River
Y CombinatorBacked by Y Combinator
FREE TEMPLATE

Productized Consulting Service Design

Five sheets and four documents that measure how often each deliverable recurs against how much of it survives, then price the spread.

Free download  ·  No account needed

Variation Analysis  ·  Coldharbour Advisory  ·  9 engagements of one service, 30 months

Two numbers per component, and they disagree

ComponentInReuseMeanWhere it lands
Report and board pack9 of 983%28.3Core, templated
Interview programme9 of 981%32.7Core, templated
Data request and validation9 of 978%25.2Core, templated
Technician utilisation9 of 922%30.0Core, input request
First-time-fix analysis8 of 931%20.4Core, input request
Travel and route density7 of 924%16.3Core, input request
Contract profitability5 of 961%23.2Priced option
Warranty and recall2 of 912%26.5Named out of scope
Reuse highReuse low
In nearly every oneCore, and a template earns its keep. 10 components, 188.4hCore, but a slot with a specified input. 3 components, 66.7h
In some of themPriced option module. 2 components, 39.2hOut of scope, named on the page. 5 components, 119.5h

The three lowest-reuse rows on a twenty-row sheet are all core, and all three are the analysis. The practice had called those three its product, and the data request, interview programme and report its plumbing. Reuse says 26 percent against 81.

Page one for this query is one instruction, repeated. Audit your last ten to twenty engagements, find the high-frequency low-customization work, package it at a fixed scope and a fixed price, then build an intake questionnaire and a delivery checklist. The most advanced version asks what you could remove and still deliver 80 percent of the value. Every one of those audits is done from memory, at the level of the whole engagement, and nothing in it is counted.

So frequency gets mistaken for reuse. Here every past engagement is broken into deliverable components and each one gets two separate numbers: how many engagements contained it, counted off the record, and how much of the artifact carried forward unchanged, measured off the files. Where the two disagree is the productization decision, and they disagree on the analysis. Coldharbour Advisory's three analysis components sit in nine, eight and seven of nine engagements at 26 percent average reuse, the three lowest rows on a twenty-row sheet.

The second half is the price. A fixed fee cannot be adjusted afterwards for what the work cost, so it is set at the p90 of the past engagement totals rather than their mean. That is 273 hours against 245: a $58,500 fee, where pricing the mean would have given $52,645 and four of nine would have run over it. The floor behind it comes from the rate card pack, the practice around it from the practice dashboard, and the boundary becomes terms in the statement of work.

255 hours at the mean, 302 at the p90, and a 41 percent narrower spread once two drivers move out

The Variation Analysis, Standard Effort Model, Scope Boundary and Triggers, and Deliverable Template Register sheets.

Variation Analysis

Inclusion off the engagement record, reuse off the delivered files. CV decides whether a driver has to be named.

RefComponentInReuseMeanCVDecision
C-01Data request and validation9 of 978%25.20.18Templated
C-02Job volume and mix baseline9 of 971%14.40.12Templated
C-03Technician utilisation9 of 922%30.00.29Input request
C-04First-time-fix analysis8 of 931%20.40.20Input request
C-05Travel and route density7 of 924%16.30.27Input request
C-06Scheduling policy review9 of 966%12.00.15Templated
C-10Interview programme9 of 981%32.70.08Templated
C-11Ride-alongs and observation8 of 974%18.00.10Templated
C-12Comparator benchmarking7 of 969%12.70.13Templated
C-13Findings workshop9 of 977%9.70.15Templated
C-14Report and board pack9 of 983%28.30.11Templated
C-15Savings model and case8 of 972%20.20.12Templated
C-16Implementation roadmap7 of 968%15.10.15Templated

Thirteen core rows shown. Two more are priced options: parts and van stock (6 of 9, 64%) and contract profitability (5 of 9, 61%). Five leave the product entirely and are named on the client page: subcontractor cost (4 of 9, 18%), telematics assessment (3 of 9, 41%), rate card review (3 of 9, 16%), warranty exposure (2 of 9, 12%) and depot network (2 of 9, 22%). Those five are 119.5 mean hours that used to be absorbed.

