Leverage and the "multiple" explained
Leverage — shown as a Leverage figure on Sessions and as Value multiple on Overview — is the ratio of the human-equivalent value of the work you shipped to what you actually spent on AI. It answers one question: how much work did you get back for every dollar of AI spend? A 12× multiple means every $1 of AI spend produced about $12 worth of human-equivalent engineering work.
Where you'll see it
Sessions page, in the stat panel above the session list: a Leverage tile sits alongside Delivered value, Spend, Tokens, and Cached for the selected date range.
Sessions page, inside a single session's detail panel: Leverage appears in the Economics section next to Tokens, AI spend, and Value.
Overview page, in the value statement at the top of the page: the last card reads Value multiple, with a line underneath like "every $1 paid → $12 of work."
Overview page, in the Spend → Value section: hovering a repo row shows that repo's leverage for the period alongside its PRs, commits, sessions, tokens, AI spend, and value.
How it's calculated
Leverage = Value ÷ what you actually paid for the AI usage behind it.
Value is the human-equivalent cost of the work — an estimate of what it would have cost to pay a person to produce the same code, at your hourly rate.
What you paid is your actual AI spend for that scope — your subscription/plan cost once you've added one, or the API-equivalent metered cost otherwise.
The multiple is formatted as a whole number once it clears 10× (e.g. "27×"), and to one decimal place below that (e.g. "3.4×").
What "Value" is based on
Each session that ATTRIBUT has scored gets a work-type classification (creation, reduction, repair, understanding, or orchestration), a confidence score, and an estimate of "effective expert-hours" (EEH) — the code it produced converted into the hours a human engineer would have needed, then dollarized at your hourly rate ($100/hour by default). That session's detail panel shows all of it: the work-type badge, the confidence percentage, the EEH estimate, and how it splits across the work types involved. Value and Leverage are estimates, not a replacement for judging the work yourself — a low confidence score is a signal to lean on that underlying detail rather than the headline number.
Reading the number
Leverage | What it means |
1× | The work was worth about what you spent on AI for it — breakeven. |
3–10× | Solid return: a few dollars of human-equivalent work for every dollar spent. |
10×+ | Large return: the estimated human-equivalent value is many times the AI spend. |
— | Not enough data to compare — see below. |
When it shows "—" instead of a number
Leverage needs both sides of the ratio to compare, so it shows a dash when either is missing:
There's no human-equivalent value estimate yet for that session or period.
AI spend for that scope is zero.
Changing the assumptions behind it
Leverage moves with your hourly rate — raise it and both the value estimate and the multiple go up. Set your own rate from the "Adjust calculations" link on the Overview page's value statement; until you do, ATTRIBUT uses a $100/hour default. The "How we measure →" link on that same statement explains the underlying methodology in more detail.