AGI allocation
The way agi allocation is calculated changed between these two points, so they are not comparable and no movement is shown.
Comparisons resume once a full period has been recorded the new way. The number itself is unaffected.
Where it stands
- People
- 57%typically 55 to 60%
- Overhead
- 22%typically 20 to 25%
- Profit
- 21%~20%, floor 10%
AGI allocation reads 21% and the whole split sits where it usually does. The full split is below.
- Window
- Stated as of Aug 23, 2026. The period the reading covers is set by its source and is not stated on this tile.
- How it is counted
- People costs and overhead costs are each taken as a share of adjusted gross income. Profit is what remains after both, so the three shares always account for all of AGI. Only the profit share is flagged against a target; the people and overhead shares are shown against their typical ranges so you can see where the profit came from.
- Source
- Financial ledger, as of Aug 23, 2026
- Drivers
- Computed from this workspace's own records.
What this is
How adjusted gross income splits across people, overhead, and profit. People costs and overhead costs are each taken as a share of adjusted gross income. Profit is what remains after both, so the three shares always account for all of AGI. Only the profit share is flagged against a target; the people and overhead shares are shown against their typical ranges so you can see where the profit came from.
Why it is on your dashboard
The core health check for the whole firm: how the income your own work earns gets divided between the people who do it, the cost of keeping the doors open, and what is left over. The first two trade off against each other freely. The one that has to be protected is the profit share, around 20%, because it is what remains after the other two have taken theirs. It sits on this board against profit share, target ~20%.
Why it is where it is
This accounts for how adjusted gross income is allocated, the whole of adjusted gross income in all.
The team's share of income sits inside its typical range. This is the share that pays for delivery.
How this was measured: This band's share of adjusted gross income, read from the served allocation split. Each point of income is counted in exactly one band.
Overhead sits inside its typical range: fixed costs are taking about the share this kind of firm usually gives them.
How this was measured: This band's share of adjusted gross income, read from the served allocation split. Each point of income is counted in exactly one band.
The protected share is at its level, which means the two shares in front of it are together taking what the firm can afford. How they split BETWEEN themselves is a choice, not a fault.
How this was measured: The remainder: adjusted gross income less the people and overhead shares. It is not measured independently, and that is the point of reading it as a split.
The three bands are shares of one whole and account for all of adjusted gross income; profit is the remainder by definition. A point taken by one band is a point unavailable to another.
What we checked, in order
Why the profit share is where it is.
Six questions about the work, and only then a seventh about revenue. The order is the whole teaching here: selling more into a delivery problem adds work at the same margin and consumes the capacity that would have fixed it.
2 of these 7 questions can be answered from what this workspace holds today. The rest name what it would take, rather than filling in with something that would be true of any firm.
- 1
Is the work being sold for less time than it takes?
First because an estimate that is wrong at the start is wrong in every period after it.
The board already measures this. Work landing inside what it was sold for is what an estimate being right looks like in aggregate, and this is that number. Read it as a learning loop rather than a score: the useful move on a miss is the conversation about why a piece of work took what it took, not a tighter review at billing. There is no ratified level on it yet, so it can be read against its own history and not against a standard.
- On-budget delivery
- 90%no target set yet, so there is no band to read it against
On-budget delivery, from this board - 2
Is the way the work runs adding time?
Second because process time is invisible on an invoice and shows up only as cost.
Not checked here
Time is stored against a client and a number of hours, with nothing recording what the hours went into. So we can see how much time the work consumed and not whether it went into revision rounds, waiting on approvals, or doing something twice, which is the distinction this question turns on.
A work type or phase on each time entry. It is the same grain that would let the board show where non-billable time actually goes, and it is a change to how time is logged as much as to what is stored.
- 3
Is the wrong seniority doing the work?
Third because senior people doing work a junior does faster is a cost problem that looks like a quality decision.
Not checked here
Hours are not joined to the person who worked them and their cost, so who did the work and what it cost to have them do it cannot be read. A blended cost per hour IS available and does not answer this: work can look cheap per hour and still be done by the wrong people.
Hours attributed to a person, joined to that person's cost rate. Both halves are on file separately and nothing joins them per piece of work.
- 4
Does somebody need training?
Fourth because one person materially slower than their peers in the same role is a specific and fixable version of the rung above.
