Automation ROI: Calculate the Real Return

Automation can save time without producing a financial return. It can increase output without improving the business. It can also appear inexpensive while creating hidden review, maintenance, and correction costs.

Calculating automation ROI requires more than comparing a software subscription with the number of hours a task currently takes.

A useful analysis must answer:

• What measurable result will improve?
• How much of the expected benefit will actually be realized?
• What does the complete system cost?
• How frequently will it fail?
• What is the financial consequence of those failures?
• How long will the automation remain useful?
• What will the solopreneur do with the time it returns?

The result should distinguish cash savings, economic value, capacity, risk reduction, and lifestyle benefits rather than combining them into one inflated number.

What Is Automation ROI?

Automation return on investment measures the value produced by an automated process relative to its total cost.

The basic formula is:

Automation ROI (%) = (Total realized benefits − Total costs) ÷ Total costs × 100

If an automation produces €3,000 of verified benefit and costs €2,000 during the measurement period:

ROI = (€3,000 − €2,000) ÷ €2,000 × 100 = 50%

The automation returned its cost plus an additional value equal to 50% of that cost.

The formula is simple. Defining “realized benefits” and “total costs” accurately is the difficult part.

Choose the Measurement Period

Automation ROI changes depending on the period being measured.

First-year ROI

First-year ROI includes setup, implementation, migration, and initial testing costs.

It is useful for deciding whether a new automation will repay its investment within the first 12 months.

Steady-state ROI

Steady-state ROI measures the expected return after the initial implementation period.

It excludes completed one-time work but includes recurring software, monitoring, maintenance, review, and correction costs.

Multi-year ROI

Multi-year ROI is appropriate when setup costs are substantial and the system should operate for several years.

Do not assume an indefinite useful life. Integrations, business processes, prices, policies, and AI models change.

For a small no-code workflow, a 12-month horizon may be sufficient. A custom system may require a longer analysis with future costs and benefits discounted to their present value.

Always label the period. “The automation has a 200% ROI” is incomplete without stating whether that refers to one month, one year, or its expected lifetime.

Establish the Baseline

ROI must compare the automated process with a credible alternative.

Measure the existing process before changing it.

Useful baseline metrics include:

• Number of tasks completed
• Active minutes per task
• Total completion time
• Waiting time
• Error rate
• Rework time
• Missed-task rate
• Exception rate
• Customer response time
• External contractor cost
• Revenue or gross profit connected to the process
• Number of interruptions
• Cost of historical failures

Use a representative period. A quiet week, launch week, holiday period, or seasonal peak may distort the result.

If volume changes substantially during the year, calculate separate normal, low, and peak periods.

Use median time when cases vary

A small number of unusually difficult cases can distort the average.

Record both:

• Median completion time
• Average completion time
• Time required by exceptional cases

The median describes the typical case. The average captures the total burden, including exceptions.

Separate active time from elapsed time

A process may take two days to finish but require only 20 minutes of active work.

Automation might reduce the two-day delay without saving many labor hours. That can still be valuable when faster completion improves customer experience, cash collection, or the speed of a later process.

Track active time and elapsed time separately.

Define the Business Outcome

Do not measure automation by the number of actions completed.

Define the business result it should improve.

Examples include:

• Reduce monthly administrative time
• Increase the number of clients served without longer working hours
• Shorten lead response time
• Reduce missed renewals
• Lower data-entry errors
• Deliver purchased products immediately
• Prevent website downtime from going unnoticed
• Reduce dependence on one person’s memory
• Improve the consistency of recurring reports
• Recover personal time without reducing output

A workflow that performs 5,000 automated actions has no meaningful ROI if those actions do not improve cost, revenue, capacity, quality, risk, or the owner’s life.

Types of Automation Benefits

Different benefits require different valuation methods.

Cash Savings

Cash savings occur when automation directly reduces an expense.

Examples include:

• Lower contractor fees
• Fewer payment penalties
• Reduced software duplication
• Lower transaction costs
• Avoided overtime
• Fewer refunds caused by operational errors
• Reduced printing, postage, or storage costs

Cash savings are the easiest benefits to defend because they can be verified through actual transactions.

Use:

Annual cash saving = Previous annual expense − New annual expense

Do not count a contractor cost as saved if the contractor continues to perform the same number of paid hours elsewhere in the process.

