Wrapping OKRs and KPIs with Value: Prioritisation in a Common Value Currency

Full text

1 Introduction

Most organisations that manage by objectives run two instruments side by side. Objectives and Key Results (OKRs) set direction: where the organisation intends to go in the next period. Key Performance Indicators (KPIs) control quality: how well the organisation performs the work it already does. Both instruments are mature and widely documented (Doerr, 2018; Kaplan and Norton, 1996). This paper does not argue against either of them.

The paper starts from a narrower observation made by the author of Æilus: all OKRs and KPIs carry different value, but in management systems they look equal. A list of objectives in a planning tool, a row of indicators on a dashboard, a set of key results in a quarterly review: each item occupies one line, one colour code, one status. Nothing in the format says that one objective matters ten times more than another, or that hitting one indicator quietly damages something the organisation needs.

The paper proposes a value layer over existing OKRs and KPIs. Each objective and indicator receives an explicit value estimate in Æ, the value currency of Æilus, together with the value system it belongs to and the anti-value it can produce. The goal setting process stays as it is. What changes is that priority becomes an explicit, revisable claim instead of an implicit one. The intended readers are managers, product and engineering leaders, operations and strategy staff who already run OKRs or KPIs and find that the system does not tell them what to do first.

Terms such as value, anti-value, value flow, value system and value transformer are used as defined in the vocabulary paper of this series. The Æ formalization itself is not re-derived here. It is set out in "Value Management Theory and Employee-Centric Competitive Advantage: A System-Level Perspective" (Bolshakov, ÆVRI, 2026), and in the Æilus page on formalization. References to axioms (A1 to A13) and theorems (T1 to T17) follow the numbering of the Value Management Theory (VMT) reference.

2 The problem: objectives that look equal

2.1 Direction and quality without weight

OKRs answer the question "where are we going". KPIs answer the question "how well are we doing it". Neither answers the question "how much does this matter, and to whom". In practice that third question is answered anyway, but informally: by whoever speaks loudest in the planning meeting, by whatever is easiest to measure, or by the order in which items were typed into a document.

Balanced scorecards were an early response to a related problem. Kaplan and Norton (1996) argued that financial measures alone describe the past and that organisations need measures across several perspectives to translate strategy into action. The scorecard broadens what is measured. It does not by itself express how much each measure is worth relative to the others, and it does not represent losses as a separate category. The same holds for OKRs as described by Doerr (2018): the method gives focus and alignment, but the relative weight of objectives is left to judgement outside the method.

2.2 Three consequences

When objectives look equal, three consequences follow. They are stated here as structural tendencies, not as measured effects.

• The important competes with the convenient. An objective that is easy to plan and report gets the same visual and procedural weight as one that is hard but decisive. Teams rationally choose the convenient one, because the system rewards completion, not significance.
• The strategic loses to the measurable. Strategic objectives (reduce structural risk, build a capability, pay down technical debt) are often harder to express as a number than operational ones (tickets closed, response time, conversion). If a number is the condition for being taken seriously, measurability starts to decide priority.
• Metrics are optimised and the business is not. Goodhart (1975) observed that a statistical regularity tends to collapse once pressure is placed on it for control purposes. In management terms: once an indicator becomes a target, people improve the indicator, and the indicator stops telling the truth about the thing it was meant to reflect.

2.3 The VMT reading

VMT gives a more precise account of why this happens. An OKR or a KPI is a measurement instrument, not value itself. Value is defined only inside a value system (A1, T1), and the same outcome can be read differently in different systems (A2). An indicator that improves inside one transformer, for example a support team's closure rate, can be anti-value for the customer system or for the engineering team that receives the escalations. Anti-value is an independent category, not merely the absence of value (A5), so an indicator that ignores it can rise while the system degrades. Local optimisation does not guarantee global improvement (T5), and there is no universal optimality criterion (T13). A dashboard of equal-looking objectives is therefore not neutral. It encodes an unspoken assumption that all items are equally valuable to the same system, which VMT says is almost never true.