Standard Effort Model

The thirteen core components totalled per engagement, before and after two drivers move out of the fixed fee. Nine of the 117 core cells are imputed at the component mean and flagged.

EngagementJob-level labour time stampCore beforeCore after
E-07 Otford Utilitiesyes215.0214.7
E-03 Penshurst Plant Hireyes222.9222.9
E-01 Halstow Building Servicesyes235.0233.7
E-09 Trottiscliffe Wateryes245.0242.7
E-05 Ightham Lift Servicesyes249.0245.7
E-02 Ockley Facilitiesyes254.9241.5
E-06 Chevening Refrigerationpartial269.1255.3
E-04 Brasted Mechanicalno302.0275.2
E-08 Ryarsh Fire Systemsno303.0272.2
Mean of the nine255.1244.9at $215, a $52,645 fee
p90 of the nine totals302.2272.8at $215, a $58,645 fee
Spread, mean to p9047.127.9a 41% narrowing
Range, low to high88.060.5published at $58,500

Set at the mean instead, four of the nine would have run over the fee, the worst by 30.3 hours and $6,516 of margin. Summing the thirteen component p90s gives 311.3 hours, only 3 percent above the engagement p90, because three of them peak on the same engagement for the same missing field.

Scope Boundary and Triggers

Every component and every driver placed. This sheet is what becomes the client-facing page, so a component with no home here is a component that gets absorbed.

Priced options and named exclusions

ComponentWhere it sitsPriceCondition
Parts and van stockPriced option$3,750Van stock on the depot item master
Contract profitabilityPriced option$5,500Revenue allocated to jobs
Subcontractor costOut, namedQuotedRate structures differ every time
Telematics assessmentOut, namedQuotedSystems differ every time
Warranty exposureOut, namedQuotedClient also installs
Depot network reviewOut, namedQuotedDepots consolidating anyway
Rate card reviewOut, namedQuotedCommercial model differs

Where the price changes

TriggerTest, before signaturePrice
Utilisation rebuilt from shift dataNo job-level labour time$4,500
Extracts late past one reissueObligation in the definitionWindow extends
Over four depots or fifteen interviewsDepot count on the extract$3,500 each

T-1 is the entry condition doing the work. Rebuilding utilisation with no job-level time stamp cost 20.5 hours on the two engagements where it happened, both absorbed. It is visible on a screen share, so it is quoted before signature.

Deliverable Template Register

Reachable hours is mean hours times reuse rate: the effort a template can touch.

RefArtifactMeanReuseReachStatus
C-10Interview programme32.781%26.5Built
C-14Report and board pack28.383%23.5Built
C-01Data request and validation25.278%19.7Build next
C-15Savings model and case20.272%14.6Build next
C-11Ride-alongs and observation18.074%13.3Build next
C-16Implementation roadmap15.168%10.3Build next
C-02Job volume and mix baseline14.471%10.3Build next
C-12Comparator benchmarking12.769%8.8Build next
C-06Scheduling policy review12.066%7.9Build next
C-13Findings workshop9.777%7.4Built
C-03Technician utilisation30.022%6.6Do not build
C-04First-time-fix analysis20.431%6.3Do not build
C-05Travel and route density16.324%3.9Do not build

159.1 of the core's 255.1 mean hours are reachable and 33.7 are captured, 13 percent of the core. The three templates that exist are the three artifacts that most looked like the deliverable. The data request sits third here and has never had one.

The largest analysis component is 30 mean hours and 6.6 reachable, so templating it is days of work to save most of a day a year. It is in 9 of 9 engagements, which is why frequency alone picks it first.

What's in the pack

01

Variation Analysis

One row per component carrying inclusion, reuse, mean hours, range, coefficient of variation, the named driver behind the spread, and which of four cells it lands in.

02

Engagement Component Matrix

The raw evidence: every component against every past engagement with hours in the cell, imputed cells flagged, plus the fee and effective rate each one collected.

