Your reliability score
Every Onflow Ads account carries one reliability score — a number from 0 to 10 that every engine on the platform reads. It is how the other side of a deal knows they can trust you, and it decides what you are allowed to do, how far your ads reach, and whether you keep the veteran rate on the marketplace fee.
It is also a policy, not just a feature: Terms §15 sets out how it moves, what it gates, what "Under Review" means, how an appeal works, and what happens to it when you close your account. This page is that section in plain language, alongside the numbers actually shipped.
You can see your own score, your full history and everything on this page live at onflowads.com/telegram/dashboard/reliability (sign-in required).
Why it exists
Two strangers agree a deal, one pays first and the other delivers afterwards. Nothing about that arrangement is safe unless conduct has consequences that carry forward. The reliability score is that memory: it makes a good record worth something, and it makes a bad one expensive — without needing anyone to read reviews, ask around or take a risk on a promise.
The basics
- Everyone starts at a neutral 8.0 — trusted from day one, with room to climb and room to slip. The reliability page says so: "Everyone starts at a neutral 8.0 — trusted from day one, with room to climb and room to slip."
- There is exactly one score per account. Buying promotions, selling them, cross-promoting, running Boost orders, hosting in the Subscriber Exchange — it is all the same number, moved by the same rules. What the platform observes in your channel moves that same number too — whether it was the Telegram bot that saw it or the reader that checks a channel's public history — and there is no second record to dodge into.
- Only the website moves the number. The Telegram bot never prices anything. When it observes a fact — a placement pulled early, a cross-post taken down, a member who left a partner's channel, a placement that served its full term — it records what it saw into a queue the website owns, and the website prices that fact from its own published table and writes the ledger entry, usually within about a minute (with one deliberate exception: an early unpin of a paid pin is priced immediately but recorded only at the end of the campaign, because until then you can still undo it — see the curable mark). A bot message about a penalty therefore says what was recorded and that it is applied on onflowads.com; it never says "your score is now N".
- Every change is written to a ledger you can read on your reliability page, with a reason and its own
OFRE-reference in the platform's records. Nothing moves your score without a line you can see. - Penalties fire only on confirmed outcomes — a real non-delivery, an upheld fraud concern, a post caught removed early. Never on an accusation alone.
- Removing an agreed post early is priced the same everywhere. Whether the post was a Paid Promotion, a Cross-Promotion or a Subscriber Exchange placement, the same check watches it for its full term and the same fixed penalty applies. Honouring a deal is worth the same everywhere, so breaking one costs the same everywhere.
- Each fact is priced once. A restarted bot, two monitors racing on the same delivery, or a queued entry replayed after a database hiccup all collapse into one ledger entry, because every entry names the thing it is about and the ledger refuses a second entry for the same thing. A confirmation followed by a delayed release credits a clean campaign once, not twice.
- Nothing you can buy raises it. Plans change your limits and your tools; only conduct changes your score.
Your score measures your conduct, not the platform's uptime. If a scheduled ad could not post because Telegram was down or an Onflow Ads system failed, your score is untouched and the buyer is made whole with a refund or a reschedule. A penalty only fires when the failure was genuinely yours — for example you removed the bot from your channel, revoked its posting rights, or deleted the post early. When the cause cannot be determined, no penalty is applied and the case goes to a human reviewer. Ambiguity resolves in your favour, never against you.
Reading the page
Everything on this page is shown live on your own reliability page, headed Your reliability score. It opens on the gauge with your number in it, your tier badge underneath, and the three gate chips that tell you at a glance which marketplaces you currently clear.
Plus plan · Advertiser. The top of your reliability page. Read the gate chips first: a tick means your current score clears that product's floor, a cross means that marketplace is locked until you recover. The tables of exact penalties, the healing rates and Your ledger are all further down the same page.
The page has five parts, top to bottom:
- The gauge and tier badge — your score out of 10 and the tier it falls in.
