How to set KPI targets

Every year, the same ritual plays out across thousands of organisations. Finance sends down a revenue target derived from investor expectations or board ambitions. Business units negotiate their share. Department heads push back with reasons why the number is unrealistic, then eventually accept something close to it. Team leads break their allocation into sub-targets using last year as a baseline plus a percentage. The resulting targets look precise. They sit in spreadsheets with decimal places and RAG statuses. But they are built on negotiation, not structural understanding.

The problem is not that the people involved lack intelligence or good intentions. The problem is that target setting without a model of how the business actually works is guesswork dressed in the language of rigour. When the sales team is told to grow revenue by 20%, nobody asks whether the pipeline, conversion rates, and average deal sizes can structurally produce that outcome. When the marketing team is told to generate 30% more leads, nobody checks whether the website traffic, content engine, and campaign budgets can support that volume. The targets exist in isolation from the system that must deliver them.

This disconnect produces predictable consequences. Teams that receive impossible targets either burn out trying or quietly game the metrics to appear on track. Teams that receive soft targets coast. The relationship between individual targets and overall business performance becomes opaque. When the company misses its annual plan, the post-mortem devolves into blame rather than diagnosis, because nobody has a model that explains which assumptions were wrong and where the plan broke down.

The debate between top-down and bottom-up target setting is one of the oldest in business planning. Both approaches have clear logic. Both fail when used in isolation. Understanding why requires looking at what each method actually optimises for and where its blind spots lie.

The solution is not to pick one approach over the other. It is to use both simultaneously and reconcile them through a shared model. A metric tree makes this possible. The executive team sets the top-level target: revenue grows from £10M to £12M. That is the top-down ambition. Then the tree decomposes that target into its structural components: customer count, average revenue per customer, retention rate, and the sub-drivers beneath each. Individual teams estimate what they can achieve at each node. Marketing projects visitor growth and conversion improvements. Sales forecasts deal velocity and win rates. Product estimates the impact of planned features on retention and expansion.

When the bottom-up estimates are assembled and rolled up through the tree, one of three things happens. If they sum to £12M or more, the targets are validated and you can proceed with confidence. If they sum to £10.5M, there is a £1.5M gap, and the conversation shifts from "try harder" to "where specifically can we close the gap, and what would it take?" If they sum to £14M, you may have set a target that is too conservative at the top, or you may have teams that are overestimating their capacity. In every case, the tree gives you a structured way to reconcile ambition with reality, and the resulting targets are coherent because they are mathematically connected from root to leaf.

A metric tree turns target setting from an exercise in allocation into an exercise in modelling. Instead of asking "what should each team aim for?", you ask "what combination of input improvements produces the outcome we want?" The difference is fundamental. The first question invites negotiation. The second invites analysis.

Consider a straightforward example. Your current annual revenue is £10M. The board wants £12M next year, a 20% increase. Revenue decomposes into Customers multiplied by Average Revenue Per User (ARPU). You currently have 2,000 customers at £5,000 ARPU. The tree immediately reframes the question: which combination of customer growth and ARPU growth gets you to £12M?

The arithmetic reveals the trade-offs. If ARPU stays flat at £5,000, you need 2,400 customers, a 20% increase. If you can grow ARPU by 9% to £5,450, you only need 2,202 customers, roughly a 10% increase. If ARPU grows by 15% to £5,750, you need just 2,087 customers, barely a 4% increase. Each scenario implies a completely different operational plan. The first demands heavy investment in acquisition. The second balances acquisition with monetisation. The third is almost entirely a pricing and expansion play.

Without the tree, leadership might simply tell both the growth team and the monetisation team to aim for 20% improvement. That would be incoherent, because a 20% increase in both customers and ARPU would produce £14.4M, overshooting the target by £2.4M. It would also be wasteful, because it asks teams to stretch on both dimensions when stretching on one might be sufficient. The tree prevents this by making the mathematical relationships explicit.

The real power emerges when you go deeper. If you need 2,202 new customers and your current lead-to-customer conversion rate is 5%, you need 44,040 leads. If you can improve conversion to 6%, you only need 36,700 leads. That distinction determines whether you need to hire three additional marketing staff or one. It determines your paid acquisition budget. It determines whether the target is achievable with existing channels or requires opening new ones. Every level of decomposition makes the plan more specific and the assumptions more testable.

There is no single correct method for setting a KPI target. The right approach depends on the maturity of the metric, the availability of data, and the strategic context. In practice, the best targets draw on multiple methods simultaneously, using each as a cross-check against the others. The five methods below cover the full spectrum from data-rich to judgement-driven.