Not checked here
Hours are not joined to the person who worked them and their cost, so who did the work and what it cost to have them do it cannot be read. A blended cost per hour IS available and does not answer this: work can look cheap per hour and still be done by the wrong people. Comparing one person against their peers needs that join before it needs anything else.
Hours attributed to a person, joined to that person's cost rate. Both halves are on file separately and nothing joins them per piece of work.
- 5
Is the shape of the team wrong for the book?
Fifth because capacity mismatched to the work is a structural cost nobody bills for.
At least one of the two per-person readings is outside its band, which is what a mismatch between the size of the team and the size of the book looks like from the outside. It does not say which way to fix it: the same reading comes from too many people for the work and from too little work for the people. One caution on the first of these. Income per person rises on its own when work moves to contractors, to an offshore team, or to tooling, because the work leaves the count of people without leaving the firm. A rise is worth checking against how the work is actually staffed before it is read as improvement.
- AGI per FTE
- $192kagainst target ~$175K, which it is outside
- AGI per AE
- $467kagainst target $400K to $500K, which it is inside
AGI per FTE, from this boardAGI per AE, from this board - 6
Is a whole service line losing money?
Sixth because it is the largest change on the list and should not be reached for until the five above have been asked.
Not checked here
Work is not classified by the kind of service it is, so revenue and cost cannot be totalled per service line. A line that loses money on every job is invisible when every job is counted together.
A service line on each piece of work, carried through to what it billed and what it cost. It is the largest of the gaps on this ladder and it answers the largest question on it.
- 7
And only now: is this a revenue problem?
Last, deliberately. This is the rung owners start at, and starting here is what turns a margin problem into a bigger margin problem: the new work arrives at the same margin and uses the capacity that would have fixed it.
Not checked here
This tile reports shares of earned income, and a share cannot separate income falling from cost rising. Both push the profit share the same way. Answering it needs the earned income level over the same window next to these shares, which this tile does not carry.
It is deliberately last, and that is the part worth keeping even while it cannot be answered here. Selling more into a delivery problem adds work at the same margin and uses the capacity that would have fixed it, so this question is the one to reach for after the six above and not before them.
Root cause
Nothing here needs a cause: the share the split protects is landing where it should. The thing to watch is drift, because a split moves a point at a time and never announces itself.
What to do now
Not answered here.
The protected share is holding on this read, so there is nothing to move this week.
This fills in when the protected share leaves its range.
What stops it coming back
Budget to the bands at planning time, so the split is an allocation you chose rather than a result you discover. A hire, a lease or a tool subscription each lands in a band before it lands in a ledger.
The structural version of a healthy split is that the bands are used as the planning frame: every standing cost is owned by a band, and a decision that would push a band past its range is visible as that before it is signed. The product can keep the split honest; deciding the shape of the firm is not a product setting. The checks above are the structural half of this. 2 of 7 can be answered from what this workspace holds. The rest (Is the way the work runs adding time; Is the wrong seniority doing the work; Does somebody need training; Is a whole service line losing money; And only now: is this a revenue problem) each name the record that would answer them. Closing those is what turns this from a diagnosis somebody reconstructs each time into one that can be walked in order.
This closes no measured driver above, and does not claim to.
What happens if this is ignored
A split off its bands does not correct itself, because the costs behind it are standing costs: a lease renews, payroll recurs, and the remainder quietly absorbs the difference every month. The failure mode is a year that feels normal while the profit share thins, followed by cost cuts aimed wherever cutting is easiest rather than at the band that drifted. Reading the split monthly is what makes the drift a decision instead of a discovery. Where this lands is not this tile. It is AGI reads $1.40M, with no band set for it. There is no band set on that tile, so it can show you the movement and cannot tell you whether the movement is already too far. A remainder squeezed at this level of income is a scale question as much as a cost one, and the income level is the other half of it. Reading the split without it is how a firm answers a size problem with a cost cut.
Is this target still telling you anything
Not answered here.
The window spans a measurement change, so its periods are not one series: compute path changed: demo-history-seed@1 to live-serving-rollup@3
How this metric is measured changed inside the period this section would look back over, so counting hits across it would be counting two different measurements as one. This fills in once the whole window sits on one basis.