- The gate chips — Paid Promotions · needs 4+, Cross-Promotion · needs 5+ and Subscriber Exchange · needs 4+, each ticked or crossed for your current score. The number on each chip is the live floor an operator can retune, and the tick is measured against that same number — so the chip and the block message you would get always agree.
- The floors and tiers — what each band means and where each product stops you.
- The tables of exactly what moves it — the penalties and the credits, with their amounts, under the headings Exactly how it drops and How it climbs. The two developer-API marks below are newer than that table and are not on it yet, so this page is the fuller list.
- Your ledger — your own history at the bottom of the page.
If your account is under review, the badge shows Under Review in place of your tier — see Account standing.
The tiers
Your tier is read straight off the score:
| Range | Tier | What it means |
|---|---|---|
| 9.0–10 | Trusted | Top ranking, full reach, extra AI text headroom and the pro AI model one tier early. Held, not farmed — see below |
| 7.5–8.9 | Reliable | A strong, dependable record. Full reach, full AI |
| 5.0–7.4 | Fair | Good standing — cross-promotion stays open. Reach at ×0.75, full AI including images |
| 4.0–4.9 | Building | Paid promotions open; cross-promotion locked. Reach at ×0.5, AI text throttled about 50%, images off |
| 0–3.9 | Low | Marketplaces locked. Reach at ×0.3, AI text at the Free baseline, images off. Rebuild before starting new deals |
The tier is a label, not a separate setting — it always agrees with the number, and it changes the moment the number crosses a band edge.
Where the floor stops you
Each product has a minimum score to start something new. Below the floor, that surface locks until you recover — but anything already in flight always finishes. The floors are shown live on your reliability page and inside any block message; currently:
| Product | Minimum score |
|---|---|
| Paid Promotions — booking and accepting, both sides of the deal | 4.0 |
| Cross-Promotion | 5.0 — the community bar is higher because no money sits in escrow, so trust carries the whole deal |
| Subscriber Exchange (hosting) | 4.0 |
Terms §15.2 counts more than the obvious as already running: continuing an existing partnership is in flight, and so is a recurring arrangement you agreed with that same partner, and the rounds that cadence sets up — because they agreed to a series, and quietly dropping your side of it would be the same unfairness a floor is meant to prevent. What locks is opening something new by hand: a new campaign, a new listing, a new offer, a new application. Where standing has fallen far enough, the platform may still pause or end a recurring arrangement — and it tells both sides when it does.
While your account is Under Review, you cannot start a new deal on any surface regardless of your score. Full detail on every state: Account standing.
If a booking request reaches a channel owner who is under review or below the Paid Promotions floor, the platform declines it for them at once — the advertiser is refunded in full and the decline is recorded as owner gated. No Ghosted a request penalty follows, because the owner was never in a position to answer. The ghost penalty is reserved for an owner who could have answered and let the offer expire.
Exactly what drops it
Every penalty has a fixed size and fires only on a confirmed outcome. These are the amounts the platform applies, and nearly all of them are printed on your reliability page as well — the two developer-API rows below are the newest and have not reached that in-product table yet. Where the two ever differ, the amount recorded on the ledger entry is the one that was applied to you.
Delivery — you accepted a deal and didn't keep it
| Change | Event |
|---|---|
| −0.30 | Ghosted a request — an owner let an offer sit until it auto-declined, never responding |
| −1.00 | Didn't post in the window — accepted a promotion but never posted it within the agreed time |
| −1.50 | Post removed early — the ad was taken down, or the channel put out of reach, before its promised duration was up. On a booking sold as a pin under the older format wording, being unpinned counts here too |
| −0.75 | Paid pin taken down early — the advertiser separately paid for the pin, and the campaign ended with the ad still unpinned. The pin charge is collected from you as soon as the unpin is confirmed — see Owing money — but this mark is not recorded until the run finishes, and re-pinning the post before then means it is never recorded at all. The ad itself is still standing, which is why this sits between the two rows either side of it |
| −0.40 | Posted over a top-of-feed ad — a format that promised hours at the top was sold, and the channel published something else inside that window. Owner only, no money effect |
Taking the ad down, unpinning a pin that was paid for, and posting over the ad are separate failures with separate marks, and the gaps between them are deliberate.