3 Æilus as a value layer

3.1 What wrapping means

The author's formulation is short: Æilus does not replace OKRs or KPIs, it wraps them with value. The procedure has three steps. First, the organisation defines any set of OKRs and KPIs it wants. Second, each of them receives a value estimate in Æ. Third, Æ becomes a common coordinate system: a way to set priorities and the main focus of management. In the author's division of labour, OKRs say where we are going, KPIs say how well, and Æilus says why and how much it matters.

Æ is not money and is not intended to represent financial value directly. The Æilus formalization page describes it as an abstract unit that makes it possible to compare otherwise incommensurable initiatives, make managerial assumptions explicit, discuss priorities beyond functional metrics, and separate value from cost, effort or activity. It is a tool of formalization, not a claim of objective or absolute value. Aggregation in Æ implies comparability, not precision.

3.2 The minimum record per objective

A bare number in Æ attached to an OKR would repeat the error the method is meant to correct: it would become one more metric. For this reason the value layer proposed here requires four fields for every wrapped objective or indicator.

1. Value system. Whose value is estimated. A KPI on release frequency may carry value in the product organisation's value system and in the customer's value system, with different estimates. The paper requires the value system to be named explicitly, together with its boundary: which actors and flows are inside it (A1, T15).
2. Planned value in Æ. The sender's interpretation: what the owner of the objective expects it to deliver if achieved. This is planned value in the VMT sense, not a forecast of fact.
3. Anti-value in Æ. The anti-value the objective can produce as a by-product, and the anti-value that grows if it is neglected. Technical debt, delay and waiting, accumulated risk, architectural degradation and coordination overhead are examples the formalization names explicitly.
4. Owner and receiver. Who owns the estimate (normally the Value Transformer Owner, VTO, of the transformer that pursues the objective) and which actor will confirm realized value at the end of the period.

With these four fields, an objective is no longer a line in a list. It is a claim about a value flow: which transformer intends to deliver which value element to which receiver, at what expected value and what expected cost in anti-value.

3.3 Comparing the incomparable

The practical promise of the value layer is that growth, stability, technical debt and quality can be put on one scale. Without such a scale, these goals are compared by narrative: growth has a revenue story, stability has an incident story, technical debt has an engineering story, and the story told best wins. With a scale, each story has to be expressed as a claim in Æ that others can read and challenge.

VMT sets the limit of this promise. Value types are commensurable, partially commensurable or incommensurable (A13). Æ does not remove partial commensurability. It forces the organisation to state how it handles it. In Æilus v2.0 this is the role of the Value Aggregation Policy, owned by the Value System Owner (VSO): a document that defines how multi-domain value is compared, aggregated and traded off, with priorities, admissibility thresholds and trade-off rules. Principle P9 requires growth decisions to state the aggregation assumptions used to justify them. Wrapping OKRs and KPIs is in effect the first and most visible use of that policy.

Some objectives will not be tradeable at all. A security or continuity constraint owned by a Domain Owner may be an admissibility threshold rather than an item to be weighed. The value layer should record such items as constraints, not assign them an Æ figure that invites trade-off.

4 Making assumptions explicit and revisable

4.1 Planned versus realized value per quarter

An Æ estimate made at the start of a quarter is a planned value. VMT states that planned and realized value generally diverge (A4) and that zero divergence is an unstable special case (T3). The value layer therefore has two columns, not one. At the start of the period the owner records planned value and expected anti-value. At the end of the period the receiving actor records realized value and the anti-value actually observed. Later, a retrospective value can be recorded when the longer consequences become visible, for example when a stability objective turns out to have prevented a class of incidents over the following year.

This turns the value layer into a learning loop rather than a scoring system. The question at the quarterly review is not only "did we hit the key result" but also "was our estimate of its value right, who confirmed it, and what should we change in the next estimate". The Planned vs Realized Value Gap Report and the Value Retrospective event of Æilus v2.0 are the natural places for this conversation.

