Fairness starts with the eligible population
The eligible population is the complete set of people or entries allowed to participate under rules fixed before selection. State the opening and closing time, geographic or membership limits, required information, and reasons an entry can be excluded. Apply the same tests to everyone. A random result cannot correct a list that omitted qualified people or included ineligible ones.
Input: 30 eligible people, each assigned one unique integer from 1 through 30. Rule: one entry per participant and one winner. Calculation: each valid number has probability 1/30 on the draw. Result: the generated number identifies one person. Interpretation: equal opportunity follows only if the mapping is complete, unique, frozen, and the number is drawn once under the announced procedure.
Decide how many entries each person receives
One entry per participant gives each eligible person equal probability. Weighted entries answer a different question: a participant with three valid entries has three times the chance of someone with one. Weighting can be legitimate when the announced rules intentionally grant multiple chances, but it must never be introduced silently. Record whether equality applies to people, entries, teams, or another unit.
Input: a list contains Alex twice because the same form was submitted twice, while 28 others appear once. Rule: one entry per participant. Process: resolve identity and remove the accidental duplicate before assigning numbers. Result: 29 unique eligible participants, not 30 entries. Interpretation: leaving both rows would weight Alex without authorization. If multiple entries were explicitly permitted, preserve and document them instead of deduplicating automatically.
Freeze the list and procedure before drawing
Close the intake, resolve eligibility questions, save the final list, and record its count before generating anything. Publish or internally approve the number of winners, whether alternates are selected, whether selection is with or without replacement, who operates the tool, and what timestamp or record will be kept. A hash or access-controlled snapshot can help show that the file was not changed, but it does not independently prove the underlying eligibility decisions were correct.
Run the announced draw once. Repeating until the output looks diverse, avoids a familiar name, or pleases observers converts random selection into discretionary selection. If a technical failure occurs, record what failed and follow a prewritten restart rule. Do not call a test click a real draw unless the procedure explicitly defines how tests are separated from production.
Choose the tool that matches the list
Use the Random Number Generator when entries have a complete, continuous numbering such as 1–30 and only one integer is needed at a time. It includes both endpoints. The current range implementation uses Web Crypto when available but has a small modulo-mapping bias and a Math.random fallback, so it should not be represented as a certified or audited draw system.
Use the Random Number Picker when the eligible identifiers themselves are numeric but nonconsecutive, or when several distinct winners must be returned without replacement. It removes duplicate numeric values. That behavior is useful only when duplicate values are accidental; it cannot represent intentionally weighted duplicate entries. Neither tool knows whether two different numbers belong to the same person, so identity-level duplicate controls remain part of list preparation.
Multiple winners and redraw rules
Input: 30 unique participants and three winners who must be different. Rule: select three identifiers without replacement. Process: use the picker with the frozen identifiers and request count 3, or remove each valid winner before the next range draw. Result: three distinct identifiers. Interpretation: every remaining participant's conditional chance changes after each removal, while no selected identifier can repeat.
Input: the selected participant cannot be contacted within the announced period. Rule fixed beforehand: mark the original selection invalid after 72 hours, retain its record, and draw one replacement from the remaining eligible nonwinners. Process: apply that rule exactly rather than choosing an alternate informally. Result: a traceable redraw. Interpretation: deciding the response window or replacement pool only after seeing the name creates avoidable discretion.
Keep a transparent record without exposing people
A useful record contains the rules version, eligibility cutoff, list count, numbering method, requested winner count, tool and settings, timestamp, operators or witnesses if applicable, returned identifiers, invalidation reasons, and any redraw. Keep enough evidence to reconstruct the procedure. Do not claim that a screenshot proves the random algorithm, but it can document what the interface displayed.
Protect privacy by publishing only what participants were told would be disclosed. Use internal identifiers rather than full names, emails, phone numbers, addresses, or account data in a public record. Limit access to the mapping between identifier and person, define retention, and delete data when it is no longer needed. Transparency means explaining the method and outcome, not exposing personal information.
Boundaries, review, and common failures
Common failures include changing eligibility after the draw, undisclosed duplicate entries, switching from equal to weighted chances, running multiple draws and choosing one, allowing the same person to win twice contrary to the rules, and inventing a redraw process after an invalid result. Another mistake is saying the tool made the process fair. Randomness addresses only the selection step; governance controls the surrounding process.
This guide covers informal classroom, community, team, and administrative selections. It is not legal advice and does not interpret promotion, sweepstakes, gambling, employment, procurement, research, or public-sector rules. A formal promotion or regulated selection needs professional review for the applicable jurisdiction and a system that meets its documentation, security, supervision, and audit requirements. NumUtility provides no certification or compliance guarantee.
A pre-draw and post-draw review
Before drawing, have a second person compare the eligibility rule with the frozen list, reconcile the number of accepted and rejected records, verify that each identifier maps to the intended number of entries, and test the selected tool with clearly marked dummy data. Delete the dummy values, reload the final frozen list, confirm the displayed count, and state aloud or in writing the winner count and replacement rule. This review reduces clerical mistakes without giving the reviewer discretion over the outcome.
After drawing, preserve the returned identifier before looking up personal details. Check eligibility using the same evidence and cutoff applied to everyone. If the winner is valid, record acceptance; if not, record the exact predeclared invalidation reason and apply the stated alternate or redraw process. Never edit the original result out of the record. A correction log should distinguish a list error discovered before selection from an invalid winner handled after selection.
For example, suppose 30 frozen identifiers yield 18, but record 18 had withdrawn before the published cutoff. If the prewritten rule excludes timely withdrawals and calls for a redraw from the remaining 29 entries, preserve 18 as the original invalid selection and draw once from those 29. If no such rule existed, pause for responsible review rather than improvising. Predictable exception handling is part of fairness because every participant should face the same rule.
When observers are involved, give them the rules and counts before revealing the selected identity. They can confirm that the operator used the frozen input and followed the declared sequence, but their presence does not turn the browser tool into a notarized or independently audited system. Describe the evidence precisely: an observer witnessed an informal procedure, not that NumUtility certified the outcome.