- −1.50 is for the ad no longer being there — deleted, or the channel put out of reach — during the duration that was paid for. It also holds the payout in escrow.
- −0.75 is for a paid pin that came down while the ad stayed up and was never put back. The advertiser keeps the substance of what they bought and is returned the pin charge in full, so the residual harm is smaller than a deletion but larger than a breach of the top-of-feed window, because money moves. It is the one mark on this page you can undo by your own action: see The one mark you can undo below.
- −0.40 is for the ad still being there but not being first — the owner published something else inside the hours the format sold as time at the top. It moves no money at all: the payout is unaffected and the placement runs its full duration.
They are checked independently, so an owner who breaks more than one promise on one placement takes each mark. How each is checked: Fraud and bot protection.
The one mark you can undo
−0.75 for an early unpin is the only mark on this page that is not recorded when we catch it. Every other row here is a finished fact by the time we see it — an ad that was deleted is deleted; a window that was breached has passed. An unpinned ad is different: the campaign is still running, and whether the pin was kept is still an open question until it ends.
So the mark is held, not applied:
- We confirm the unpin on two consecutive readings, about five minutes apart.
- The money moves straight away. The pin charge comes out of your payout at that point and it does not come back, whatever you do next. You were paid to keep a pin and there was a stretch of the campaign where it was not there.
- Your score does not move. Nothing is written to your ledger.
- Re-pin the post and leave it pinned to the end of the run, and no mark is ever recorded. A re-pin is confirmed on two consecutive readings, exactly like the unpin — you cannot flick it back on for one check.
- If the campaign ends with the ad still unpinned, the mark is recorded then.
- Unpin it a second time after re-pinning and the mark is back on the table, with no further charge — the pin was already returned in full.
The notification you get when we catch it says all of this, and links straight to the booking.
Fraud and disputes
| Change | Event |
|---|---|
| −2.00 | Fraud concern upheld — a concern raised against you was investigated and upheld; the heaviest single mark |
| −0.50 | Raised a false fraud concern — you reported fraud and the claim did not hold up |
| −0.50 | Weaponising disputes — a repeated pattern of rejected fraud concerns (a third rejection in 90 days, applied at most once per 30 days), not a one-off |
| 0.00 | Concern dismissed — inconclusive — the reviewer could not establish fault either way. Nobody's score moves; an audit line is written on the owner's ledger so the outcome is visible, but it changes neither the number nor the event count |
| −1.00 | Campaign stopped by our team — an admin stopped a campaign because of your conduct in it: a report against you was upheld, or a breach was found on review. A person looked, so it is never a check's guess — but it is one broken run, not a fraud finding |
Cross-Promotion
| Change | Event |
|---|---|
| −0.30 | Left a CP slot unconfirmed — you attached to a campaign and never confirmed within 48 hours. Once per campaign, on top of the join strike and cooldown the campaign itself applies |
| −1.50 | CP post removed early — took the cross-post down before the agreed retention window |
| −1.00 | Left the partner channel early — you left (or left and rejoined) your partner's channel before the cross-promotion ended, while their post was still up. Lighter than removing the post, because the partner's ad was still being read |
Subscriber Exchange
| Change | Event |
|---|---|
| −1.50 | SubX post removed early — a hosted placement, or your half of a swap, was pulled before its agreed hold ran out |
Boost and referrals
| Change | Event |
|---|---|
| −1.00 | Boost not delivered — reserved in the table for a Boost order the seller owned that failed and was refunded for cause. Boost orders are fulfilled by a provider rather than a member, so nothing fires it automatically today; the row exists so a ledger entry written under it can label itself |
| −2.00 | Chargeback — a confirmed fraudulent chargeback or payment reversal |
| −2.00 | Referral abuse — confirmed self-referral or fake-signup abuse |
AI misuse — evidence-backed and confirmed by a human reviewer. Automated systems never dock a score on their own.
