Quick definition: Completion rate is the percentage of a defined population that finishes a specified process, task, form, lesson, workflow, or sequence within an agreed time window.
What is completion rate?
Completion rate measures progress from a genuine start or eligible opportunity to a defined end state. It is used for checkout flows, registration forms, onboarding, surveys, training, video lessons, support workflows, applications, and product tasks. The numerator must represent actual completion, not a click on “finish” or a screen that appears after an error.
The metric is not universally a conversion rate. Completion rate focuses on finishing a bounded process; conversion rate refers more broadly to a desired action among an eligible population. A checkout completion rate can be a funnel stage conversion from checkout start to paid order. A course completion rate can be an educational outcome without being a commercial conversion. A completion event can also be shallow if the process is designed poorly, so teams should establish that completion represents a useful customer outcome.
Completion rate differs from success rate when a process permits several outcomes. A support workflow might be completed when a ticket is closed, but success requires the customer’s issue to be resolved without reopening. A form can be completed but rejected by eligibility checks. Name the metric according to the outcome it truly describes.
Completion-rate formula and denominators
The general formula is:
completion rate = unique eligible units that complete process / unique units that start or are eligible for process × 100
The denominator should match the question. “Form completion among starters” diagnoses form friction. “Completion among all assigned eligible users” measures the total effect of offering a redesigned process. A workflow that is optional needs an eligibility rule; a mandatory workflow may use all units reaching a preceding required state. State whether an abandoned but resumed task can complete, whether repeated attempts count once, and how long a unit has to finish.
| Metric | Numerator | Denominator | Use |
|---|---|---|---|
| Form completion | Submitted valid forms | Form starters | Form usability diagnosis |
| Task success | Tasks completed with success state | Eligible task attempts | Product workflow quality |
| Course completion | Learners finishing required content | Enrolled eligible learners | Learning program progress |
| Experiment completion | Assigned users completing outcome | All assigned eligible users | Causal product decision |
Use a consistent unit. If a person can submit several forms, a per-attempt completion rate differs from the share of unique people who complete at least one. Both may be useful, but they should not share a label. For account workflows, one user’s completion can make the account complete; choose account-level aggregation if that is the product outcome.
Completion rate in A/B testing
Completion rate is often a suitable primary metric for an experiment that changes the process itself: reducing form fields, improving error messages, reordering setup, adding saving and resuming, or clarifying instructions. Select the metric at the level of the intended decision. If a change affects who begins a process, completion among starters is post-treatment conditional and should be a diagnostic; completion among assigned eligible users is usually the primary causal measure.
Pair completion with outcome quality. A shortened loan application can increase submitted forms while decreasing approval quality; an onboarding flow can increase “done” states while users fail at the first real task; a survey can be completed faster with lower-quality responses. Relevant guardrails include validation errors, time to completion, rework, support contacts, downstream conversion, refunds, and retention. This guide to primary and guardrail metrics explains the discipline.
Predefine the observation window and statistical plan. When a process takes several days, do not compare recent entrants before both arms have equivalent follow-up. Sample-size calculation should use the baseline completion rate and the smallest difference that would justify implementation; our sample-size article covers the inputs. Report counts, rates, percentage-point differences, and confidence intervals rather than a winner label alone.
Worked completion-rate calculation
A software company tests inline validation in an account-verification flow. The primary outcome is successful verified completion within 24 hours among all new accounts assigned to the flow. Control has 6,000 assigned eligible accounts, with 3,900 verified completions. Treatment has 6,050 accounts, with 4,174 completions.
control completion = 3,900 / 6,000 = 65.0%treatment completion = 4,174 / 6,050 = 69.0%absolute difference = +4.0 percentage pointsrelative lift = 6.2%
Among accounts that start entering details, completion rises only one point. That can coexist with a four-point all-assigned lift because treatment also encourages more eligible accounts to start. Both findings are useful, but the all-assigned result estimates the total product impact. The team checks validation errors, fraud review rate, account activation, time to completion, and support contacts before rollout.
Completion data-quality caveats
Completion frequently spans client and server systems. A browser confirmation might fire even when the server rejects the record; a backend status can change later after asynchronous validation. Prefer a durable success state from the system of record, or retain separate “submitted,” “accepted,” and “completed” events. Preserve attempt identifiers and timestamps to deduplicate retries and distinguish resumed sessions from separate attempts.
Timeouts and censoring affect the denominator. A user assigned near the end of a reporting period may not have had 24 hours to complete. Exclude or delay analysis using a pre-specified maturity rule applied equally to both variants. Timezone changes, daylight-saving transitions, and job retries can alter timestamp logic; test boundary conditions explicitly.
Accessibility and device differences may cause completion gaps that aggregate reporting hides. Check that assistive technologies, low-bandwidth connections, mobile browsers, and localized formats can complete the process. Segment investigation can find implementation defects, but predefine decision-critical segments and avoid treating every post-hoc difference as conclusive.
Common completion-rate mistakes
- Counting a button click as completion: use a confirmed final state.
- Using starters after a treatment changes starts: this hides total experience impact.
- Mixing attempts and unique users: repeated retries alter the rate.
- Ignoring incomplete follow-up: newer entrants may not have enough time.
- Optimizing speed alone: rapid completion can create mistakes or low-quality records.
- Changing validation rules during comparison: the definition becomes inconsistent.
Frequently asked questions
Is completion rate the same as conversion rate?
Not necessarily. Completion rate is finishing a defined process; conversion rate is a broader desired outcome. A process completion can be one funnel conversion stage.
Should the denominator be starters or eligible users?
Use starters to diagnose the process, and eligible assigned users to assess the total effect of offering a variant. Label the distinction.
What should count as completed?
Use the durable final state that the business and customer recognize, such as accepted submission, successful verification, or fulfilled task.
Can completion rate exceed 100%?
Not for a unique-unit rate with consistent definitions. A per-attempt or total-event measure can exceed 100% only if the denominator is different and should be named differently.
Why can completion rise while satisfaction falls?
The flow may pressure customers, hide consequences, or accept low-quality submissions. Measure errors, rework, support, and feedback.
Summary
Completion rate is the share of a defined population that reaches a confirmed end state for a bounded process. Define the unit, start or eligibility condition, final event, retries, and observation window. In A/B tests, use all assigned eligible units for the decision, investigate conditional process rates carefully, and protect quality outcomes.
Sources
- Nielsen Norman Group: Task completion
- UK Government Digital Service: Measure performance
- NIST: Binomial distribution