before anything depends on it |
| Message tags | Available for updates outside the window | Mostly gone Account, post-purchase and event update tags withdrawn in 2026 | Only the human agent tag remains: a person, not an automation, can reply for up to 7 days |
| Replying to comments | Keyword reply straight into Messenger | Same, with limits One private reply per comment, within 7 days | The first message must carry a reply button. The window only opens when the person replies or taps, and an "open website" button does not count |
| Meta's paid Marketing Messages | Did not exist | Not for the UK The UK is not on the supported countries list | Not an option, however it gets pitched |
| ManyChat pricing | Plan in use in 2025 | Changed March 2026 Priced by active contacts a month | Around 5,000 active contacts sits in the Business tier, about $69 a month billed annually in US pricing. UK price to confirm |
| Welcome sequence | Six messages over eight days after opt-in | Breaks the window rule Messages two to six land outside the 24 hours | Rebuild so each message is triggered by something the person does (Part six) |
| Broadcasts | Whole-list broadcasts, then A/B/C tests | 49.2% delivered with a Meta warning | Broadcast only to people whose window is open, or to people who have opted in to marketing messages |
| What the system is for | Opt-ins and site visits (the flow switched focus twice in July and August) | Draw sell-through | The draw-closing flow becomes the main flow, because that is where margin sits (Part three) |
| Purchase data | Phase 2 of the agreement, pricing to be confirmed | Prerequisite | Moved to the start: none of the personal messages work without knowing who has played |
| Scope | Messenger, Instagram, WhatsApp and TikTok | Facebook only | This plan covers the Facebook Page and Messenger. Other channels are left out on purpose |
| Projections | 400,000 followers × 1.5% = 6,000 new customers a month, used as an example | Replaced | Numbers get rebuilt from the real figures requested in Part eleven, not from follower counts |
Screenshots from the account
Data window shown on the screenshots: 24 June to 8 July 2025.
| Flow | Sent | Delivered | Opened | Clicked |
|---|---|---|---|---|
| Broadcast to the whole list Ticket discount offer | 3,405 | 49.2% | 99.8% | 94.4% |
| Triggered by a comment Nightly freebie post | 298 | 84.6% | 100% | 99.6% |
The only difference between those two rows is that the second group had just commented, so their window was open. Messaging open windows delivers at around 85%. Blasting the list delivers at around 49% and earns a policy warning.
Even so, the broadcast still produced roughly 1,580 clicks from 3,405 names. The same list by email would have produced somewhere around 100. The channel is worth protecting.
Contacts going inactive at that rate is the strongest argument for the cycle over the list. A list that loses two thousand people a fortnight cannot be the thing the plan rests on.
| Segment | Sent | Opened | Clicked | Open rate | Clicked, of opens | Clicked, of sent |
|---|---|---|---|---|---|---|
| A | 3,152 | 1,430 | 1,168 | 45.4% | 81.6% | 37.1% |
| B | 2,979 | 806 | 720 | 27.1% | 89.4% | 24.2% |
| C | 3,138 | 608 | 518 | 19.4% | 85.1% | 16.5% |
| All | 9,269 | 2,849 | 2,424 | 30.8% | 85.0% | 26.2% |
Once someone opens, they click at 82% to 89% whichever segment they are in. The spread between segments is almost all in whether the message is opened, and on Messenger an unopened message is very often an undelivered one. The segment that did best is the one most likely to have had open windows. That is worth checking in the account before the next broadcast.
Mobile converted at about 20%. Desktop converted at under 1% while carrying 25% of traffic. If that has not been fixed, every flow that sends people to the site on desktop is leaking.
The opt-in funnel. The notes say 71% opened, 97% of those clicked and 98% of those opted in, "meaning almost 48%" opted in. Multiplied through, those rates give about 67%, not 48%. One of the four figures is wrong.
Open rates. The screenshots show 99.8% and 100% opened; the A/B/C test shows 30.8%. They are probably measured differently (of delivered against of sent), but it needs confirming before either is quoted.
Dates. The screenshots show a 3,405-message broadcast in the 24 June to 8 July window, but the change log has ManyChat set up on 30 June and the first broadcast on 30 July. Either the screenshot dates or the log need checking.