| Change | Event |
|---|---|
| −1.00 | AI misuse — AI-made copy or creative used in an upheld scam or spam campaign |
| −2.00 | AI abuse — confirmed prompt-injection, jailbreak or banned-content attempts against the AI tools |
The developer API — the same pair, one surface further out. A key is only another way to act on the platform, so misusing one moves this same score rather than a private counter nobody else can read. Priced level with the AI pair on purpose: scripting fraud through an integration is not a lesser wrong than typing it.
| Change | Event |
|---|---|
| −1.00 | An abusive pattern of automation — you ran an abusive pattern of automation through the developer API, upheld on review |
| −2.00 | Serious misuse of the developer API — sharing or reselling a key, or scripting fraud through it, upheld on review |
Nothing automatic docks your score here. The developer-API desk proposes a flag into the same review queue the AI content classifier files into, and the score moves only when a person confirms it — the same rule the AI pair above follows. A proposal that is rejected never touches your score at all, and a rejected one leaves nothing behind.
Because these two are the newest entries in the table, your reliability page has no short label for them yet. When one is applied it appears in your ledger as the sentence it was recorded with rather than as a two-word label — see Reading your ledger — and it carries an Appeal button like every other penalty. Blocking a key is a different act with a different remedy: it stops that key working, it is not a mark on you, and it never touches your score.
Cancelling a booking before it goes live — before or after the owner accepted it — is a full refund with no fee and no reliability effect on either side. That is true whichever party cancels. Earlier versions of these pages listed a −0.30 advertiser penalty for cancelling an accepted booking; it is not in the platform's table and nothing applies it. Once a placement is live, it can no longer be cancelled at all — a placement that then runs badly is a fraud concern, not a cancellation.
A cross-promotion running with several partners is checked partner by partner, so a single act that breaks all of them at the same moment — removing the bot from your channel, or taking away its admin rights — is detected once per partner and marked once per partner. On a campaign with ten partners that would be ten early-removal penalties for one act, which is enough on its own to take a score to the bottom of its range.
Terms §15.1 binds the platform against that: where one cause broke several placements in the same campaign at the same time, that is one incident carrying one penalty. Appeal any one of those entries and every mark beyond the first is reversed in full. What your partners are owed is settled separately and is not reduced because your score was corrected.
Letting a booking offer expire without answering it inside the accept window is applied from the clock alone. The clock cannot tell why you were away, and it does not stop when the platform is down — background checks keep running through a maintenance window, including one opened on the very product you needed. If the Services, or the part of them you needed, were unavailable to you for a material part of that window, appeal it and it will be reversed. On this one class of penalty, you have to tell them.
These automatic penalties are the basic level, not the limit. For deliberate fraud, scams, chargeback abuse or coordinated manipulation, Onflow Ads reserves the right to permanently freeze payouts and the account, to delete the account and its data with no payout or refund, and in extreme cases to pursue legal action. See Trust and safety. None of this touches you if you deal in good faith.
The one discretionary movement
Not every abuse arrives as a filed concern. Where a senior reviewer sees clear abuse that nobody complained about, they can move a score by hand, up or down. Terms §15.1 publishes that power rather than letting you find it on your ledger, and it publishes the limits with it:
- A single discretionary adjustment is capped at ±2.00 points.
- Only a Master Admin can make one.
- A written reason is mandatory and is stored on the entry you can read.
- You are notified when one is applied.
- How often any one reviewer can make one is itself rate-limited — at most 10 an hour per reviewer.
- It can be appealed exactly like any other penalty.
It is not a route around the published table — it is the case the table does not reach. On your ledger it appears as Admin adjustment, with its reason attached.