Owner Time Recovered

Automation may return time to the solopreneur without reducing a cash expense.

Use:

Annual hours recovered = Annual task volume × Net minutes saved per task ÷ 60

The word net matters. Subtract:

• Preparation time
• Review time
• Exception handling
• Corrections
• Monitoring
• Maintenance

If a task previously required 12 minutes and now requires four minutes of review, the net saving is eight minutes—not 12.

Research provides useful benchmarks, but it cannot replace measurement of the specific workflow. A Federal Reserve analysis found that surveyed generative AI users reported average savings equal to 5.4% of their working hours, approximately 2.2 hours in a 40-hour week. These were self-reported results across varied occupations, not a guaranteed saving for an individual business.

Valuing Solopreneur Time

A solopreneur’s hour does not have one universally correct financial value.

Possible methods include:

Avoided replacement cost

Use the amount that would otherwise be paid to a competent contractor or employee for the task.

This is appropriate when automation directly replaces outsourced work.

Contribution value

Use the profit the solopreneur can realistically generate by redirecting the recovered hour to paid or revenue-producing work.

Do not use a theoretical maximum rate if there is insufficient demand to sell the additional capacity.

Opportunity-cost value

Use the value of the best realistic alternative use of the hour.

That alternative may be:

• Client delivery
• Product development
• Sales
• Content with measurable expected value
• Strategic work
• Recovery and rest

The last category has real personal value but may not create a financial return.

Use a utilization factor

Not every recovered hour becomes productive capacity.

Use:

Economic time benefit = Hours recovered × Hourly value × Utilization factor

If automation returns ten hours per month but only six are realistically redirected to valuable work, the utilization factor is 60%.

This prevents theoretical capacity from being counted as realized economic value.

Capacity Gains

Automation can allow a solopreneur to handle more work without increasing working hours.

Use:

Additional capacity = Previous capacity ÷ Remaining manual share − Previous capacity

Suppose a standardized delivery process required two hours per client and automation reduces the owner’s involvement to 90 minutes.

The owner’s time requirement falls by 25%. If demand exists and every other constraint remains unchanged, capacity could increase from four clients in eight hours to approximately five clients.

Capacity is not revenue.

Additional capacity creates financial value only when:

• Sufficient demand exists
• The owner accepts the additional work
• Another bottleneck does not prevent delivery
• Quality remains acceptable
• The additional work produces positive contribution margin

Report unused capacity separately from realized sales.

Revenue Gains

Automation may contribute to revenue by:

• Responding to leads faster
• Reducing abandoned purchases
• Following up consistently
• Preventing failed renewals
• Improving product availability
• Increasing delivery capacity
• Personalizing relevant offers
• Recovering otherwise lost transactions

Do not count the full additional revenue as benefit.

Use contribution profit:

Revenue benefit = Incremental revenue × Contribution margin

If automation contributes €5,000 in additional sales with a 60% contribution margin, the economic benefit is €3,000 before automation costs.

Establish attribution

Revenue can change because of seasonality, pricing, traffic, promotions, market conditions, or unrelated business improvements.

Use one or more of these methods:

• Compare similar periods
• Introduce the automation to only part of the process
• Compare eligible and ineligible cases
• Track a defined conversion event
• Use unique campaign or workflow identifiers
• Measure the same customers before and after
• Delay other material changes during the test

Avoid attributing every improvement after implementation to the automation.

Quality Improvements

Automation can reduce variation and prevent avoidable mistakes.

Possible quality benefits include:

• Fewer missing fields
• More consistent formatting
• Correct calculations
• Fewer duplicate records
• More complete documentation
• Faster detection of anomalies
• More consistent customer communication
• Better compliance with an approved checklist

Value quality through the cost it avoids.

Use:

Quality benefit = Reduction in errors × Average cost per error

The cost per error may include:

• Correction time
• Refunds
• Customer support
• Replacement work
• Lost sales
• Payment charges
• Reputational recovery
• Professional fees

Do not assign an arbitrary monetary value to “better quality” without explaining how that quality affects the business.

Risk Reduction

Some automations protect against low-frequency but expensive events.

Examples include:

• Backups
• Security monitoring
• Expiration alerts
• Payment reconciliation
• Duplicate-payment checks
• Access revocation
• Required-document validation
• Website availability monitoring

Use:

Expected annual loss = Probability of failure × Financial consequence

The estimated benefit is the reduction in expected annual loss after the control is introduced.

