A strategy initiative can finish on time and still leave the question that justified its funding unanswered. A new service process may be in place, but customers may keep encountering the same problem. At the next review, leaders must decide whether to keep investing, change the approach or stop. Completion tells them what was delivered. It does not, by itself, tell them what changed.
The practical test is to connect the delivered work to the result it was meant to produce. Who was meant to encounter it? What would those customers or employees do differently? What evidence would change the next resource commitment? These questions make the output-to-outcome distinction useful for a decision, rather than just a choice of metrics.
Output, outcome and impact
An output is work the organization delivers: a launched service, a completed training program or a new process. An outcome is a change for customers or the organization, such as more customers completing a task or staff resolving an issue correctly on the first attempt. An outcome need not take years to appear; some changes can be observed soon after delivery. Impact describes a wider result the initiative is meant to support, such as improved retention or financial performance. That wider result may depend on several changes beyond the initiative itself.
The University of Wisconsin–Madison Division of Extension makes a similar distinction in its logic-model guidance. It separates activities and what they produce from changes in learning, behavior and longer-term conditions. For a strategy review, the distinction guards against treating the existence of an intervention as evidence that it worked.
These six hypothetical pairings show how a delivery measure can be joined to a change worth investigating. None of the outcomes follows automatically from its output.
| Area | Output delivered | Outcome to examine |
|---|---|---|
| Customer service | A revised help page | More customers resolve a specified problem without contacting support |
| Customer onboarding | A new account setup flow | More eligible customers complete their first useful task |
| Operations | A revised order handoff procedure | Fewer orders require correction after the handoff |
| Operations | Training for maintenance teams | More faults are diagnosed correctly on the first visit |
| Strategy | A new partner offering | More target customers buy and use the combined offering |
| Strategy | A redesigned approval process | Relevant decisions reach an accountable owner sooner, without more reversals |
Each pairing also exposes an assumption. A help page can reduce support contacts because customers solve their problem—or because they cannot find a way to ask for help. A faster approval process can improve decisions—or simply pass them through with less scrutiny. Leaders need a measure that reflects the intended change, alongside evidence that would reveal an unwanted one.
Follow the assumption to the next commitment
Consider a hypothetical retailer that funds a new online account setup flow. Leaders expect simpler setup to help new customers place a first order without assistance, increasing the chance that they return. The initiative has a chain of distinct claims: the retailer delivered the flow; eligible customers encountered it; more of them placed a first order; and, later, more returned. A successful launch confirms only the first claim.
The critical assumption is that difficulty with setup was preventing customers from reaching their first order. If customers instead leave because delivery slots are unavailable, improving setup could raise account completions without changing orders. If first orders rise but repeat orders do not, setup may have removed an initial obstacle without addressing why customers would use the service again. The decision depends on where the chain breaks.
Richard P. Rumelt’s argument in Good Strategy/Bad Strategy helps explain the role of the milestone. Coherent action responds to a diagnosed challenge. Launching the flow matters if setup was an important constraint and the new flow addresses it. Meeting the launch date cannot establish either proposition. The review should test the diagnosis, not only the project team’s delivery.
Before rollout, the retailer can assign an owner to each link and agree on the baseline and review window. This hypothetical plan connects each observation to a decision:
| Link in the chain | Baseline | Owner | Indicator and review date | Decision informed |
|---|---|---|---|---|
| Delivery and reach | Whether eligible customers could complete the previous setup journey, and how many reached it | Product lead | Check after launch whether the new flow works and reaches the intended customers | Fix access or delivery before judging customer response |
| First useful behavior | Share of eligible new customers who placed a first order without assistance before rollout | Customer lead | Review the same share after customers offered the new flow have had time to order | Investigate the barrier if behavior does not improve; continue a limited test if it does |
| Business result | Repeat-order rate for new customers before rollout, measured over a defined period | Commercial lead | Review repeat orders after new-customer groups have had the same opportunity to return | Consider wider investment, adaptation or stopping in light of the result and other explanations |
The baseline does more than provide a number to beat. It defines the customers and time period against which a later observation will be read. If the retailer counts all visitors before launch but only customers who complete setup afterward, the apparent improvement may reflect a changed denominator rather than changed behavior. The owner’s job is to make that distinction visible at the review, not merely report a favorable figure.
Suppose the flow works and reaches customers, but first orders do not improve. A broad rollout would be hard to justify on delivery alone. Leaders could examine where customers still abandon the journey and whether setup was the main barrier. They might revise the flow if it still creates friction, or redirect the next test toward delivery availability if that appears to be the constraint. An unchanged outcome calls for investigation, not an automatic stop.
Now suppose first orders rise. A limited continuation may be justified while the retailer checks repeat orders. That supports funding the next test, rather than every proposed extension of the project.
Ask what else could explain the change
Even a well-defined outcome measure cannot, on its own, show what caused the result. First orders could rise because of a promotion, seasonal demand or better delivery availability. Comparing the new rate with its baseline establishes that something changed; it does not isolate the effect of setup.
If feasible, the retailer could roll out the flow in stages and compare customers offered it with a suitably comparable group not yet offered it. That would strengthen the assessment only if differences between the groups and changes over time were examined carefully. Where a credible comparison is impractical, leaders can still check whether the expected sequence appeared and investigate competing explanations. As John Mayne’s contribution-analysis approach emphasizes, confirmed delivery alone makes a weaker case than evidence of the changes the initiative was expected to produce. Such a review can support a reasoned judgment about contribution, not proof that the initiative caused every observed gain.
The rigor required should fit the decision. A small, reversible test may warrant a cautious continuation on a promising early signal. A large, difficult-to-reverse commitment demands stronger evidence that the behavior has changed and that the initiative plausibly contributed. In either case, leaders should state what remains uncertain before deciding how much more to commit.
Keep the measures that explain where progress stopped
Retain delivery and reach measures after launch. If an outcome stalls, they help distinguish a solution that failed to reach customers from one that reached them but did not help. First-order completion then shows whether behavior changed while the retailer waits for repeat-order evidence. Each measure locates a different problem; none should become a substitute for the result the investment was meant to support.
At the next review, put the funding request beside the weakest supported link in that chain. If customers never reached the flow, fund the delivery fix; if they used it without placing more orders, revisit the diagnosis before expanding it. Where first orders improved, specify what the next period of funding must establish about repeat use. The review should end with a commitment proportionate to what has changed, and a clear question the next investment must answer.