1. 1

   Historical trend

   Take last year's performance and extrapolate forward, often with a growth modifier. If revenue grew 15% last year, a historical-trend target might be 15-18% this year. This method is grounded in reality and easy to justify, but it anchors teams to past performance. It cannot account for step-change investments, market shifts, or compounding effects. Use it as a baseline sanity check, not as the primary method.

2. 2

   Benchmark-based

   Compare your metrics against industry peers, market averages, or best-in-class performers. If the median SaaS net revenue retention rate is 110% and yours is 95%, a benchmark-based target might aim for 105% within twelve months. This method is useful when you lack internal history or when you suspect your current performance is significantly above or below market norms. The limitation is that benchmarks are often poorly sourced, context-dependent, and lag behind real-time conditions.

3. 3

   Model-based

   Use the metric tree to calculate what is achievable given specific assumptions about each input. If you know your traffic growth trajectory, conversion rates, and retention curves, you can model the range of outcomes and set a target within that range. This is the most rigorous method because it forces you to state your assumptions explicitly, making it possible to diagnose exactly where reality diverged from the plan. It requires a well-structured metric tree and reliable data at each node.

4. 4

   Aspiration-based

   Set a stretch goal that represents a step change in performance, often tied to a strategic milestone. "Reach £50M ARR to qualify for Series C" or "achieve 90% customer satisfaction to unlock enterprise deals." Aspiration-based targets can energise teams and signal strategic intent, but they are dangerous when disconnected from a model. A stretch goal that is structurally impossible does not motivate. It demoralises. Use aspiration-based targets only when paired with a model-based reality check.

5. 5

   Constraint-based

   Determine the minimum performance needed to keep the business viable or to achieve a critical objective. What is the lowest acceptable retention rate before unit economics turn negative? What is the minimum revenue growth rate needed to avoid a down round? Constraint-based targets define the floor rather than the ceiling. They are particularly useful for defensive metrics where the goal is not to maximise but to stay above a threshold. Every organisation should know its constraint-based targets even if it aims far above them.

The strongest target-setting processes use all five methods as inputs into a single conversation. The historical trend tells you where momentum is carrying you. Benchmarks tell you where the market sits. The model tells you what is structurally achievable. Aspirations tell you what the business needs strategically. Constraints tell you what the business cannot afford to drop below. The final target should sit within the range defined by these five perspectives, with the model-based method carrying the most weight because it is the most testable and the most directly connected to the levers your teams actually control.

Even organisations that invest seriously in target setting fall into patterns that undermine the exercise. Most of these mistakes stem from treating targets as isolated numbers rather than as nodes in an interconnected system. Recognising these traps before your next planning cycle can save months of misaligned effort.

The language organisations use about targets reveals their culture. In blame cultures, targets are commitments. "You committed to 20% growth and you delivered 14%. What went wrong?" In learning cultures, targets are hypotheses. "We predicted 20% growth based on these assumptions. We achieved 14%. Which assumptions were wrong and what does that teach us?" The difference is not semantic. It fundamentally changes how people behave.

When targets are treated as commitments, teams optimise for hitting the number at all costs. They game metrics, hide bad news, avoid ambitious goals they might miss, and focus on short-term results even when they come at the expense of long-term health. When targets are treated as hypotheses, teams are incentivised to set honest predictions, surface problems early, and learn from variances rather than disguise them. The quality of information flowing through the organisation improves dramatically because people are not afraid of what the data will show.

This reframing is grounded in behavioural science. Research on psychological safety, pioneered by Amy Edmondson at Harvard, consistently shows that teams perform better when they feel safe to take risks and acknowledge failure. Treating targets as hypotheses creates exactly that environment. A missed target is not a personal failure. It is a signal that the model was incomplete, that an assumption was wrong, or that the external environment shifted in a way the plan did not anticipate. The appropriate response is curiosity, not punishment.

The metric tree supports this reframing structurally. When a top-level target is missed, the tree allows you to trace backwards through the branches to find exactly where the variance originated. Revenue fell short by £500K. The tree shows that customer acquisition was on track but ARPU declined because the plan-mix shifted toward lower tiers. That is a specific, actionable finding. It tells you that the pricing strategy or upsell motion needs attention, not that the team failed.

Practically, this means your target-setting process should include explicit documentation of the assumptions behind each target. What conversion rate are you assuming? What retention rate? What average deal size? When the quarter ends, review the assumptions as rigorously as you review the outcomes. Build the habit of asking "what did we learn?" before asking "what do we do next?" Over time, this shifts the culture from one that fears targets to one that uses them as a tool for continuous improvement. The targets become better each cycle because the model becomes more accurate, and the model becomes more accurate because people are honest about what it got wrong.