However, expected value should not be used to justify unacceptable catastrophic exposure.

An automation with a 0.1% probability of causing an unrecoverable business loss may remain inappropriate even when its calculated average cost appears low. Use technical restrictions, approvals, insurance, or a manual process for consequences outside the business’s risk tolerance.

Faster Cycle Times

Automation can create value by shortening the time between:

• Enquiry and response
• Invoice and payment
• Purchase and delivery
• Data availability and decision
• Error occurrence and detection
• Content approval and publication
• Customer request and resolution

Faster cycle time may improve revenue, customer satisfaction, or working capital.

Measure the economic effect rather than assigning money to speed automatically.

For example:

• Did faster invoicing reduce average payment time?
• Did faster lead response improve qualified conversions?
• Did immediate delivery reduce support requests?
• Did earlier alerts reduce downtime?

If the faster process changes no downstream outcome, report the time improvement without converting it into financial ROI.

Lifestyle Return

For a solopreneur, an automation may be worthwhile because it protects evenings, reduces interruptions, or makes the business easier to leave unattended.

Possible lifestyle outcomes include:

• Fewer weekend checks
• Shorter administrative days
• Less context switching
• More predictable working hours
• Reduced anxiety about forgotten tasks
• Easier travel
• Better recovery
• Less dependence on the owner being constantly available

These benefits should not be forced into a speculative monetary value.

Track them separately:

• Personal hours reclaimed
• Out-of-hours interventions
• Number of recurring interruptions
• Days the business operates without manual attention
• Subjective workload or stress score

A workflow can have modest financial ROI and strong lifestyle ROI. That may be entirely appropriate for a business designed around independence rather than maximum scale.

Calculate Total Automation Cost

The denominator of the ROI calculation must include more than the visible subscription.

Discovery and Process Design

Before building, the owner may need to:

• Observe the existing process
• Clean data
• Document rules
• Standardize inputs
• Define exceptions
• Establish permissions
• Identify the authoritative source
• Create completion criteria

Value this time even when the owner performs it personally.

Implementation

Implementation costs may include:

• Workflow configuration
• Custom development
• Integration work
• Data migration
• Authentication setup
• Prompt and instruction development
• Template creation
• Testing
• Documentation
• Professional support

Include unsuccessful experiments that were necessary to reach the working implementation.

Software and Infrastructure

Recurring costs may include:

• Automation platform subscription
• AI model usage
• API charges
• Database hosting
• Storage
• Monitoring
• Logging
• Security tools
• Additional application tiers
• Transaction fees
• Backup services

Use expected production volume rather than the low volume observed during testing.

Review and Approval

Human oversight is a real operating cost.

Use:

Annual review cost = Annual cases × Review minutes per case ÷ 60 × Hourly value

If the system produces 1,000 outputs requiring three minutes of review, it creates 50 hours of annual review work.

This cost may still be lower than performing the complete task, but it must be included.

Exception Handling

Estimate:

• Percentage of cases requiring intervention
• Average time per exception
• Cost of recovering from failed actions
• Cost of unresolved cases
• Number of false escalations

Use:

Annual exception cost = Annual cases × Exception rate × Average exception cost

Exception rates often determine whether an automation succeeds economically.

Maintenance

Automations require maintenance when:

• An API changes
• A field is renamed
• An authentication token expires
• A website changes
• A business rule changes
• A model is updated
• An integration fails
• Volume increases
• A new exception appears
• The source data changes

Record actual maintenance time rather than assuming the workflow will operate indefinitely without attention.

Downtime and Recovery

Include:

• Lost transactions
• Delayed delivery
• Manual fallback work
• Investigation time
• Restoration time
• Customer communication
• Duplicate or missed actions

A workflow that saves ten hours during normal operation but requires a full day to repair every month may have negative ROI.

Security, Privacy, and Compliance

Additional costs may include:

• Access controls
• Data-processing agreements
• Privacy review
• Professional advice
• Audit logs
• Security testing
• Data minimization
• Incident response
• Compliance documentation

These are not optional overhead when the workflow handles personal, confidential, financial, health, or regulated information.