What raises it
The everyday way up is simple: finish deals cleanly.
| Change | Event |
|---|---|
| +0.20 | Clean delivery (owner) — you delivered and the promotion completed cleanly, start to finish |
| +0.10 | Clean campaign (advertiser) — a campaign you booked ran clean and completed; the buyer earns trust too |
| +0.05 | SubX hosting completed — you carried a Subscriber Exchange placement, or your half of a swap, for its full term |
| +0.05 | SubX placement ran clean — your own exchange placement ran end to end |
Two honesty rules apply to the way up:
- Trusted is held, not farmed. Above 9.0, each clean completion is worth steadily less as your score approaches 10, so nobody can pump a permanent 10 with a churn of cheap deals.
- Self-dealing counts for nothing. Buying your own listing earns neither side any score — including two accounts that share one Telegram identity.
Freezes, holds and other lines that move nothing
Some ledger entries record a state, not a score change, so the history stays complete:
| Label | What it records |
|---|---|
| Score frozen for review | A reviewer placed a manual hold on the account — between 1 hour and 30 days, 72 hours by default — with a reason. While it stands you show as Under Review |
| Review hold lifted | A reviewer cleared a hold early |
| Concern dismissed — inconclusive | A fraud concern against you ended without a finding either way. Nothing was taken |
| Accounts merged — lower score kept | Two accounts of yours were joined (for example a starter account folded into your website account). The merged account keeps the lower of the two scores and the whole history of both, and the entry shows the difference so nothing is hidden |
| Carried over from a removed account | A below-baseline record resumed on a re-registered identity — see the rules below |
| Recovered ledger entry | A score move that could not be written at the moment it happened was queued and written later, with its original intent intact |
The freeze that a fraud concern itself places is not a ledger entry of its own: it is the hold described in Account standing, lasting the full 14-day concern window and lifting the moment the concern is decided — upheld, rejected or dismissed.
Three ways to reverse a drop
A penalty is never permanent. The first two routes below can run at the same time.
1. Earn it back
The fastest route, and the only one that works above 8.0. Every clean deal adds to your score right away — +0.20 as an owner, +0.10 as an advertiser. A single upheld-fraud mark takes real work to undo, but delivery after delivery, it comes back.
2. Let a clean streak heal it
Even with no new deals, a below-baseline score heals itself automatically — at most once a day — as long as you stay clean. You have to go quiet long enough first, and how fast it then heals depends on the worst unforgiven mark on your record in the last 120 days. These are the figures published on your reliability page:
| Worst recent mark | Clean streak first | Then heals |
|---|---|---|
| Under 0.50 — a ghost, an unconfirmed CP slot, the exclusive-window −0.40 | 14 days | +0.10 per day |
| 0.50 to 0.99 — a false dispute, a paid pin taken down | 21 days | about +0.067 per day |
| 1.00 to 1.99 — non-delivery, an early removal, leaving a partner channel | 30 days | +0.05 per day |
| 2.00 — an upheld fraud, a chargeback, referral abuse, AI abuse | 45 days | +0.025 per day |
Limits that keep healing honest:
- Healing never crosses 8.0 on its own — anything above neutral has to be earned with real deals.
- After an upheld fraud the ceiling is held at 6.0, easing back toward 8.0 only over the following 90 days.
- A frozen account does not heal at all until it clears, and neither does an account with a fraud concern still open against it — even after the timed hold has lapsed. Time spent under review is not credited afterwards.
- A new penalty resets the clean streak — the quiet window starts again.
- When many accounts are waiting to heal, the platform takes turns by who was healed least recently, not by lowest score, so nobody is starved of their daily step.
3. Appeal a penalty you think is wrong
Every standing penalty on your ledger has an Appeal button, and a human reviews every appeal. Terms §15.4 and Refunds §24 set a window of 30 days from the day the penalty was applied; the button itself does not count days — the reviewer applies the window. You can file at most three appeals a day, because every one lands in a human queue. If a mark is overturned, the exact points are restored, untapered — a genuine restore can even carry you back above the recovery ceiling, because it returns standing you had already earned. A forgiven mark also stops slowing your clean-streak healing. Full walkthrough: Appeals and fraud concerns.