The margin on a competition is ticket revenue minus the cost of the prize. Illustratively: sell £100 of tickets against a £75 prize and the draw makes £25.
The prize cost is fixed the moment the draw is announced. The revenue is not.
| Sell-through | Ticket revenue | Prize cost | Margin |
|---|---|---|---|
| 100% | £100 | £75 | £25 |
| 85% | £85 | £75 | £10 |
| 75% | £75 | £75 | £0 |
| 60% | £60 | £75 | -£15 |
Illustrative ratios only. Palladium's real prize-to-ticket ratio is one of the figures requested in Part eleven.
A draw does not fail gently: it crosses from profitable to loss-making over a narrow band of sell-through. And every ticket sold above break-even is close to pure margin, because the prize does not get more expensive. Entries landing late in a draw that is most of the way to filling are worth several times those landing early.
The 2025 flows were built around opt-ins and site visits. On the economics, the draw-closing flow is the money flow: it pushes extra entries into draws already running, against a prize already paid for, at the point where each ticket is worth most.
The "Paradise isn't getting the love it deserves" message from the 2025 plan is exactly this, and it should be one of the first things switched on.
| Trigger | Works? | Notes |
|---|---|---|
| Comment on a Page post | Reliable | Keyword or any comment, on a chosen post or any post. Published, scheduled and boosted posts. Not in Groups, not on reposts, and comments made before the flow was switched on are not picked up. |
| Comment during a Facebook Live | Yes | The draws already run live on Facebook |
| Comment on an ad | Yes, with limits | Only relevant if ads are in scope |
| Messenger keyword | Reliable | Any incoming message containing a word |
| Conversation starters | Yes | Up to four tappable questions, shown only to people who have never messaged the Page |
| Default reply | Reliable | The FAQ catch-all, already built in 2025 |
| m.me links, ref links and QR codes | Reliable | Can carry which draw, post or email they came from |
| Ads that open Messenger | Yes, with limits | Only if ads are in scope |
| Facebook Story reply or reaction | No trigger | Whether a Story reply lands as an ordinary message the default reply would catch is untested |
| Page follow or like | No | |
| Reactions or likes on posts | No | A reaction gives no route to message someone |
| Shares | No | And rewarding shares breaks Meta's promotion rules (Part ten) |
| Website chat plugin | Gone | Retired in 2024. m.me links do the same job |
Meta allows one private message per comment, sent within 7 days of the comment. That message does not open the 24-hour window on its own. The window opens only when the person replies or taps a reply button, and an "open website" button does not count.
The 2025 "Yes, notify me next time!" button did this correctly. Every trigger flow keeps that step.
| Route | UK Page? | Notes |
|---|---|---|
| Messaging inside the 24 hours | Yes | Any message, promotional included, after they message, tap or react |
| Messenger Lists (recurring notifications) | Yes | The UK is one of the countries Meta kept when it switched these off elsewhere on 10 February 2026. The opt-in has to be asked for inside the window. Frequency is limited. Test in the account first |
| Human agent tag | Manual only | A person replying by hand, up to 7 days after the customer's last message. No automated messages |
| Broadcast channel on the Page | Unconfirmed | Meta offers these to larger Pages. UK availability not confirmed, and ManyChat cannot post to one |
| Meta's paid Marketing Messages | No | The UK is not a supported country |
| Sponsored Messages | No | Withdrawn in 2024 |
| Update message tags | No | Withdrawn in 2026 |
Three things get tested in the account in week one before any flow depends on them: Messenger List opt-in and sending from a UK Page, whether a Story reply reaches ManyChat, and whether a broadcast channel is available.
