Reference
Frameworks behind kpi-architecture
Part of the KPI Architecture skill · loaded on demand from SKILL.md
Read this when you need the original method in depth: why Kaplan and Norton built the scorecard the way they did, how a strategy map is constructed step by step, what the public McKinsey material actually says about tracking transformation value, and where each method breaks.
Contents
- The Balanced Scorecard (Kaplan and Norton, HBR 1992)
- From measurement system to management system (1996)
- Strategy maps (HBR 2000; book 2004)
- Lead and lag measures, and the cause-and-effect hypothesis
- Measure selection: the published guidance and the practical rules
- Value realisation tracking: what McKinsey has published
- Stage gates L0 to L5 as described publicly
- The three-tier metric hierarchy: milestone, operational, value
- Goodhart, Campbell and the gaming problem
- Where each framework fails
- Source list
1. The Balanced Scorecard (Kaplan and Norton, HBR 1992)
Robert S. Kaplan and David P. Norton, "The Balanced Scorecard: Measures That Drive Performance", Harvard Business Review, January-February 1992. The article came out of a year-long study of a dozen companies (the "Measuring Performance in the Organization of the Future" project sponsored by KPMG's Nolan Norton Institute).
What it claims. Financial measures alone are a lagging, backward-looking view of a business. They report the consequences of decisions made months or years earlier and say nothing about whether the organisation is building the capabilities that will produce next year's results. Managers who steer on financial measures alone over-invest in short-term profit and under-invest in customers, processes and people. A scorecard that "balances" financial with non-financial measures, and outcomes with drivers, gives a "fast but comprehensive view" and forces trade-offs to be visible.
The four perspectives and the question each asks (1992 wording).
| Perspective | Question in the article | What it captures |
|---|---|---|
| Financial | How do we look to shareholders? | Survival, profitability, growth, shareholder value |
| Customer | How do customers see us? | Time, quality, performance and service, cost, as the customer experiences them |
| Internal business | What must we excel at? | The processes, decisions and actions that produce the customer outcomes |
| Innovation and learning | Can we continue to improve and create value? | The organisation's ability to launch new products, create value, improve efficiencies |
In the 1996 book and later work the last two were renamed "internal business process" and "learning and growth". The mechanism is unchanged.
The mechanism. Three things make the scorecard work when it works.
- Forced balance. Putting the four perspectives on one page makes it impossible to hit a financial target by quietly degrading a customer or process measure without the degradation being visible. The scorecard is a device for seeing trade-offs, not for avoiding them.
- Limiting the set. The original claim is that a handful of measures per perspective, chosen because they matter for the strategy, beat a hundred measures chosen because the data exists. Kaplan and Norton's later practical guidance settled around 20 to 25 measures in total for a business unit.
- Translating strategy into operational terms. Each measure is supposed to be derived from a strategic objective, so that arguing about the measure is a way of arguing about the strategy. This is what separates a scorecard from a KPI dashboard.
The failure it was built to prevent. Managing the business through the rear-view mirror: cost-cutting and financial engineering that looks good for six quarters and hollows out the franchise.
2. From measurement system to management system (1996)
Kaplan and Norton, "Using the Balanced Scorecard as a Strategic Management System", HBR January-February 1996, and the book The Balanced Scorecard: Translating Strategy into Action (Harvard Business School Press, 1996).
The 1996 work reframed the scorecard from a measurement tool to a management process with four loops:
- Translating the vision. Turning strategy statements into objectives and measures that people can act on.
- Communicating and linking. Cascading objectives to departments and individuals, and linking rewards to measures.
- Business planning. Aligning budgets and initiatives with scorecard targets, so that the budget funds the strategy rather than last year's cost base plus inflation.
- Feedback and learning. Reviewing not only whether targets were hit but whether the cause-and-effect assumptions behind the scorecard held. This is "double-loop learning" (the term comes from Argyris): the review can change the strategy, not only the actions.
The important addition for kpi-architecture is the third loop. A scorecard without an initiative portfolio and a budget link is a poster. This is where the value-realisation ledger (section 6) attaches.
3. Strategy maps (HBR 2000; book 2004)
Kaplan and Norton, "Having Trouble with Your Strategy? Then Map It", HBR September-October 2000, and Strategy Maps: Converting Intangible Assets into Tangible Outcomes (Harvard Business School Press, 2004). The Strategy-Focused Organization (2001) sits between them.