An overturned mark returns your points, returns a good-faith deposit the same mark forfeited, and restores your veteran streak — the 100 clean campaigns (or 100 clean placements) behind the veteran rate on the marketplace fee — because a forgiven mark does not interrupt the count. It does not refund the difference between the standard rate and the veteran rate on fees already charged while the mark stood, and it does not reinstate allowances that were throttled in the meantime. Refunds §24 states this outright: a fee charged at the standard rate because of your standing is not a fee levied in error. A reviewer can also restore only part of a penalty — every partial restoration, a later full one and an undo all point at the same original entry, and together they never hand back more than the penalty took. A mark counts as reversed exactly when the whole amount has come back. If the platform cannot write the reversal at that moment, the reviewer is told and nothing changes; the intent is queued and written later, still linked to the original entry.
What a low score costs beyond the gates
Even above every floor, the score keeps working quietly — it is a dial, not just pass or fail:
| Area | Effect |
|---|---|
| Marketplace ranking | A higher score lifts your listings up the marketplace automatically |
| Ad reach | Below 7.5 your ads reach a fraction of the full drop count: ×0.75 at Fair, ×0.5 at Building, ×0.3 below the floor. You still run — just narrower — so you can earn your way back rather than being frozen out |
| Marketplace fee | A reliability penalty restarts your veteran streak — the 100 clean campaigns on one channel (or 100 clean placements as an advertiser) that earn the veteran rate — so the account pays the standard rate on every new booking until the streak is rebuilt; the standard rate is 9% and the veteran rate is 7%. A penalty overturned on appeal restores the streak by itself |
| AI tools | Full at Fair and up; text throttled about 50% and images off at Building; text at the Free baseline below 4.0; AI off entirely while under review. Trusted earns extra text headroom and early access to the pro model |
| The developer API | Your key's requests-per-minute rate is banded on the same thresholds as the AI allowance above: refused outright while you are under review; clamped below 4.0 to the lowest rate any plan sells the API at, which is 60 a minute today; halved and rounded up at 4.0 to 4.9; exactly your plan's rate from 5.0 to 8.9; a fifth more at 9.0 and above, and never less than one extra call |
| Paid Promotions suspension | Delivery faults only — not posting, a removed post, a broken pin, a breached window, a ghosted request, a removed cross-post, leaving a partner channel, an early exchange removal — over the last 30 days, ignoring anything since reversed, decide whether the bot suspends you from Paid Promotions delivery. A carried-over record, a rejected concern or a disputes pattern never counts toward that |
Alongside those bands, API access has a minimum score of its own, in the same operator-tunable family as the three marketplace floors above. It ships at 0, which means off: no key is refused today for the score behind it. Unlike the marketplace floors it has no shipped predecessor to inherit, so any other starting value would have taken every integration below it dark on deploy day, mid-call, with nobody having asked for that — turning it on has to be a deliberate act. The bands still apply either way, because they narrow a rate rather than close a door.
The rate is worked out fresh on every call, from your live plan and your live score, so a score that recovers speeds your integration back up on its own with nothing to reissue. Your plan's own keys-and-rate figures are on The Boost API.
Reading your ledger
The bottom of your reliability page lists your 50 most recent score changes — the amount, a plain-language label, and beneath them the date and the score you were left on. It is your recent history, not your entire one. You can reload it up to 60 times a minute.
