Creator Agency Metrics That Matter, and Which Decorate

Revenue is the number every agency steers by, and it is the one number that cannot tell you whether anything you did this week worked. Across our corpus of 42,852 conversations, what predicts whether a thread converts is where the offer sits inside it, a thing you can measure on any inbox, on any shift, for any person.

Most agency dashboards are built from numbers that describe the past accurately and steer nothing.

Which numbers actually steer an agency?

The ones that survive three tests. A number earns a place on the board you make decisions from only if it is:

1. Comparable between two people doing comparable work, on different inboxes.
2. Leading: it moves before the money does, not after.
3. Actionable: a decision you can take this week changes it.

Revenue fails the first two and passes the third only by accident. Here are the same three tests applied to the numbers most agencies already have:

Number	What it tells you	What it hides	Steers?
Revenue per chatter	What the inbox produced	Which inbox they were handed	No
Conversion inside a volume band	How comparable threads are worked	Little, if the band is honest	Yes
Where the offer lands in the thread	The habit our data ties to conversion	Nothing: it measures behaviour directly	Yes
Share of the inbox opened per week	Whether the quiet half exists to you	Which threads were chosen	Yes
Coverage of the evening block	Whether you staffed the hours that convert	How well those hours were used	Yes
Average response time	Whether live threads go cold	Which threads got the fast reply	Partly
Messages sent per shift	Typing volume	Everything that matters	No
Total fan count	Audience size	Whether any of them buy	No

The four rows in the middle share something: none of them is money. They describe what the team did, which is the only thing you control.

What is the finding, and what does it rest on?

The measured result behind the third row: across 42,852 conversations covering 1.5 million messages and 38,879 sales, an offer placed before the sixth message converts roughly a third worse than the same offer placed later. The best position sits after about ten exchanges, which is the subject of when to pitch a sale.

What was compared: where the first offer sat in a thread, against whether a sale followed.

What the measurement does not show, which matters more than the result:

• It is observational. Nobody assigned chatters to conditions. It is an association strong enough to act on, not a proof of cause.
• It does not identify who typed. If experienced chatters also pitch later, part of this is seniority showing up as message order.
• It is a population average. Your creator has one audience, one price grid and one voice, and can sit anywhere in the spread around it.
• It measures the sale, not the year afterwards. A habit that lifts conversion today and tires a fan out faster would look identical to a good one here.
• It says nothing about money. The corpus holds conversations and sales, not payroll, margins or prices. The full boundary is in what our data does not show.

What makes this a metrics story is not the finding. It is that offer position is a number your team generates every day and nobody records, while revenue, which everybody records, could not have told you any of it.

Why can two chatters’ revenue not be compared?

Because it mostly measures allocation. Two people on your roster are not running the same experiment: the differences between their inboxes outweigh the differences in their skill.

What sits between a chatter and their revenue, before any question of ability:

• Who they inherited. A thread with a purchase history behind it converts more easily than a cold one, and nobody on the team chose which they got.
• How many high spenders are in the pile. One whale makes a month and their absence unmakes it. Neither is a performance event.
• Which hours they work. Evenings maximise conversion and the small hours collapse it, so a rota decision shows up in a person’s revenue as if it were skill. See what time fans buy.
• How many threads they were given. Volume and rate move in opposite directions, and totals hide the trade.
• Whose account it is. Price tier and niche differ enough that cross-account comparison mostly measures the creator.

The correction is arithmetic rather than clever: compare inside a band. Split the roster into thirds by threads handled, compare conversion only within a third, and the ranking usually changes. The full grid for doing that to people rather than to the agency is the chatter performance scorecard.

Which numbers are decoration?

The ones that go up when the team is busy. Busy and effective produce the same reading on these, which is exactly why they are comfortable to look at:

• Messages sent. Rises when threads are sprayed and when they are worked well. Used as a target it produces the first.
• Total fans or subscribers. Says nothing about whether any of them buy, and can grow for months while revenue falls.
• Average response time as a target. Improves fastest by answering one-liners and skipping the conversation nearest a decision.
• Gross revenue as a headline. Before the platform’s cut and before commission it is not the money anyone receives (see net revenue).
• Month-on-month totals with no volume attached. More threads worked worse beats fewer threads worked better, every time.

None of these are lies. They just cannot tell you what to do differently on Tuesday.

What do you measure tomorrow morning?

Four things, none needing a new tool. The point is to generate the numbers you do not currently have: a week of discipline, not a project.

1. Offer position. For every thread where an offer went out, record how many exchanges preceded it. A tally is enough. This is the leading indicator our data ties to conversion.
2. Conversion inside a band. Threads that received an offer, and how many bought, grouped by volume band rather than by person.
3. Share of the inbox opened. Distinct threads opened this week against threads that exist. Left unmeasured, the quiet half of the inbox stays at zero.
4. Evening coverage. Hours staffed in the block that converts, against hours staffed overall.

Then translate one of them into money as your own worked example, never as a benchmark:

Line	Where it comes from	Your number
Threads that received an offer last month	Your tally	___
Of those, how many converted	Sales log	___
Threads where the offer landed before the sixth message	Your tally	___
Your average order value	Platform statement	___
Value of moving those early pitches later	Multiply the three above out yourself	___

The last row is deliberately blank. We measure conversations, not your prices, and a number supplied by us there would be an invention.

How do you check this against your own numbers?

By running the comparison on your own inbox. The method finishes inside a week:

• Pick one closed week and pull the threads where an offer was made.
• Split them in two (offers before the sixth message, offers after) and compare the conversion of each group. Adjust nothing yet.
• Hold volume roughly equal across the groups, or you are measuring shift luck again.
• Repeat on a second week. One week of thread-level data is noisy at any agency’s scale.
• Then change one thing. Move the offer later on one account, leave another alone, and compare over a month.

If your split disagrees with ours, your split wins. That is the correct outcome, and the reason a metric you can check beats one you can only admire.