KPI Tree and Driver Map Template
Four documents and six sheets that rank your drivers by what each one has actually released, rather than by a leverage score.
Free download · No account needed
How Sensitivity Is Calculated
The ranking is arithmetic and reproducible from Monthly History alone. Read this before arguing with it.
The problem with elasticity
Brightwell's identity is Sessions x Enquiry rate x Lead-to-order rate x Average order value. Differentiate a product with respect to any one factor and the elasticity is exactly 1. Lift any driver ten percent and revenue rises ten percent, whichever one you picked.
So the Elasticity column reads 1.00 on all four rows. That is the structure of the identity rather than a rounding artifact. It also means the usual instruction to work your highest-leverage driver is an instruction to choose between four identical numbers.
What replaces it
The drivers are identical in what a one percent move produces. They are nothing like identical in how far they have ever moved. Sessions have ranged from 33,900 to 53,600 across twenty-four months. Average order value has ranged from $1,836 to $2,014 and never left that band.
| Driver | Own P90 | Elasticity | Releases | Share of gap |
|---|---|---|---|---|
| Sessions | 51,110 | 1.00 | $85,097 | 83.1% |
| Lead-to-order rate | 24.82% | 1.00 | $45,662 | 44.6% |
| Enquiry rate | 3.37% | 1.00 | $41,193 | 40.2% |
| Average order value | $1,992 | 1.00 | $24,955 | 24.4% |
Identical elasticity. A 3.4x spread in what a best-ever month is worth. Nothing on that table closes the $102,389 gap, and that finding only exists because the levels are bounded by history instead of chosen.
Every guide to building one of these ends by telling you to work the driver with the most leverage. In a multiplicative tree that instruction is empty. Revenue is the product of its drivers, and the elasticity of a product to any one of its factors is exactly one, so lifting sessions ten percent and lifting average order value ten percent both lift revenue ten percent. The elasticity column on this pack's tree reads 1.00 on every leaf row.
What differs between drivers is range. Brightwell Supply's sessions swung 36.4 percent between their tenth and ninetieth percentile over twenty-four months, while average order value swung 7.7 percent and never left a ten percent band. So the ranking here is realized leverage: the dollars released by moving one driver to a level it has actually reached, holding the others at their medians. Identical elasticity, and a 3.4x spread in what a best-ever month is worth.
Written for whoever owns the operating plan and has to hand four teams a number each. Reach for it when a target has landed and the room is already arguing about which lever to pull. Every level here carries a date and a count of the months already spent at or above it, which turns an assigned number into a commitment somebody can accept or decline. The definitions underneath belong in a metric register, the weekly operating view in a RevOps dashboard pack, and the version a leader actually reads in an executive one-pager.
What you get
KPI Tree
The identity as a table with the level, the parent and what each node multiplies against. Leaf drivers carry an owner; intermediate nodes deliberately do not, because a node two owners meet at produces somebody accountable for a number they cannot touch.
Driver Sensitivity
One row per driver with its own P10, median, P75, P90 and best observed month, the P10 to P90 swing, and the realized leverage that swing is worth. The elasticity column sits beside them reading 1.00 throughout, which is the point.
Commitment Scenarios
Every single driver at its own P90, every pair and every triple at P75, and the co-occurrence bound. Seventeen rows that answer whether one team can reach the target, which is usually the question underneath the request.
Driver to Owner Map
The level asked for, the level accepted, the months already spent at or above it, and the levers inside and outside that person's authority. Counter-offers are recorded with what they cost, because a recorded decline is a working outcome.
Leading Indicator Register
Each candidate correlated against its driver at every lag from zero to three months, with the measured lead, the overlap count and a verdict. Rejected candidates keep their numbers so they do not come back next quarter as suggestions.
Monthly History
Twenty-four months of sessions, leads, orders, revenue and the five indicator series. Every percentile in the pack is reproducible from this one sheet, so a driver value nobody can trace back to it does not belong in a scenario.
How Sensitivity Is Calculated
Why elasticity is a constant on a multiplicative tree, what realized leverage measures instead, why the ninetieth percentile rather than the maximum, and why the commitment level is lower again.
What the Tree Cannot Tell You
Five questions the map will be asked and cannot answer, each with a number rather than a caveat. The largest is co-occurrence, and it is worth 24.3 percent in the worked example.
How it works
- 1
Send the target and the history
The number the board asks about, and whatever monthly export you have. River writes the identity and proves it multiplies back to one real month before anything else runs.
- 2
Measure each driver's own range
Twenty-four months per driver, rates recomputed from their underlying counts. Percentiles, the swing, the months at or above P75, and the longest consecutive run at that level.
- 3
Rank on what has happened
Each driver moved to its own P90 alone, then pairs and triples at the commitment level. The output is the smallest set of drivers that actually closes the gap.
- 4
Put a level to each owner
One note per driver carrying the level, the months already there, and the levers they control. Accepted, accepted with a condition, or declined with the counter-offer priced.
Frequently asked questions
Is this free, and what format are the downloaded files?
Free, and no account is needed for the download. Documents arrive as .docx and sheets as .csv, so they open in Word, Pages, Google Docs, Excel, Numbers or Sheets with no conversion step. Edit with AI is the other button: it installs the same pack as a private space and fills it in from your own history.
Why is every driver's elasticity exactly 1.00?
Because the tree is a product. Lift any factor of a product by one percent and the product rises one percent, whichever factor you picked. It is arithmetic rather than a property of your business, which is why an elasticity column cannot rank drivers and why this pack ranks on observed range instead.
Why the ninetieth percentile rather than the best month ever?
Because one month may have had a trade show in it. P90 of twenty-four months lands near the third-best, which the business has demonstrated more than once. Twenty-four observations is the working level and twelve the floor, with the count on every row so a short series stays visible.
Can I just multiply a monthly P90 by twelve for an annual plan?
No, and the pack shows why. None of Brightwell's four drivers has ever held its own P90 for two consecutive months. At P75 sessions managed four in a row and two drivers never managed two. An annual commitment needs the sustained-run count, not the single-month percentile.
Does a driver need more than an owner's name against it?
Yes, and there is precedent outside operating reviews. The SEC expects a company presenting a key metric to say how management uses the metric in managing or monitoring the business. That third element is what a driver map usually skips, and the commitment note supplies it.
How do you know a leading indicator actually leads?
By testing every lag rather than assuming one. The register correlates each candidate against its driver at zero to three months and takes the strongest. The OECD's own methodology locates an average lead the same way, at the peak of the cross correlation function, while warning that correlation size cannot be the only test.
Does this space define the metrics or build the dashboard?
Neither, deliberately. Naming and defining them comes first in the metric definitions pack, and specifying what gets built comes after in the dashboard spec tool. Keeping them apart is what stops a ranking conversation turning into a definitions argument halfway through.
Find out whether one team can reach your target
Send the number and whatever history you have. River writes the identity, ranks your drivers on what each has actually released, and names the smallest set that closes the gap.
Edit with AI