Working harder, selling more, or changing a process does not necessarily mean a business has improved. First, specify which result matters, who benefits, and what you will not sacrifice to achieve it.
Improvement does not begin with a software tool or an impressive-looking metric. It begins with a verifiable business decision. A restaurant might cut waste by preparing fewer meals than its customers need. An app might gain registrations through promotions while losing active users. A repair shop might serve more customers while generating many more complaints. In each case, activity has increased, but progress remains uncertain. The American Society for Quality (ASQ) describes continuous improvement as the ongoing effort to improve products, services, or processes through incremental or breakthrough changes [R01]. UK government guidance on measuring service performance recommends starting with a service's purpose and the needs it serves rather than selecting metrics first [R02].
This article explains how to define an improvement worth pursuing, not how to build a dashboard or select every KPI formula. It applies to both solo businesses and larger teams; the amount of documentation changes, but the need for an observable outcome does not.
1. Distinguish an activity, a change, an outcome, an improvement, and growth
| Concept | What it describes | Lunch delivery example |
|---|---|---|
| Activity | Work performed or effort invested | Reviewing recipes and shifts in three meetings |
| Change | Something modified in the operation | Preparing smaller batches at different times |
| Outcome | What happens after the change | 31 of 520 prepared meals are discarded |
| Improvement | A favorable result against a baseline without unacceptable harm | A lower waste rate without more stockouts or lower contribution |
| Growth | An increase in scale, customers, capacity, or reach | Opening a second delivery window |
An activity may be necessary for improvement without producing one. A favorable result is also insufficient if it sacrifices a critical requirement. “Cut complaints by 30%” looks positive until we discover the complaint form stopped working. Growth may be sensible, but it does not replace judgment about quality, liquidity, or viability; deciding whether to expand is addressed later in this pillar.
Also distinguish a one-off correction—remaking a defective order—from a system improvement—changing the workflow so defects happen less often. Both matter, but they require different evidence.
2. Decide whom the improvement serves and why it matters
Start with one question: Which customer experience, economic result, or obligation do we want to protect or improve? Then separate demonstrated needs from the business owner's assumptions. Leadership can legitimately propose a priority; it should not present that priority as a proven customer need without evidence.
Use a short chain:
- Need or problem: “Waste consumes resources we could use to fulfill orders customers actually request.”
- Beneficiaries: the business (fewer losses), customers (availability maintained), and staff (reasonable workloads).
- Desired outcome: fewer prepared meals end up being discarded.
- Scope: lunch products at one location and defined service windows, not every business activity by default.
- Decision enabled: retain the new preparation pattern if the test succeeds; modify or reverse it otherwise.
Improvements need not maximize revenue. A service may prioritize safety, compliance, punctuality, or shorter waiting times. Sometimes maintaining good service while reducing cost or risk is the improvement. In other cases, cost-cutting is unacceptable because it undermines an essential condition.
3. Establish a baseline that does not mislead you
A baseline is observed performance before the change, with a defined population, time period, and source. It is not a remembered impression or an average chosen because it makes the proposal look attractive. To establish it:
- Specify the period (for example, four consecutive operating weeks), the unit (meals, not customers), and comparable operating days.
- Record sources: production logs, completed deliveries, unmet requests, ingredient costs, and incident records.
- State exclusions: refunds, orders canceled before preparation, test products, and closed days.
- Preserve numerators and denominators alongside percentages. Do not compare 12/100 with 120/1,000 while ignoring different populations or conditions.
- Record concurrent changes: menu, prices, weather, hours, or marketing campaigns. A baseline does not eliminate those factors; it makes them easier to identify.
GOV.UK guidance recommends using prior performance as a baseline, looking at trends, and segmenting results because averages can conceal differences among groups [R02]. If records are not yet adequate, begin measuring. A hypothetical number is useful for teaching a method, not for justifying a real investment.