4.2 Interpretation as a managed object

Making the estimate explicit also makes it attackable, which is the purpose. Under principle P10, interpretations of planned and realized value must be formed in good faith, criteria must be explicit, and realized value must be validated by the receiving actor. Systematic inflation of planned value or deflation of realized value is treated as a methodological violation. An Interpretation Change Log records when and why an estimate was revised. Over several quarters this log shows whether an organisation's estimates are converging on what receivers confirm, or drifting.

5 From v1 to v2

The first generation of Æilus (v1.x, 2022) already prioritised work using value-to-cost ratios and value-based reporting. According to the Æilus versions page, its application across different domains exposed several limitations. Value was treated as measurable but was not formally defined, so stakeholders used incompatible interpretations without noticing. Measurement instruments such as points, money and KPIs were sometimes mistaken for value itself, which led to metric-driven distortions. System boundaries were unclear. Differences between planned and realized value were observed but not explained.

These limitations matter directly for the approach in this paper. A naive v1-style wrapping, a single "value score" per OKR, risks exactly the confusion the versions page describes: the Æ figure becomes another KPI to be optimised. Æilus v2.0, built on VMT, changes four things that the value layer depends on:

• value is defined formally through an ontology and axioms, so every estimate must name its value system;
• interpretation is explicitly managed, with planned, realized and retrospective value kept apart;
• sustainability is a primary constraint, so an objective that raises planned value while violating participation conditions (A8, A9) is not admissible however high its Æ;
• responsibility is role-based: the VTO owns transformer-level estimates, the VSO owns the aggregation policy and the system-level balance between growth and stability.

In short, v1 offered a way to rank work by value. v2 adds the conditions under which such a ranking can be trusted and challenged.

6 Adoption procedure

The following procedure is a methodology design derived from the sources above. It has not been validated empirically.

1. Fix the value system. The VSO names the value system to which the objectives belong, its actors and its boundary. If objectives serve several systems, for example the organisation and its customers, each system is named, and the paper recommends estimating separately rather than blending.
2. Keep the existing goals. Take the current OKRs and KPIs as they are. Do not rewrite them to fit the value layer. Wrapping that requires rewriting goals is a redesign, not a wrap.
3. Agree an initial aggregation policy. The VSO drafts a short Value Aggregation Policy: which domains are compared on one scale, which are admissibility thresholds, how partial commensurability is handled, and what reference anchor gives Æ its scale (for example, one agreed reference objective set at a round number).
4. Estimate planned value and anti-value. Each VTO records, for each objective or indicator, the four fields of section 3.2. Estimates are made relative to the reference anchor, not in isolation.
5. Align interpretations. Hold an Interpretation Alignment Session in which receivers see the estimates and challenge them. The aim is explicit criteria, not consensus.
6. Prioritise by value and anti-value. Order objectives by net expected value under the aggregation policy, with constraints applied first. Record where the order differs from the original plan and why.
7. Record realized value. At period end, receivers record realized value and observed anti-value. Gaps are discussed, not punished.
8. Revise. Update estimates, the aggregation policy and the Interpretation Change Log. Retrospective value is added when longer consequences appear.

7 Illustrative example

The table below is entirely hypothetical. It describes an imagined software product organisation in one quarter. The value system is the product organisation's value system, with its customers and engineering teams as actors inside the boundary. All figures are round, invented numbers chosen to show the mechanics. They are not findings, measurements or recommendations.

Read as a list of equal objectives, these five items would compete for the same attention. Read with the value layer, three things become visible in this invented case. First, availability is not weighed at all: under the hypothetical aggregation policy it is an admissibility threshold owned by a Domain Owner. Second, the technical debt objective has a modest planned value but a large anti-value if neglected, so its net importance is higher than its planned value suggests. Third, the closure-rate KPI carries high anti-value if pursued, because pressure on closure can push unresolved issues back to customers and to engineering. That is the Goodhart pattern stated in value terms.

The end-of-quarter column illustrates the asymmetry of plan and fact (A4). The growth objective realized less than planned. The documentation objective realized more. Neither result is a failure of the method. Both are inputs to the next estimate and to the Interpretation Change Log.