Labels you will see on entries:
| Label | Meaning |
|---|---|
| Clean delivery | You delivered a promotion cleanly as the owner |
| Clean campaign | A campaign you booked as the advertiser ran clean |
| Ignored a request | An offer sat unanswered until it auto-declined |
| Did not post in the window | Accepted a promotion but never posted it in time |
| Post removed early | The delivery monitor caught the ad taken down before its promised duration was up |
| Paid pin taken down early | A pin the advertiser paid for came down before the campaign ended; the pin was refunded to them |
| Posted over a top-of-feed ad | Something else was published in your channel inside an ad's exclusive window |
| Fraud concern upheld | A concern raised against you was investigated and upheld |
| Fraud concern rejected | A concern you raised did not stand |
| Repeated rejected concerns | A pattern of concerns that did not stand |
| Concern dismissed — inconclusive | A concern against you ended with no finding; nothing was taken |
| Left a CP slot unconfirmed | You attached to a cross-promotion and never confirmed your side |
| CP post removed early | You took a cross-post down before the agreed retention window |
| Left the partner channel early | You left your partner's channel before the cross-promotion ended |
| SubX post removed early | A Subscriber Exchange placement came down before its term |
| SubX hosting completed / SubX placement ran clean | A Subscriber Exchange run that finished cleanly |
| Boost not delivered | A Boost order you owned failed and was refunded for cause (reserved — nothing applies it automatically) |
| Chargeback | A confirmed fraudulent payment reversal |
| Referral abuse | Confirmed self-referral or fake-signup abuse |
| AI misuse / AI abuse | Human-confirmed misuse of the AI tools |
| Clean-streak recovery | Automatic healing after a quiet period |
| Restored — not at fault | An appeal was overturned and the exact points returned |
| Forgiven | An appeal was accepted as an honest mistake and the points returned |
| Admin adjustment | A manual correction by the Onflow Ads team, with its reason attached |
| Score frozen for review / Review hold lifted | A manual hold was placed, or cleared, by a reviewer |
| Accounts merged — lower score kept | Two of your accounts were joined and the lower score carried |
| Carried over from a removed account | A below-baseline record resumed on a re-registered identity |
| Recovered ledger entry | A queued score move written after the fact, with its original intent |
An entry the page has no short label for prints the reason it was recorded with instead — a full sentence naming what happened. That is the record, not a fault, and it carries an Appeal button like any other penalty.
Tags on an entry:
- Reversed — the penalty was overturned or forgiven and its points returned
- Appeal under review — you contested it and the team is reviewing
- Appeal declined — a reviewer looked at your appeal and the penalty stands; each entry is appealed once
- An Appeal button — available on any standing penalty you have not contested yet
Each entry has its own OFRE- reference in the platform's records, and that is what identifies one specific mark if the team quotes one to you or you write in by another route. The row on your page shows the amount, the label, the date and the score after — not the reference. You never need it to appeal, because appealing from the entry already says which one you mean.
If you have no activity, the ledger reads: "No activity yet — your score sits at the neutral 8.0 until you complete deals."
Rules and edge cases
- A brand-new account is not penalised for having no history. 8.0 is the starting point, not a score you have to earn up to.
- Deleting a post after the paid window has fully elapsed is fine. The deal is complete — only deletion during the window is the −1.50 penalty.
- Low views are not a penalty. Reliability penalises non-delivery of the placement, never market view variance. Underdelivery remedies, where a plan offers them, are financial.
- Both sides of every paid deal are scored. Owners earn and lose on delivery; advertisers earn and lose on how they run the deal they started — and on how they use the concern process.
- Each penalty lands once per order — but one order can carry more than one kind. The delivery monitor's flag can only apply its penalty a single time on any given order, and a second look at the same violation cannot dock you for it twice. Different failures on one placement are separate marks, though: publishing over an ad inside its exclusive window (−0.40), unpinning a paid pin (−0.75) and taking the ad down early (−1.50) are separate promises broken, checked independently.
- A penalty that falls due while you are Under Review still applies. The freeze hides your tier; it does not suspend the rules.
- The score is clamped to 0–10. It cannot go below 0 or above 10 however many events land.
- A restore returns the exact points, not a rounded amount — you are returned to the score you were on immediately before the penalty.