03

Standard Effort Model

The core at the mean and at the p90, before and after the drivers move, with the fixed fee derived from the second figure and the floor rate it has to clear.

04

Scope Boundary and Triggers

Every component placed as in the fee, a priced option, or named out of scope, plus each trigger with a test that is checkable before signature and a price.

05

Deliverable Template Register

The artifacts ranked by reachable hours, mean hours times reuse rate, so templates get built in the order the arithmetic gives rather than the order that feels right.

06

Service Definition

The client-facing boundary: the fee, the window, what arrives, what the client provides, what is named out of scope with its price, and where the price changes.

07

Delivery Playbook

The internal version: the entry check before signature, the week-by-week hour allocation, and how the low-reuse analysis components get a specified input instead of a template.

08

Sales Materials

The one-page summary and the objections, with every claim traced back to a row on the variation analysis rather than to a number somebody liked the sound of. A live bid still goes through the proposal pack.

09

How the Product Was Derived

The full worked example: nine engagements, twenty components, two drivers moved, the fee at the p90, and what the product deliberately does not fix.

10

Frequency is not reuse, and the difference is the product

The standing rule. No component enters the core on inclusion alone, and a wide component gets a named driver before the fee is published. A driver that fires later is a change order.

How to use it

  1. 1

    Open in River, or take it blank

    Open the pack and send the past engagements of one service, or download the four Word documents and five CSV sheets and work the arithmetic yourself.

  2. 2

    Decompose into components, not phases

    One artifact per component, fine enough that client-specific variation lands in its own row. Then hours per component per engagement, at whatever grain the records hold.

  3. 3

    Count inclusion, measure reuse

    Inclusion off the engagement record. Reuse off the delivered files against the previous client's version of the same artifact, counting adapted content as half.

  4. 4

    Price the p90, then move the drivers

    The fee comes from the p90 of the engagement totals, because a fixed price is one not subject to adjustment for the contractor's cost experience. Then each named driver becomes an obligation, an exclusion or a trigger.

Frequently asked questions

Is this template free?

Yes. The download is Word documents and CSV sheets, no account and no card. Edit with AI is the optional half: River decomposes your past engagements, measures both numbers per component and derives the fee. The rest sit in the template library.

What format are the downloaded files?

Four Word documents and five CSV sheets in one zip. The sheets carry the columns that do the work: inclusion, reuse rate, mean and range of hours, coefficient of variation, the named driver, and the four-way decision. Excel, Numbers and Google Sheets open them directly.

How many past engagements do I need?

Eight or more makes a p90 worth publishing a price against, because a fixed price belongs where the risk can be predicted with an acceptable degree of certainty. Four to seven is directional and the sheet says so on the row. Below four, one bad engagement moves the p90 further than any driver you could name.

Why measure reuse when I already know what recurs?

Because the two answers differ, and predictably. In the worked example the three components the practice called its product average 26 percent reuse while the three it called plumbing average 81. Templating on frequency alone builds the wrong three artifacts first.

Why price at the p90 instead of the average plus a buffer?

A fixed price is one that places maximum risk and full responsibility on the contractor, so you hold the whole distribution rather than its middle. Set at the mean, four of the nine past engagements would have run over the fee, the worst by $6,516 of margin.

What if a component varies too much to include?

Name the driver, then choose an instrument. A fixed price is defensible where the uncertainties are identified and their cost impact estimated. Rebuilding utilisation with no job-level labour time cost 20.5 hours twice, so it is quoted at $4,500 rather than absorbed.

Does a fixed fee mean I earn less?

Not in the example. The fee plus options at the historical attach rate is $64,056 against a bespoke mean of $62,778. What changed is the boundary: 119.5 mean hours of previously absorbed work are named and quoted, and the effort range fell from 88 hours to 60.

Measure it before you package it

Take the Word documents and CSV sheets blank, or send River the past engagements of one service and get the four-cell decision and a fee with a distribution behind it.

Edit with AI