4. Turn a vague wish into an observable outcome
“Improve customer service” does not say what changes for customers. “Sell more” does not say whether the goal is net sales, contribution, or collected cash. An actionable definition includes outcome, starting point, target, deadline, population, and constraint:
Over the next four comparable weeks, reduce the share of prepared lunches discarded from a 12% baseline to no more than 7%, while keeping unmet requests at 4.5% or lower and maintaining contribution per fulfilled order after waste costs.
These percentages are teaching assumptions, not recommended industry targets. The objective identifies what would be observed; it does not yet establish which intervention is best. Choosing a change requires diagnosis and testing, which other articles in this pillar cover.
Test the wording by asking someone who did not help write it: What is measured? Which orders are included? Over which dates? What data source is used? What counts as success? What result would force the team to stop?
5. Add guardrails so a “better” number cannot hide a worse business
The primary measure represents the intended result. A guardrail, or countermetric, checks what must not deteriorate. A measure can be secondary in a report yet critical from a risk perspective.
| Primary ambition | Risk of optimizing one number | Necessary protection |
|---|---|---|
| Less waste | Preparing too little and turning customers away | Unmet requests / observable demand |
| Faster deliveries | Poor packing or unsafe pressure on drivers | Incomplete orders, safety, and complaints |
| Higher sales | Discounts that eliminate contribution | Contribution per order, cash, and returns |
| Better repeat purchasing | Rewarding purchases that would have happened anyway | Incremental contribution and incentive cost |
| More website registrations | Acquiring users who never obtain value | Real activation and later use, not sign-ups alone |
A legal requirement or safety constraint should not be traded against extra points in a scoring model. It is a mandatory boundary, not a variable to exchange for profitability. If a business cannot verify a critical boundary, it cannot claim full success.
6. Recognize different types of improvement without confusing methods
Correction: restores an individual case to an agreed condition. Incremental improvement: modifies part of a process and tests whether the benefit persists. Variation reduction: prevents unusually bad days even if the average changes little. Breakthrough change: significantly alters capabilities or operations and may require additional investment, training, and controls.
Continuous improvement is not synonymous with total quality management (TQM) or with the plan-do-check-act (PDCA) cycle. ASQ distinguishes broader approaches such as Lean, Six Sigma, and TQM and explains PDCA as a practical cycle for change and review [R01][R03]. Here we define the direction of improvement; a separate Pillar 7 guide to PDCA will explain how to introduce, evaluate, and sustain changes. You do not need to formally adopt one of those frameworks to establish a baseline and make better decisions.
7. Handle competing objectives explicitly
Small businesses rarely have unlimited resources. When comparing alternatives, record what each option gains, what it risks, and who bears the cost. Speed may require overtime, extensive variety may increase inventory, and a loyalty program may reward people who were already frequent customers.
Do not settle conflicts through arbitrary averages. First exclude options that violate safety, privacy, commitments, or capacity constraints. Then compare the feasible ones against explicit criteria: customer value, contribution, cash flow, reversibility, effort, and quality of evidence. “We do not know yet” can also be a responsible decision when it identifies what observation would change the answer.
8. Worked example: waste, stockouts, and contribution at a coffee shop
Entirely hypothetical teaching case: “Almuerzo Cerca.” These figures describe no actual business and are not market forecasts. We compare two four-week periods using the same record definitions. Assume S/30 per delivered lunch, S/18 variable cost per sold lunch, and S/10 already spent on ingredients and preparation for each discarded lunch, without double counting. Simplified contribution excludes fixed expenses and taxes.