Switching and Exit Costs

Consider what happens when:

• The provider raises prices
• A required feature moves to another plan
• Data cannot be exported cleanly
• The service closes
• The workflow must be rebuilt elsewhere
• A model no longer supports the required behavior

A low initial price can hide expensive dependence.

AI-Specific ROI Costs

AI automation creates additional uncertainty because its outputs may vary.

Include:

• Model usage
• Prompt and context preparation
• Retrieval infrastructure
• Evaluation datasets
• Human verification
• Incorrect classifications
• Unsupported generated claims
• Model changes
• Performance monitoring
• Additional security controls
• Higher exception handling

Model inference may be one of the smallest cost components. According to Stanford cost data, the cost of querying a model performing at approximately GPT-3.5 level fell more than 280-fold between November 2022 and October 2024.

Lower model prices can improve unit economics, but they do not remove integration, review, correction, data, and governance costs.

A cheap incorrect output remains expensive when a person must investigate or repair its consequences.

Do Not Trust Perceived Productivity

Users may feel faster without completing work faster.

A 2025 METR experiment found that experienced open-source developers took 19% longer to complete selected tasks when allowed to use early-2025 AI tools. Before the experiment, they expected AI to make them 24% faster; afterward, they still believed it had made them approximately 20% faster.

The study examined a specific group, set of repositories, and generation of tools. It should not be generalized to all coding or automation work.

Its broader lesson is highly relevant: self-reported speed is not a sufficient ROI measure.

Record actual start times, completion times, accepted outputs, corrections, and failures.

Build the Full ROI Formula

For a first-year calculation, use:

Total realized benefit = Cash savings + Realized time value + Contribution profit + Error reduction + Risk reduction

Total first-year cost = Setup + Software + Usage + Review + Exceptions + Maintenance + Downtime + Security

Then:

First-year ROI (%) = (Total realized benefit − Total first-year cost) ÷ Total first-year cost × 100

Do not count the same benefit twice.

If recovered time creates additional client revenue, count either:

• The value of the recovered time, or
• The contribution profit produced with that time

Counting both would duplicate the return.

Risk-Adjusted ROI

Expected automation failures should be included in the calculation.

Use:

Expected failure cost = Annual runs × Material failure probability × Average financial loss

Then:

Risk-adjusted net benefit = Gross benefit − Operating cost − Expected failure cost

This approach works for frequent failures with measurable average consequences.

For low-frequency catastrophic events, use a separate risk limit. Do not rely exclusively on expected value.

Examples requiring separate controls include:

• Unrestricted payments
• Destructive database actions
• Disclosure of confidential information
• Legally binding communications
• Unsafe product guidance
• Publication across high-visibility channels

Break-Even Volume

Break-even volume identifies how many successful runs are required to recover the initial investment.

First calculate:

Net benefit per run = Manual cost per run − Automated variable cost per run − Expected exception cost per run

Then:

Break-even runs = Initial implementation cost ÷ Net benefit per run

Suppose implementation costs €600 and each successful run produces €5 of net benefit:

Break-even volume = €600 ÷ €5 = 120 runs

If the task occurs ten times per month, the expected payback period is approximately 12 months.

If the underlying process may disappear within six months, the project probably does not justify its setup cost.

Payback Period

Payback measures how long the automation takes to recover its initial investment.

Use:

Payback period = Initial investment ÷ Monthly net operating benefit

An automation costing €1,000 to implement and producing €250 of monthly net benefit has a four-month payback period.

Payback is useful for cash planning but does not measure benefits after the investment has been recovered.

Two automations can have the same payback period but very different useful lives, risks, and long-term returns.

Worked Automation ROI Example

Consider a hypothetical solopreneur who prepares four performance reports each month.

Existing process

• Four reports per month
• 75 minutes per report
• Five owner hours per month
• Owner time valued at €50 per hour

Monthly baseline cost:

5 hours × €50 = €250

Automated process

The workflow retrieves data and performs verified calculations. The owner reviews the analysis and recommendations.