What a strategy map is. A one-page diagram of the strategic objectives (not measures) in the four perspectives, connected by arrows that state cause-and-effect hypotheses. Read from the bottom up: learning-and-growth objectives enable internal-process objectives, which deliver the customer value proposition, which produces the financial outcomes.
The standard architecture in the 2004 book.
- Financial perspective splits into a productivity strategy (improve cost structure, increase asset utilisation) and a growth strategy (expand revenue opportunities, enhance customer value). Both feed long-term shareholder value. The tension between them is deliberate: the map should show which one leads in the strategy period.
- Customer perspective states the value proposition. Kaplan and Norton borrow Treacy and Wiersema's three generic propositions (operational excellence: lowest total cost; product leadership; customer intimacy / complete customer solutions) and add "system lock-in". The proposition is described by attributes: price, quality, availability, selection, functionality, service, partnership, brand.
- Internal perspective groups processes into four clusters: operations management, customer management, innovation, and regulatory and social. A strategy emphasises one or two clusters; the map should make that emphasis visible.
- Learning and growth has three components: human capital (strategic job families and their readiness), information capital (systems and data the strategic processes need), and organisation capital (culture, leadership, alignment, teamwork).
The construction procedure (top-down).
- Start from the financial objectives and the time frame (what the board must see in the P&L, by when).
- Choose the customer value proposition and state it as objectives the customer would recognise.
- Identify the few internal processes at which the organisation must excel to deliver that proposition. These are "strategic themes"; three or four is typical.
- Identify the human, information and organisation capital those processes require, and where the current gap is.
- Draw the arrows. Each arrow is an if-then statement that someone in the room is willing to defend.
- Only then choose measures: at least one for each objective, so that the arrow between two objectives becomes a testable pairing of a lead measure and a lag measure.
The mechanism. The map converts a strategy narrative into a set of falsifiable hypotheses. If the map says "pricing capability in the sales force leads to higher pocket margin", then the scorecard can show, quarter by quarter, whether the lead measure moved and whether the lag measure followed. When it does not follow, either the execution failed (the lead measure did not move) or the hypothesis is wrong (it moved and nothing happened). The map is what lets you tell the two apart.
The failure it was built to prevent. Scorecards with 30 unlinked measures where nobody can say which ones drive which, so the quarterly review is a recital of numbers rather than a test of the strategy.
4. Lead and lag measures, and the cause-and-effect hypothesis
Kaplan and Norton use "outcome measures" (lag) and "performance drivers" (lead). The distinction is relative, not absolute: gross margin is a lag measure for a pricing initiative and a lead measure for EBITDA.
Rules that follow from the mechanism:
- Every arrow on the map should be represented by at least one lead measure at the tail and one lag measure at the head. If you cannot name the pair, the arrow is an aspiration.
- Lead measures must move faster than lag measures. If both move on the same cycle, the "lead" is not giving early warning.
- Lead measures are the ones owners can act on within the review period. Lag measures are the ones the board cares about. A scorecard with only lag measures cannot be managed; one with only lead measures cannot be trusted.
- The time lag between the two should be stated (for example "two quarters") because that is the point at which failure of the lag to follow the lead becomes evidence against the hypothesis rather than patience.
5. Measure selection: the published guidance and the practical rules
Kaplan and Norton's published guidance is sparse on selection rules, so the practical rules below are consulting practice consistent with the mechanism rather than a quotation.
Few. Around 20 to 25 measures for a business unit scorecard, four to seven per perspective. McKinsey's 2017 transformation study found that only 29 percent of the metrics companies tracked were actually used during the programme; the rest were noise that cost time to collect and diluted attention.
Causal. Every measure sits on a map objective, and every objective is on a chain that reaches the financial perspective. An orphan measure is deleted or the map is wrong.
Owned. One named person per measure who can explain the number and is accountable for moving it. Committees do not own measures. If two people share a measure, split it or pick one.
Actionable. The owner can move the measure within the review cycle with resources they control. "Market share" is rarely actionable for a branch manager; "share of wallet in assigned accounts" may be.
Definable. A formula, a data source, a frequency, a baseline and a target exist and are written down in a measure-definition sheet. A measure that cannot be defined precisely will be reinterpreted every quarter.
Balanced. Lead and lag; financial and non-financial; outcome and health (quality, safety, engagement) so that hitting the outcome by breaking something shows up.