- Published values can change, going forward only. A change to what a penalty costs applies to events after the change; an entry already on your ledger keeps the amount recorded against it. A change to a floor reaches further, because a floor is tested against the score you have at the moment you act — so raising one can lock a surface you could use yesterday without anything on your record having changed (Terms §15.6).
- A bad record survives account deletion. Deleting an account that leaves with a below-baseline score records a one-way hash of the identity and the floor it must resume at — kept for 45 days, or 90 days after an unforgiven upheld fraud. Registering again with the same identity inside that window resumes at that floor rather than at 8.0, and the carried-over amount appears on the new ledger as its own entry with its reason. A good-standing account leaves nothing behind. Deleting and re-registering more than once keeps the worst floor and the longest window.
- Merging accounts never launders a score. When a Telegram starter account is folded into your website account, the merged account keeps the lower of the two scores, the full history of both, and the event count of both.
Frequently asked questions
What score do I start with?
8.0, the same for everyone.
Is there a separate score for what the Telegram bot sees?
No. One score per account. When the bot's check finds a placement was removed early, it records that fact into a queue the website owns; the website prices it from its own table, writes the ledger entry and moves the score — usually within a minute. You appeal it exactly the same way as anything else. The bot's own message tells you what was recorded and that it is applied on onflowads.com; if it could not record the fact at all, it says the event was logged for review and that nothing has been applied yet.
The same is true of the one channel check the bot does not do. The top-of-feed exclusivity check reads your channel's public history through a different reader — not the bot, and not from an admin position — and its −0.40 still lands on this one score, in this one ledger, with the same Appeal button on it. There is no second record anywhere.
Can I buy a higher score?
No. Plans change your limits, fees and tools — never your score. The only inputs are conduct and time.
Will my score drop if Telegram goes down and my ad cannot post?
No. System and platform failures never cost you reliability; the buyer is refunded or rescheduled instead. Only failures that are genuinely yours are penalised, and unclear cases go to a human with no penalty applied. The one exception to watch is the accept-window penalty, which is applied from the clock alone — appeal that one if an outage caused it.
How do I get back above 8.0?
Only by completing real deals — +0.20 per clean delivery as an owner and +0.10 as an advertiser, with +0.05 credits for clean Subscriber Exchange runs. Automatic healing stops at 8.0; everything above neutral is earned.
How long does healing take?
You need a clean streak first — 14 to 45 days depending on the worst recent mark — and then it heals at a published daily rate, at most once a day. The table above has the exact figures, and the same table is on your reliability page.
My label changed to "Under Review". What happened?
A fraud concern involving your account is open, or a hold was placed on your wallet. Your tier is hidden, the score stops healing, withdrawals pause and you cannot start new deals until it resolves. See Account standing.
Someone raised a concern against me that I know is false. Does it hurt my score straight away?
No. An open concern moves your score only if it is upheld. While it is open, your score is frozen and hidden rather than reduced. If the reviewer cannot establish fault either way, the concern is dismissed and nobody's score moves.
An old entry has scrolled off my ledger. Can I still see it?
The page shows your 50 most recent events. If you need one that has fallen off — and it is still inside the 30-day appeal window — contact [email protected] with the date and the amount.
Can I see the score of a channel I am about to buy from?
The score feeds marketplace ranking and the trust signals shown on listings. Your own full ledger is private to you.
Where can I dispute an entry?
From the entry itself — the Appeal button in Your ledger, within 30 days of the penalty and up to three times a day. See Appeals and fraud concerns.
Related
- Account standing — the states that can block you regardless of your number
- Appeals and fraud concerns — contesting a mark, step by step
- Fraud and bot protection — the checks that produce most of the penalties above
- Owing money — what a paid pin taken down early or a forfeited cross-promotion cover does to your wallet
- Marketplace entry gates — the floors alongside every other gate on joining a marketplace
- The Boost API — the keys and per-minute rates your score now bands
Next: the score is only half the picture. Account standing covers the states that can block you regardless of your number — Under Review, support restricted, suspended, banned, connection refused — and how each one ends.