• 25 minutes of owner time per report
• 100 minutes per month
• 1.67 owner hours per month

Net monthly time saving:

5 hours − 1.67 hours = 3.33 hours

Monthly time value:

3.33 × €50 = €166.50

Costs

• Initial setup: eight hours × €50 = €400
• Software and model usage: €35 per month
• Maintenance: 0.4 hours × €50 = €20 per month
• Expected correction and recovery cost: €10 per month

Monthly operating cost:

€35 + €20 + €10 = €65

Monthly net benefit:

€166.50 − €65 = €101.50

Payback period:

€400 ÷ €101.50 = approximately 3.9 months

First-year ROI

First-year benefit:

€166.50 × 12 = €1,998

First-year cost:

€400 + (€65 × 12) = €1,180

First-year net benefit:

€1,998 − €1,180 = €818

First-year ROI:

€818 ÷ €1,180 × 100 = approximately 69%

This is an economic ROI based on owner time valued at €50 per hour.

If the recovered time is neither sold nor redirected to economically valuable work, the automation does not create €1,998 in cash. It still returns approximately 40 personal hours during the year, which can be reported as a separate lifestyle or capacity benefit.

Use Sensitivity Analysis

ROI depends on assumptions that may be wrong.

Test at least three scenarios.

The conservative scenario reveals that the project becomes unattractive when less time is saved or owner time has a lower realizable value.

Test sensitivity to:

• Task volume
• Minutes saved
• Review time
• Exception rate
• Hourly value
• Software price
• Model usage
• Maintenance
• Useful life
• Revenue attribution
• Conversion rate
• Contribution margin

If a small change makes ROI negative, the investment is fragile.

Separate Cash, Economic, and Lifestyle ROI

A clear solopreneur dashboard should report three outcomes.

This prevents a time-saving workflow from being presented as cash profit while still recognizing why it may be valuable.

For example:

• Cash return: €300 per year
• Economic capacity returned: €1,500 per year
• Personal time reclaimed: 36 hours per year

The figures answer different questions and should not be added automatically.

Measure Output Acceptance

For AI-assisted workflows, generated output is not the relevant unit. Accepted output is.

Track:

First-pass acceptance rate

Outputs accepted without correction ÷ Total outputs

Correction rate

Outputs requiring material correction ÷ Total outputs

Abandonment rate

Outputs discarded ÷ Total outputs

Cost per accepted output

Total generation, review, correction, and system cost ÷ Accepted outputs

A system producing 1,000 drafts at low model cost may be less economical than one producing 300 drafts with a much higher first-pass acceptance rate.

Track Exception Economics

An automation may perform well on normal cases while losing money on exceptions.

Track:

• Standard cases
• Automatically resolved cases
• Human-reviewed cases
• Failed cases
• Reversed actions
• Unresolved cases

Calculate the contribution of each group.

Example:

The 5% failure group may consume more time than the 15% review group.

Averages can conceal this concentration.

Do Not Confuse Performance With Revenue

Business surveys frequently report stronger effects on performance than on revenue.

An OECD survey found that 65% of surveyed SMEs using generative AI said it improved employee performance. Smaller shares reported that it enabled scaling—35%—or increased revenue—26%.

These are self-reported associations, not proof that generative AI caused the outcomes.

They nevertheless illustrate why ROI measurement should follow the full chain:

Tool use → Task improvement → Capacity or quality improvement → Business outcome → Financial value

Many automation projects create the first or second result without reaching the final one.

Compare Automation With the Real Alternative

The alternative is not always “continue doing everything manually forever.”

Compare the proposed automation with:

• Eliminating the task
• Reducing its frequency
• Using a template
• Batching the work
• Improving the existing process
• Delegating to a contractor
• Buying specialized software
• Building a simple rule-based workflow
• Using AI assistance without automation
• Accepting the current cost

A €2,000 custom workflow may have positive ROI compared with manual work but poor ROI compared with a €20 monthly specialized application.

Choose the highest-value practical alternative, not merely the most technically impressive solution.

Common Automation ROI Errors

Counting all saved time as cash

Owner time creates cash value only when it reduces an expense or produces additional contribution profit.

Using revenue instead of profit

Additional sales still carry delivery, payment, support, refund, and product costs.

Ignoring review

A person checking every output is part of the operating system and its cost.

Ignoring maintenance

Integrations and business rules change. Historical maintenance should be included as soon as it becomes available.

Excluding setup time

The owner’s implementation time has an opportunity cost even when no invoice is paid.

Using vendor benchmarks as forecasts

Results from another company, occupation, or experiment do not establish the return for a different task.

Measuring generated volume

More messages, drafts, records, or reports do not prove economic value.

Double-counting benefits

Recovered time used to generate additional profit should not be counted again as both time value and profit.