Gaming-resistant. See section 9. Pair any measure that can be gamed with the measure that would reveal the gaming (revenue with margin, closure rate with reopen rate, orders within pricing guidance with discount exceptions granted).
6. Value realisation tracking: what McKinsey has published
The public sources are McKinsey articles, chiefly:
- Michael Bucy, Tony Fagan, Benoit Maraite and Cornelia Piaia, "Keeping transformations on target", McKinsey, March 2017. Based on a review of 18 transformations at 13 organisations drawn from McKinsey's transformation database.
- Kevin Laczkowski, Tao Tan and Matthias Winter, "The numbers behind successful transformations", McKinsey, October 2019.
- McKinsey's broader writing on the "transformation office" and the "chief transformation officer" role.
What these sources describe, paraphrased:
A transformation office (TO). A small central team, often led by a chief transformation officer reporting to the CEO, that runs the initiative pipeline. It sets a weekly review cadence to monitor initiative progress against milestones, records the value captured, and unblocks owners. The 2017 article's rule of thumb is that around 80 percent of initiatives should receive a specific action update each week; initiative owners typically carry about three initiatives at a time.
An initiative-level value ledger. Every initiative has an owner, a baseline, a target, a stage, milestones with dates, and a value figure that is validated by finance rather than asserted by the owner. Value is tracked at the initiative level, then aggregated, then reconciled to the P&L. Finance signs off at the gates where value is claimed.
Value attrition and pipeline coverage. The 2017 article reports that initial impact estimates fall by roughly 70 percent cumulatively between L1 and L5. The implication the authors draw is that a programme needs a pipeline worth more than three times its target at the early stages to land the target. The 2019 article adds that companies in the top quartile set gross targets of 75 percent or more of trailing earnings and "restocked" their initiative portfolio by about 70 percent after year one.
Sand versus boulders. In both articles, roughly half of programme value comes from small initiatives (each under about 0.5 percent of total value). Programmes that track only the large initiatives miss half the value and lose the engagement of the wider organisation.
Timing. Initiatives typically start about 2.5 weeks late and finish about four weeks late. Value phasing should therefore be planned month by month and re-forecast, not assumed from the business case.
Metrics discipline. Only 29 percent of tracked metrics were actually used. Fewer, better-defined metrics reviewed on a fixed rhythm beat comprehensive dashboards.
What these sources do not provide, and this skill does not invent: internal templates, proprietary software details, or client-specific figures. Where the skill describes a ledger layout or a cadence, that is consulting practice consistent with the published material.
7. Stage gates L0 to L5 as described publicly
The 2017 article describes six stages. Paraphrased definitions, with the gate test this skill applies:
| Stage | Public description (paraphrased) | Gate test to enter the stage | Who signs |
|---|---|---|---|
| L0 Idea | Collect as many ideas as possible regardless of feasibility or size | Logged with a rough size and a sponsor name | TO |
| L1 Validated | Worth pursuing; the owner has validated the value assumptions and feasibility | Owner named; value hypothesis with a baseline and rough mechanism; no obvious blocker | Owner and TO |
| L2 Business case | A solid business case built and approved by finance | Baseline agreed with finance; target and phasing signed; double-count check passed; one-off costs stated; milestones drafted | Finance |
| L3 Planned and in execution | Milestones defined, monthly value capture scheduled | Milestone plan with dates and owners; lead measures set with a target; value phased month by month; first milestone started | TO and sponsor |
| L4 Implemented | All milestones completed; finance assesses value delivery | All milestones closed; value visible in the operational lead measures; finance test of P&L visibility under way | TO |
| L5 Realised | Realised value appears in the financials and is reasonably certain to persist | Value identified in management accounts against the agreed baseline; sustained for an agreed period (commonly three months); budget or forecast adjusted so the value cannot be spent twice | Finance (CFO or controller) |
Two rules that make the gates useful rather than bureaucratic:
- Value counted is value at L5 only. Everything earlier is forecast. Reports must show the stage split, so that "we have identified $40M" cannot be read as "we have delivered $40M".
- Gates move backwards. An initiative that stalls at L3, or whose L4 value fails the finance test, goes back a stage and the ledger shows the movement. Hiding regressions is what makes ledgers untrustworthy.