| Measure | Baseline: four weeks | Subsequent observed period: four weeks |
|---|---|---|
| Meals prepared | 500 | 520 |
| Meals discarded | 60 | 31 |
| Meals delivered | 440 | 489 |
| Requests unmet due to unavailable product | 20 | 21 |
| Waste / prepared meals | 60/500 = 12.00% | 31/520 = 5.96% |
| Unmet requests / (deliveries + unmet requests) | 20/460 = 4.35% | 21/510 = 4.12% |
| Contribution after waste costs | 440×(30−18)−60×10 = S/4,680 | 489×(30−18)−31×10 = S/5,558 |
| Contribution after waste / delivered order | 4,680/440 = S/10.64 | 5,558/489 = S/11.37 |
The later period would provisionally satisfy the waste goal of 7% or less, the unmet-request guardrail of 4.5% or less, and the financial protection of at least S/10.64 contribution per delivered order. However, volume, product mix, and customer behavior may also have changed. This comparison alone does not prove that the process change caused the results. Review seasonality, missing records, menu changes, the selection of operating days, and conditions of implementation.
Three counterexamples that are NOT acceptable improvements: (a) cutting waste to 3% by leaving 12% of requests unmet; (b) reaching 6% waste with negative contribution because of overtime; (c) claiming success by selecting only quiet days after the change and all days before it. Each hypothetical case violates the agreed comparison or guardrails.
9. Decide whether an intervention is worth doing now
An initiative deserves attention when the problem is material, a feasible action exists, and evidence supports acting at an acceptable level of risk. Ask: What happens if we do nothing? What will the test cost? What other work must wait? Can we stop if losses increase? Who will evaluate the result?
If a problem is small, current performance is adequate, and intervention costs more than its plausible benefit, maintaining and observing may be responsible. When evidence points to an immediate safety risk, preventive action can be urgent even without a long historical series. Later we will define reproducible KPIs and methods for studying causes; do not mistake this initial definition for an automatic approval of a change.
Reusable template: improvement definition sheet
Copy this table into a document or spreadsheet. The example column uses the hypothetical case above; replace it with actual records before making business decisions.
| Field | What to enter | Hypothetical example |
|---|---|---|
| Problem and beneficiaries | What affects whom | Lunch waste; business and customers |
| Business purpose | Why it matters | Protect availability and contribution |
| Population and period | Included units and timeframe | Lunches at one location over four weeks |
| Baseline and evidence | Value, numerator, denominator, source | 60/500 = 12%; production log |
| Time-bound target | Observable change | ≤7% in the next four comparable weeks |
| Guardrails | What must not deteriorate | Unmet requests ≤4.5%; contribution ≥S/10.64/order |
| Mandatory boundaries | Safety, legal obligations, promises | Do not sell unsafe products or hide requests |
| Owner and review | Who interprets results and when | Operations lead; weekly review |
| Stop condition | What forces a change or pause | Unmet requests exceed agreed limit |
| Available decision | Keep, revise, investigate, stop | Adopt only if evidence supports it |
Minimum approach without software: record daily meals prepared, delivered, discarded, and requested but unavailable; summarize the numerators and denominators weekly. If you need to combine cash, order, and app records without counting them twice, use the business data integration guide. The task here is first to identify what source is missing.
Common mistakes, limitations, and next decision
Do not label a hard-to-implement change as an improvement simply because it took effort. Do not compare percentages without denominators, copy targets from unrelated firms, interpret silence as satisfaction, offset safety violations with sales, exclude costs outside the chosen period, or infer causality from one before-and-after comparison. Regulated or safety-critical operations require appropriate controls and applicable checks before changes are tested.
What you should have by the end: a reviewable sheet that defines a valuable result, a baseline, a time-bound goal, guardrails, and an available decision. You can now decide whether an initiative deserves a trial; this sheet alone cannot establish that an intervention works.
References
- [R01] American Society for Quality (ASQ). Continuous Improvement. https://asq.org/quality-resources/continuous-improvement/
- [R02] UK Government Digital Service. How to set performance metrics for your service. https://www.gov.uk/service-manual/measuring-success/how-to-set-performance-metrics-for-your-service
- [R03] American Society for Quality (ASQ). PDCA Cycle. https://asq.org/quality-resources/pdca-cycle