OnlyFans chatter metrics

A defensible reply timer starts with an eligible inbound fan message and stops at the first meaningful human reply. It should not run while the conversation is unassigned, outside the person's agreed shift, blocked by the platform or waiting for a manager approval the chatter cannot provide. System messages, duplicate fan pings and the chatter's own follow-ups do not belong in the same calculation.

1. 1Show the distributionMedian and a slow-tail percentile reveal the normal experience and the conversations being abandoned. One average can hide both a fast majority and a handful of disastrous waits.
2. 2Split controllable from uncontrollable timeAssignment, shift and approval states decide whether a delay belongs to the chatter, the rota or the manager. Do not dock someone for a clock they could not stop.
3. 3Pair speed with outcomeRead response time beside repeat replies, purchases, flags and fan retention. Faster is useful until it starts producing generic answers and avoidable mistakes.
4. 4Review examples, not only ranksSample the fastest and slowest conversations. A number tells you where to look; the exchange tells you whether coaching, coverage or process is the answer.

For each chatter, count eligible priced offers, paid unlocks and the value unlocked after a fixed observation window. Exclude free messages, tips, failed sends and duplicate records. Keep send date attached to the offer so a late purchase returns to the original cohort rather than flattering whichever shift happened to be online when the transaction appeared.

Then split the funnel by creator, fan segment, offer and price band. A chatter working a warm, high-spend book should not be compared raw with one covering cold or lapsed fans. Attribution is useful only when it describes the opportunity each person received as well as the sale they produced.

A priced message sent is offer value. An unlock is observed gross. Realised net is the creator-side amount after the platform's treatment, and reconciled revenue is the statement-backed transaction classification used for money that moves. These labels answer different questions; combining them produces a payout dispute sooner or later.

Fan value needs the same discipline. Put attributable tips, unlocked PPV, purchase count, recency and repeat behaviour beside the conversation, but state what is missing. Subscription revenue is not automatically attributable to a chat thread, and a message-history cap makes spend a floor rather than an exact lifetime total.

Discounting can rescue an offer or quietly train the whole book to wait. Track the original price, final price, who approved the change and whether the fan bought again. Review both discount frequency and discounted value; ten tiny exceptions and one enormous concession are different management problems.

Refunds and chargebacks arrive after the conversation that caused them. OnlyFans' current terms say a successful refund or chargeback can be deducted from Creator Earnings, while its Fan–Creator contract preserves applicable refund rights and tells fans not to charge back except for a good-faith dispute. That makes reversals a necessary quality signal, but not automatic proof that the chatter was at fault.

• Link the reversal to the original transaction, offer and conversation.
• Separate refunds, disputes and confirmed chargebacks; they are not interchangeable labels.
• Review the promise made, the content delivered and any manager approval before assigning responsibility.
• Hold commission calculations to the reconciled ledger rather than clawing provisional pay back through a spreadsheet.

The current OnlyFans terms are explicit that a creator remains legally responsible for account content and compliance when someone else helps operate the account. A team scorecard therefore needs a manager-review lane for risky work: off-platform payment or contact attempts, prohibited meeting language, underage indicators, refund promises, unexplained deletion patterns and unusual discounting.

A signal is a routing decision, not a verdict. Store the message, actor, source rule, confidence, timestamp and manager outcome. That record lets the agency tune false positives, coach repeated problems and clear a chatter when the allegation does not survive review. An editable screenshot folder cannot do that.

Keystrokes, mouse movement, online dots and screenshots measure visible activity. They cannot tell whether the right fan received the right reply, whether a PPV promise was accurate, whether the fan bought again or whether the revenue survived reconciliation. Once people are ranked on presence, they optimise for presence.

There is a narrower and better record: the work product and its result. Attribute each outbound message to a person, keep the preceding fan context, connect priced offers to purchases and reversals, and preserve review decisions. That is enough to coach commercial performance without turning the job into webcam proctoring.

• Use message count to explain workload, never as proof of quality.
• Use active time to diagnose a rota only when the clocking method is disclosed and reliable.
• Never convert a missing event into a zero; unknown and none are different states.
• Do not let a composite 'integrity score' replace the underlying evidence and human review.

FanHelm's personal chatter view shows Today, This week and This month plus a dense 30-day history. It counts outbound sent messages authored by the signed-in chatter, attributed transactions, gross sales observed and commission observed. The view is scoped from the authenticated identity, not a chatter ID supplied by the browser, and explicitly says that it is not a payroll or payout statement.

Response time currently comes from the chatter's recorded shift averages. It is an average, not yet the median and slow-tail scorecard described above; FanHelm also leaves 'waiting over ten minutes' unknown because the current conversation record has no waiting-since field. A visible gap is better than a fabricated service-level metric.

Managers and admins have a real QC message feed with the attributed sender, creator, fan, preceding inbound message, team reply, average reply time, PPV price, purchased state and message status. A separate manager-only flag feed surfaces banned-phrase and anomaly detections with confidence and a route back to the conversation. The current UI supports review; it does not pretend every flag has already been adjudicated.

Account analytics separately report observed net and gross earnings and withhold the PPV unlock rate until the message scan is complete and its coverage reconciles. FanHelm does not yet present per-chatter realised-net, discount, refund or chargeback rates as finished dashboard metrics. Those are the next scorecard layers; current screenshots use labelled demo data wherever an example is needed.