Ignoring failed runs

The average successful task time says little when failures create expensive recovery work.

Assuming constant volume

Seasonality, business contraction, or process changes may reduce the number of future runs.

Ignoring opportunity cost

Time and money invested in one automation cannot be invested in another project simultaneously.

When Negative Financial ROI May Be Acceptable

Some automations are justified primarily by resilience, quality, or personal boundaries.

Examples include:

• Reliable backups
• Security alerts
• Expiration reminders
• Compliance records
• Emergency monitoring
• Access revocation
• Protection of uninterrupted personal time
• Reduction of stressful out-of-hours checks

Document the reason explicitly:

“This workflow is maintained for business continuity, although its direct financial ROI is negative.”

That is more useful than manipulating the value of time until the calculation becomes positive.

When to Stop an Automation

Define stop conditions before implementation.

Possible conditions include:

• ROI remains negative after the agreed test period
• Review time exceeds the expected limit
• Exception rate rises above a threshold
• Maintenance exceeds the monthly benefit
• A provider price increase breaks the economics
• The process changes materially
• Error consequences become unacceptable
• A simpler replacement becomes available
• The workflow no longer supports a useful output

Previous setup spending is a sunk cost. Do not maintain an unproductive automation merely because time or money has already been invested.

Review Automation ROI Regularly

Review a new workflow frequently during its first operating period.

During the first month

Track every failure, correction, exception, and manual intervention.

After three complete cycles

Recalculate time saved, operating cost, acceptance rate, and payback using actual data.

Each quarter

Check pricing, volume, maintenance, risk, and continued relevance.

After material changes

Recalculate when:

• The process changes
• The model changes
• An integration is replaced
• Prices increase
• Volume changes substantially
• A new market or product is added
• The automation receives more authority

ROI is not established permanently at launch.

Frequently Asked Questions

What is automation ROI?

Automation ROI is the percentage return produced by an automated process after subtracting its complete implementation and operating costs from its realized benefits.

How do you calculate automation ROI?

Subtract total automation costs from total realized benefits, divide the result by total costs, and multiply by 100. Use the same measurement period for benefits and costs.

What costs should automation ROI include?

Include process design, implementation, software, API usage, owner time, testing, review, exception handling, corrections, maintenance, monitoring, downtime, security, and switching costs.

How should a solopreneur value time saved?

Use avoided contractor cost, realistic contribution value, or the opportunity cost of the best alternative use. Apply a utilization factor when only part of the recovered time will be redeployed economically.

Is time saved the same as money saved?

No. Time becomes a cash saving only when it reduces an expense or produces additional contribution profit. Reclaimed personal time can still be a valuable lifestyle benefit.

What is a good automation ROI?

There is no universal percentage. The required return depends on setup cost, payback period, risk, useful life, alternative investments, and whether the benefit is financial, strategic, or personal.

How long should automation take to pay back?

Simple, low-risk workflows should generally justify short payback expectations because they may change or become obsolete. More durable systems can support longer periods when their benefits and useful life are well established.

How do you calculate automation payback?

Divide the initial investment by the monthly net operating benefit. The result is the number of months required to recover the setup cost.

How do you calculate break-even automation volume?

Divide implementation cost by the net benefit produced by each successful run. The result is the number of runs required to recover the investment.

How do you measure AI automation ROI?

Include model usage, context preparation, evaluation, human review, correction, false outputs, monitoring, and changing model performance. Calculate cost per accepted result rather than cost per generated result.

Can an automation have positive ROI without increasing revenue?

Yes. It may reduce expenses, recover valuable time, prevent errors, shorten cycle times, reduce risk, or protect personal time.

When should automation be discontinued?

Discontinue or redesign it when recurring costs exceed benefits, exceptions become excessive, the process changes, failures become unacceptable, or a simpler alternative produces a better return.

Measure What the Automation Actually Returns

Automation ROI is not the number of hours a tool claims to save.

It is the verified value remaining after implementation, review, exceptions, maintenance, failure risk, and opportunity cost are included.

For a solopreneur, that return may appear as cash, profit, capacity, resilience, or personal time. Each is valuable, but each should be measured honestly and reported separately.

The strongest automation is not the one that performs the most actions. It is the one that repeatedly returns more value than it consumes while keeping its errors, complexity, and consequences within acceptable limits.