8. The three-tier metric hierarchy: milestone, operational, value
A practical structure that reconciles the scorecard (section 1 to 4) with the ledger (sections 6 and 7):
| Tier | What it measures | Example | Where it lives | Frequency |
|---|---|---|---|---|
| Milestone (activity) | Whether the work is being done | Sepsis order set live in all five hospitals by 30 June | Initiative plan | Weekly |
| Operational KPI (lead) | Whether the work is changing behaviour or performance | Pathway order-set adherence, percent | Scorecard, internal or learning perspective | Weekly or monthly |
| Value (lag) | Whether the change is showing in the financials | Supply cost per case reduction in operating EBITDA, dollars | Ledger and scorecard financial perspective | Monthly, validated by finance |
Activity metrics are necessary for execution management and useless as evidence of value. The most common failure in transformation reporting is presenting tier 1 as if it were tier 3 ("we have completed 85 percent of milestones" presented next to the target number, inviting the reader to assume 85 percent of the value). The hierarchy forces each metric to be labelled with its tier.
9. Goodhart, Campbell and the gaming problem
Charles Goodhart (1975) observed, in the context of monetary targets, that a statistical regularity tends to collapse once pressure is placed on it for control purposes. Donald Campbell (1976) made the parallel point for social indicators: the more an indicator is used for decision-making, the more it will be corrupted and the more it will distort the process it is meant to monitor.
For scorecard design this has three consequences:
- Any measure attached to pay or promotion will be optimised in the narrowest way that satisfies the definition. The definition sheet must therefore close the obvious loopholes (what counts, what is excluded, who validates).
- Pair measures so that the cheap way to hit one shows up in the other. Revenue with margin; on-time delivery with cost per delivery; call handling time with first-contact resolution; orders within guidance with exception approvals granted.
- Prefer measures that are hard to move without doing the real work. Share of wallet is harder to game than number of customer visits.
10. Where each framework fails
The scorecard fails when it is populated bottom-up from available data (every function nominates its favourite metrics) rather than top-down from the strategy; when perspectives are treated as four silos rather than a chain; when measures have no owners; or when it is reviewed as a report rather than as a hypothesis test.
The strategy map fails when every objective is connected to every other (a map with 40 arrows says nothing), when objectives are written as activities ("implement CRM") rather than outcomes ("sales force knows each account's total spend"), or when the map is drawn by consultants and the executives who must defend the arrows were not in the room.
Value tracking fails when value is counted before finance has validated it; when baselines are set after the initiative started (so the initiative claims the market's improvement); when two initiatives claim the same margin; when savings are "realised" but the budget still contains the cost so the money is spent elsewhere; or when only boulders are tracked and the sand that carries half the value is never logged.
All three fail when the measurement layer is built after the plan is fixed. Measures chosen after the fact are chosen to make the plan look good.
11. Source list
- Kaplan, R. S. and Norton, D. P., "The Balanced Scorecard: Measures That Drive Performance", Harvard Business Review, January-February 1992 (reprinted July-August 2005).
- Kaplan, R. S. and Norton, D. P., "Putting the Balanced Scorecard to Work", HBR, September-October 1993.
- Kaplan, R. S. and Norton, D. P., "Using the Balanced Scorecard as a Strategic Management System", HBR, January-February 1996.
- Kaplan, R. S. and Norton, D. P., The Balanced Scorecard: Translating Strategy into Action, Harvard Business School Press, 1996.
- Kaplan, R. S. and Norton, D. P., "Having Trouble with Your Strategy? Then Map It", HBR, September-October 2000.
- Kaplan, R. S. and Norton, D. P., The Strategy-Focused Organization, HBS Press, 2001.
- Kaplan, R. S. and Norton, D. P., Strategy Maps: Converting Intangible Assets into Tangible Outcomes, HBS Press, 2004.
- Kaplan, R. S. and Norton, D. P., The Execution Premium, HBS Press, 2008.
- Bucy, M., Fagan, T., Maraite, B. and Piaia, C., "Keeping transformations on target", McKinsey and Company, March 2017.
- Laczkowski, K., Tan, T. and Winter, M., "The numbers behind successful transformations", McKinsey and Company, October 2019.
- Goodhart, C., "Problems of Monetary Management: The U.K. Experience", 1975.
- Campbell, D. T., "Assessing the Impact of Planned Social Change", 1976.
- Treacy, M. and Wiersema, F., "Customer Intimacy and Other Value Disciplines", HBR, January-February 1993 (source of the value-proposition types used in strategy maps).