| Built in 2025 | Verdict | Why |
|---|---|---|
| Freebie keyword flows ("comment your favourite colour") | Keep | This is the flow that delivered 84.6%. The nightly freebie is the cheapest engine for reopening the window every day. |
| Default reply and FAQ answers | Keep, review | Answers still need checking against how the site works now (wallet, instant wins, claims). |
| Secret competition link flow | Keep | Gives people a reason to message, which opens the window. |
| Customer tagging and segments | Keep | Becomes the base the purchase data attaches to. |
| "Welcome from Skye" sequence (six messages, eight days) | Rebuild | Good copy, wrong timing. Message one stays as the opt-in reply. The rest become replies to things people do (Part six). |
| Whole-list broadcasts | Stop as the main tactic | 49.2% delivered and a policy warning. Open a window first, then send. |
| Sponsored Messages for re-engagement | Not available | Withdrawn by Meta in 2024. Replaced by a Messenger List opt-in built into every flow. |
| SKYE10 code, charity page and "nominate a cause" buttons | Check | These were marked "needs set up on site" in 2025. Confirm they exist before any flow links to them. |
| Facebook Group | Page posts only | Automation runs on Page posts, not in a Group. The Group is where the community is, so trigger posts go on the Page and get shared into the Group. |
Check the account for any live Meta or ManyChat warnings, the delivery rate over the last 90 days, and which flows are still running. If the eight-day sequence or scheduled broadcasts have been running all year, they may have been sending into closed windows the whole time.
| # | Trigger | What ManyChat does |
|---|---|---|
| 1 | Comment on a freebie or draw post | Public reply, private reply with a button, tap opens the window, offers the Messenger List opt-in, tags the contact |
| 2 | Messenger List message, plus a "last chance" post | Closing alert to people who opted in, comment keyword for everyone else, tags closing responders |
| 3 | Comment during a Live draw | Confirms, teases the next draw, offers the opt-in while the whole audience is watching |
| 4 | Messenger keyword, conversation starters, default reply | Answers instantly, hands to a person when there is no match, tags by question type |
| 5 | Winner m.me link | Thanks the winner and asks for a short video |
| X | All of the above | Tags every contact by source, draw and keyword, and sends it to Palladium's own database |
Trigger: comment on a Page post
The flow that delivered 84.6% in 2025, kept as it is. The nightly freebie gives people a reason to comment every day, which keeps windows opening every day rather than weekly.
The one addition: every flow ends by offering the Messenger List opt-in ("want a heads-up before this one closes?"), so the person can still be reached after the 24 hours.
Trigger: Messenger List message and a comment post · the money flow
Facebook has no Story trigger, so this runs in two halves. People who opted in to closing alerts get a Messenger List message. Everyone else gets a "last chance" Page post with a comment keyword, which reopens their window.
On the economics this is the most valuable flow in the plan. It drives entries into a draw already running, against a prize already paid for.
Trigger: comment during a Live
The draws already run live on Facebook, so the moment exists and is unused. Comments trigger a message confirming and teasing the next draw. It is the best moment to collect Messenger List opt-ins, because the most engaged players are all there at once.
Trigger: keyword, conversation starters, default reply
The default reply and FAQ answers built in 2025 are the base. Add conversation starters for the four questions asked most. When there is no match, a person picks it up, and has 7 days to reply by hand.
Someone typing "how does this actually work" has already decided they are interested. Instant, complete answers at 11pm on a Sunday are a sales mechanism, not admin.
Trigger: m.me link sent with the win notification
Shares cannot trigger a flow, and rewarding them breaks Meta's rules. Instead, the win email or text carries an m.me link that opens Messenger. The flow thanks the winner and asks for a short video. The moment someone has just won is when they are most likely to say yes.
These fire on what someone does or fails to do once they are in the system. None of them are possible without the purchase data connection in Part eight.
Cheapest improvement in the build
Someone taps the button and then nothing happens. You have 24 hours and most setups use it once.
| Timing | Message |
|---|---|
| +2 hours | Soft check, did the link work |
| +8 hours | Social proof, last winner, tickets remaining |
| +20 hours | Last call before the window shuts |
This is where the best of the "Welcome from Skye" copy goes: the "is this for real?" explanation and the winners feed fit the +8 hour slot.
Trust and mechanics
In the contact list, never bought. The message is how it works, is it legitimate, previous winners, what happens if you win, and the charity story. SKYE10 belongs here, if it exists on the site.
Frequency and basket size
Bundle offers, new draw alerts, their favourite prize category. The 2025 themed-code approach ("you have been chosen", two bundle prices with named codes) fits here.
Win-back, personal
"I'm feeling friendly today. Giving a few lovely ladies, yes that includes you, 10% off to come back and play." Aimed at a segment, never broadcast to everyone.