8 Risks and safeguards

8.1 False precision

A number in Æ looks exact even when it is a rough judgement. Readers may treat a difference of 40 and 45 Æ as meaningful. VMT is clear that aggregation implies comparability, not precision, and that interpretation can scale without confirmation by realization (T16). Safeguards: use coarse, round values; prefer ordering and ratios over decimals; show planned and realized value side by side so that the size of typical gaps calibrates how much a figure can be trusted; never present Æ as financial value.

8.2 Gaming the Æ estimate

Any figure that decides priority invites manipulation. A team that wants its objective funded can inflate planned value; a team that wants to look good can understate anti-value. This is the Goodhart problem moved one level up, and interpretational distortion of this kind accumulates anti-value (T17). Safeguards from Æilus v2.0: realized value is recorded by the receiver, not the sender (P10); systematic inflation is a methodological violation, not a negotiation tactic; the Interpretation Change Log makes each revision traceable; the Æilus Master watches for cargo-cult adoption; and the VSO owns the aggregation policy, so no single transformer sets the scale.

8.3 Replacing judgement with a score

A ranked list can become a way to avoid discussion. The value layer is meant to make discussion possible, not to end it. Constraints, participation conditions and domain policies sit outside the ranking. When the ranking and experienced judgement disagree, the disagreement is a signal to revisit the estimate, the boundary or the aggregation policy (T9, T15).

8.4 Maintenance cost

Æilus states that every artifact must have an owner, an update loop and a decision purpose, otherwise it becomes passive documentation and creates anti-value through maintenance cost and false confidence. A value layer that nobody updates after the first quarter is worse than none.

9 Limitations

This paper is conceptual. It is built from the author's published materials on Æilus and VMT and contains no empirical data. The adoption procedure and the example are designs, not reports of practice. The paper does not show that wrapped OKRs lead to better outcomes than unwrapped ones; that would require comparative studies over several periods, with planned and realized values recorded by independent receivers. The approach depends on the quality of the aggregation policy, which this paper describes only in outline. It also assumes a minimum of trust: in organisations where estimates are routinely used against their authors, P10 will be hard to sustain. Finally, Æ is designed for internal complex systems. It does not replace economic models of markets, price formation or financial forecasts.

10 Conclusion

OKRs set direction and KPIs control quality, but neither says how much an objective matters or to whom. When that question is left implicit, the convenient beats the important, the measurable beats the strategic, and metrics improve while the business does not. Æilus addresses this by wrapping existing goals with value: each objective and indicator receives a planned value and an anti-value estimate in Æ, tied to a named value system and an owner, and later confronted with realized value confirmed by the receiver. The method does not make value objective. It makes the assumptions behind priorities explicit, comparable and revisable, under an aggregation policy owned at system level. Its main risks, false precision and gaming, are the same ones that affect any metric, and the safeguards are those Æilus v2.0 introduced after v1 showed that instruments can be mistaken for value. Goals can be as many as an organisation wants. The value layer is a disciplined way of saying which of them matter, and of checking that claim every quarter.

References

• Bolshakov, V. Value, Anti-Value and Value Flows: A Working Vocabulary for Practitioners. ÆVRI, 2026.
• Bolshakov, V. Value Management Theory and Employee-Centric Competitive Advantage: A System-Level Perspective. ÆVRI, 2026.
• Bolshakov, V. Æilus does not replace OKR or KPI. It wraps them with value. aeilus.tech (news, ID 801), 2025.
• Bolshakov, V. Æilus as a Formalization. aeilus.tech/methodology/aeilus-as-a-formalization/, 2026.
• Bolshakov, V. Evolution of Æilus. aeilus.tech/methodology/versions/, 2026.
• Bolshakov, V. Æilus Methodology v2.0: Principles, Roles, Processes and Events, Artifacts and Reports. aeilus.tech/methodology/, 2026.
• Bolshakov, V. Theory of Value Management (VMT). aeilus.tech/theory-of-value-management/, 2026.
• Doerr, J. Measure What Matters. 2018.
• Goodhart, C. A. E. Problems of Monetary Management: The U.K. Experience. 1975.
• Kaplan, R. S. and Norton, D. P. The Balanced Scorecard: Translating Strategy into Action. 1996.