Backed by Y CombinatorThree documents and three sheets that track measured benefits against a locked baseline, then net out what a confounder, not the engagement, actually caused.
Free download · No account needed
What you get
Pack of 6 files: 3 documents, 3 spreadsheets
Benefit Realization Tracking TemplateBenefit Register
| Ref | Benefit | Baseline | Target | Actual | Attributable |
|---|---|---|---|---|---|
| B-01 | [Benefit name] | [Value, date, source] | — | — | — |
| B-02 | [Benefit name] | [Value, date, source] | — | — | — |
| B-03 | [Benefit name] | [Value, date, source] | — | — | — |
Baseline is locked before anything changes. Attributable stays blank until a confounder is actually checked, never filled in from the baseline gap alone.
Every benefits realization template treats the arithmetic as the whole job: baseline, target, actual, a percent-realized column. That arithmetic was never the hard part. The hard part is that most registers get built after work already started, from whatever the consultant and the client can still remember about where things stood. A client's own operations team can unravel a claim like that in one sentence about what else happened that quarter. This pack starts from the baseline the business case already locked, before anyone had a reason to remember it favorably.
Two pieces of public evaluation guidance say why that matters, in language written for government evaluators rather than consultants. HM Treasury's Magenta Book says a baseline has to be collected before an intervention starts, so a counterfactual can be estimated later, not reconstructed from memory once results are already known. New Zealand Treasury's guidance on managing benefits treats attribution as a separate step from measurement entirely, distinguishing a benefit an initiative caused directly from one it merely contributed to alongside other factors.
Kestrel Bay Seafood, a fictional seafood packer, locked a baseline of $22,000 a month in packing-line overtime on 3 March 2026, before a revised shift schedule went live. At the 90-day check-in, overtime had fallen to $9,830, a $12,170 swing against the baseline. But packing volume had also dropped 22% after a wholesale account delayed an order, unrelated to the schedule. Netting out what that volume drop would have cost anyway leaves $9,090 attributable to the schedule change, not $12,170, a $3,080 difference that a naive comparison would have handed to the wrong cause.
Benefit RegisterKestrel Bay Seafood · baseline 3 Mar 2026
| Ref | Benefit | Baseline | Target | Actual | Attributable |
|---|---|---|---|---|---|
| B-01 | Packing-line overtime premium | $22,000/mo | $10,100/mo | $9,830/mo | $9,090/mo saved |
| B-02 | Case-pack rework and scrap rate | 3.8% | 2.5% | 2.9% | 0.9 pt improved |
| B-03 | Expedited freight spend | $4,300/mo | $1,800/mo | Not yet measured | Window opens 5 Sep 2026 |
The naive movement on B-01 alone was $12,170 a month. $9,090 is what actually survived the confound check.
Attributable stays blank until the Attribution Note tests for a confounder, never filled in straight from the baseline gap.
Attribution NoteB-01, packing-line overtime premium, 90-day check-in.
| Value | |
|---|---|
| Naive movement | $12,170/mo saved |
| Confounder identified | Packing volume fell 22% after a client delayed an order |
| Historical relationship | $140/mo overtime per point of volume change |
| Counterfactual overtime | $18,920/mo |
| Attributable movement | $9,090/mo saved |
$22,000 baseline plus (−22.0 × $140) = $18,920 counterfactual. $18,920 minus $9,830 actual = $9,090 attributable.
Realization ReportBenefit Realization Report
Overtime measured $9,830/mo against the $22,000 baseline, a $12,170 movement on its face. Packing volume also fell 22% in the same window after a wholesale account delayed a recurring order. Netting out what that volume drop would have done to overtime on its own leaves $9,090/mo attributable, 76% of the $11,900 projected.
The report leads with $9,090, not $12,170, because the larger number would not survive a question from anyone who tracks the client's own order book.
Every benefit named to a baseline value, date and source locked before the recommendation went live, against its target.
The exact metric definition, data source and extraction method for every benefit, written before the first actual arrives.
Tests every movement against a named, evidenced confounder before calling it attributable, with the counterfactual arithmetic shown.
The one summary a sponsor reads start to finish, leading with the attributable figure rather than the naive one.
Accounts for whatever gap remains against target once the confound check is done, naming the specific rollout cause.
The Kestrel Bay Seafood example showing why a $12,170 naive saving and a $9,090 attributable one differ by a confounder.
Open in River and send what you baselined before the recommendation went live, the value, date and source for each benefit, plus the business case target.
Every benefit gets one row with a real baseline value, date and source, never reconstructed after the fact once work has already started.
When an actual arrives, it gets tested against a named, evidenced confounder before being called attributable, never assumed equal to the naive figure.
A live sheet the client's team can rerun, plus the Realization Report and Variance Explanation written from what the register actually found.
Yes. The zip is Word documents and CSV sheets, no account and no card. Edit with AI is the optional half: the agent locks each baseline from what you send, writes the measurement method, and tests every movement against a confounder once an actual arrives. More sit in the template library.
Three Word documents and three CSV sheets in one zip. The mechanism note, Benefit Realization Report and Variance Explanation open in Word, Pages or Google Docs. The Benefit Register, Measurement Method and Attribution Note open in Excel, Numbers or Google Sheets with every column intact.
A normal tracker compares actual against baseline and stops, crediting the engagement with everything that moved. This pack adds a required step: naming a specific, evidenced confounder and computing what it alone would explain, so the reported figure is what the engagement attributably caused, not what the calendar happened to produce.
A named, evidenced alternative explanation tested against data the client already has, not a category like "market conditions." A wholesale account delaying an order and moving a plant's shipped volume is a confounder. "Nothing else seems to have changed" is not a check; it is the absence of one.
The register carries "Not yet measured" across the actual and attribution columns rather than a blank or a guess, with the date the window opens stated plainly. Nothing gets reported for that benefit until real data exists to test.
This space starts once a recommendation is live and its business case baseline is locked, and it runs until every measurement window closes. The handover pack covers transferring the deliverables themselves, and the findings report is usually where the recommendation itself originated.
Written by
Co-founder & CEO, River
Chandler has spent the last six years working in AI, building tools for millions of users and Fortune 100 companies. In 2025, he co-founded River to build a new kind of operating system for AI agents.
Take the Word documents and CSV sheets blank, or send River what you baselined and get back a Benefit Register with the measurement method already written.
Edit with AI