Highest-converting segment
Clicked through, landed on the site, did not complete. Needs the website to send the abandoned basket back into ManyChat. It was in the 2025 agreement as "cart abandonment notifications into messaging platforms" and was never built.
Direct margin protection
"Paradise isn't getting the love it deserves so I'm doing 10% off just for you. Don't tell the others." A draw that is underselling gets pushed to a chosen segment before it crosses break-even.
Cheaper than win-back
2,019 contacts went inactive in one fortnight. A flow that fires before someone lapses costs far less than recovering them afterwards.
Retention starts at purchase one
A first-time buyer gets a short sequence aimed at a second purchase. A purchase does not open the Messenger window by itself, so the first message has to invite a reply ("tap to get your ticket numbers here") that does.
Segments B, D and G are by definition people who have not been in touch recently, which is exactly the 49.2% case.
So none of these can be a straight broadcast. Each reaches people one of two ways: through a Messenger List opt-in collected earlier, or a window-opener first, such as a comment prompt aimed at that segment, then the real message inside the open window at around 85% rather than 49%.
Before sending anything
Warnings on the account, delivery rate for the last 90 days, how many contacts are still active, which flows and sequences are still running, and what the ManyChat plan costs now. Switch off anything sending outside the window.
Prove the numbers again
One freebie post, the keyword flow, the opt-in button, the in-window follow-up. Measure five things: comments, delivery rate, open rate, opt-in rate and click-through to the site. That gives a clean 2026 baseline against the 2025 figures.
The actual product
Build the trigger flows around the rhythm the business already has, with the draw-closing flow first. Rebuild the welcome sequence. Connect purchase data from the website into ManyChat.
Where the 2025 ideas finally run
With purchase data flowing back, the Layer two flows switch on: never played, active, lapsed, almost bought, draw rescue, going quiet, post-purchase.
| Option | Verdict | Why |
|---|---|---|
| ManyChat, already in place | Keep | It absorbs Meta's changes for you. Now priced by active contacts: around 5,000 is the Business tier, about $69 a month billed annually in US pricing. UK price to confirm |
| Purchase data connection, website to ManyChat | Build this, phase 2 | Small piece of work, and a prerequisite for every personal message |
| Fully custom Messenger integration | No | App review, ongoing maintenance, and you inherit every change Meta ships |
| AI assistant answering questions | Phase 3 | In the 2025 agreement. Worth it once the FAQ answers are right, because "is this legit, how does it work" is what stands between a sceptic and a first entry |
What is worth building is the connection pushing every purchase between the website and ManyChat: which draw, how many tickets, when, and whether a basket was abandoned. ManyChat only sees Facebook behaviour, so without purchase data there is no segmentation, and the personal messages that made the 2025 plan different cannot run.
If Klaviyo is still in use, it is the obvious place for that data to meet email, which the 2025 agreement already named.
Not available. Withdrawn by Meta in 2024. Any price per thousand messages describes a channel that no longer exists.
Not available to UK businesses. The UK is not on Meta's supported countries list.
Against Meta's rules. Comment to enter is fine.
Not possible. None of them give a route to message someone.
Delivered 49.2% and triggered a warning. Broadcasts have a place, but only into open windows or to people who have opted in to marketing messages.
Not supported. Trigger posts go on the Page.
Not credible. The 2025 "6,000 new customers a month" was an example built on an assumed 1.5%. The numbers get rebuilt from real figures.
Out of scope for this plan. Facebook first, and only once the Messenger system is running cleanly.
| Figure | Their number | What it lets us work out |
|---|---|---|
| Cost per acquisition | If a Messenger-acquired customer costs less than the current CPA, that is the business case in one line. | |
| Average order value | What a click out of Messenger is worth, and which bundle the offer should push. | |
| Spend over three months | The value of a recovered lapsed player. | |
| Length of retention | Where players drop off, so the going-quiet flow fires before that point. | |
| Margin on a competition | What one extra late entry is worth. Turns the draw-closing flow into a pound figure. | |
| New customers per month | The base this is measured against. |
Run the account check, switch off anything sending outside the window, then relaunch on the next freebie with the in-window follow-up. It costs almost nothing and gives a clean 2026 baseline against 2025.