A recipient-choice program can look healthy in a launch report while leaving its most important question unanswered: of the people who were actually invited and had enough time to act, how many completed a choice? A raw count of selections is not a completion rate, and a percentage without a cohort, observation window, exclusions, and cost boundary cannot guide a useful decision. This guide gives gift-program owners a reproducible way to define that denominator, inspect the path from invitation to choice, and decide whether to change the invitation, available options, follow-up, or operational support. Its companion workbook calculates from your own records. It does not supply an industry benchmark or imply that a particular completion rate is universally good.
Define the decision and the cohort before calculating a rate
Write the decision in a sentence before opening a spreadsheet. “We need to decide whether to change the invitation for next quarter’s employee milestone program” is testable. “We want to improve engagement” is not. The program owner should name the population, the invitation route, the choice offered, the time allowed, and the next decision date. An analyst can then decide which events belong in the numerator and denominator. Operations must confirm what each system status means; a vendor’s “sent” status might mean accepted for delivery, not actually delivered to a recipient.
A cohort is a group with a common entry rule and a comparable opportunity to complete. A monthly send batch may be one cohort, but a program spanning employee appreciation, customer onboarding, and partner recognition usually needs separate groups. Those audiences may differ in eligibility, permission to receive gifts, sender familiarity, catalog relevance, and contact routes. Combining them can obscure a local problem. Record the cohort assignment at invitation creation and do not change it after seeing the outcome. When a person receives several legitimate invitations for different occasions, retain distinct invitation IDs but state whether the analysis unit is an invitation or a unique recipient. Do not switch units between the numerator and denominator.
The minimum schema is intentionally small. A one-row-per-invitation input is easier to audit than a joined export with names, addresses, free-text preferences, and support notes. The Giftpack measurement framework provides a broader view of campaign KPIs; this workbook narrows the question to the choice stage and the period in which the recipient could act. If your source contains repeated status updates, derive the first qualifying event per invitation ID in a controlled preparation step. Keep the raw log elsewhere with access appropriate to its sensitivity.
| Field | Definition for this workbook | Owner’s check |
|---|---|---|
| Cohort ID | Stable group assigned when the invitation is created | Does it represent one comparable offer and route? |
| Invitation ID | Unique key for one genuine invitation opportunity | Are corrections and resends linked to the original? |
| Sent and choice dates | Observed invitation and completed-choice events | Are times normalized to one reporting time zone? |
| Eligibility flag | Documented inclusion decision, entered as 1 or 0 | Was the rule fixed before outcomes were reviewed? |
| Costs | Selected invitation and fulfillment costs in one stated currency | Are missing, refunded, and zero amounts distinguished? |
The table is a contract for analysis, not a request to add all five fields to a gifting platform. A data steward can prepare a pseudonymous measurement extract. A privacy owner decides whether the purpose, retention period, access, and cross-border movement are appropriate. A program owner signs off on the cohort and exclusion definitions. A finance owner signs off on cost scope and currency conversion. Keep those approvals outside the public workbook if the underlying records are sensitive.
Build a denominator that has had a fair chance to complete
The simplest headline is completed eligible invitations divided by eligible invitations. It is also the easiest to distort. An invitation sent yesterday should not be counted as a failure in a seven-day completion report. Choose an observation date and a completion window, then include an invitation in the denominator only when its full window has elapsed. A choice within the window is a completion. A choice after the window remains a real recipient action but does not retroactively change the earlier-window metric. Report it separately or use a longer-window sensitivity view. This is right-censoring in practical terms: the record is not yet fully observable at the cutoff.
The workbook sets an editable observation date and a seven-day default window. Both are assumptions, not external benchmarks. It marks recent rows as censored, then shows eligible invitations, completions, and the rate for the mature set. It also calculates seven-, fourteen-, and thirty-day views using a separate mature denominator for each window. Do not compare those three percentages as though they all describe the same set of invitations; the longer window excludes more recent sends. A program can improve the apparent thirty-day rate while serving a different, older population. The correct comparison is between cohorts with matching rules, windows, and data cutoffs.
Eligibility must also be decided without hindsight. Legitimate exclusions might include test invitations, duplicate sends that reused one opportunity, cancelled programs before the recipient had a usable link, or people ruled ineligible under a documented policy. “Did not choose” is not an exclusion. Neither is “hard to reach” if the program intended to invite that population. Record exclusion reason codes in a controlled source and summarize them in aggregate; do not place sensitive explanations or personal details in the public template. Review whether an exclusion policy removes more of one region or recipient group than another.
Delivery is a separate event. If a link was not deliverable, decide whether the invitation was ever truly at risk of completion. Use a documented rule and keep both counts visible: intended eligible invitations and confirmed delivered invitations. If only confirmed deliveries form the denominator, a poor delivery operation can produce a deceptively high completion rate. The workbook retains both steps so the owner sees whether a change belongs in address quality, email deliverability, sender recognition, catalog choice, or follow-up. It should never silently present delivery-adjusted completion as the whole-program rate.
Prepare the event trail without carrying personal data into analysis
Export only the records needed to answer the declared question, from systems and routes that the organization is authorized to use. In a preparation copy, replace addresses and names with a stable invitation key. Normalize event dates to a single reporting zone and document that zone beside the cohort. Deduplicate repeated updates by invitation ID and event type before importing. A reminder may generate another email event without generating another opportunity; it should not add a new denominator row. A replacement gift may generate a new fulfillment event but still belong to the original choice opportunity.
Run the preparation in an order that prevents subtle leakage and double counting:
-
Identify the invitation population from the authorized program list, not from the subset of people who opened an email.
-
Remove known test and void records under predeclared rules; retain counts and reasons in a restricted audit log.
-
Resolve duplicate IDs and correction chains. If two real invitations share an ID, stop and fix the source key rather than letting a formula guess which one wins.
-
Normalize dates and currency units. Check that a delivered, reminder, or choice date cannot precede the initial invitation without an explained migration case.
-
Map each invitation to one cohort and an eligibility flag. Resolve unknown eligibility with an owner; do not coerce it to zero.
-
Import only the necessary measurement fields. Keep recipient contact details, address, free-text preference, and support narrative out of the workbook.
-
Reconcile row counts and total costs to the restricted source report, then retain the extract version and cutoff used for the decision.
The workbook’s row checks flag duplicate IDs, missing sent dates, impossible date order, negative costs, and invalid eligibility values. A zero fulfillment cost is allowed: it might represent a digital option, an in-kind item, or a cost booked elsewhere. Zero is not the same as missing; the finance owner must document its meaning. A row marked for checking is not silently included in clean-rate calculations. The QA sheet is an exception list, not a permission to delete inconvenient outcomes. A team with more than the template’s prepared rows should extend formulas and tests before adding data, or rebuild a governed extract with the same definitions.
The recipient preference-center guide explains why choice, address collection, and promotional permission should remain distinct. The benchmark can measure the choice event without asking recipients to surrender unrelated information. If the analysis is broken down by small teams or countries, suppress or aggregate groups that might reveal an individual’s decision. A rate can be mathematically correct and still be inappropriate to publish or share with a manager.
Read the funnel as a set of possible failure points
The workbook separates sent invitations, confirmed delivery, recorded choice, and completion within the chosen window. These are not interchangeable statuses. A sent-to-delivered gap points toward routing, bounce handling, address quality, or source-system semantics. A delivered-to-choice gap may involve recognition of the sender, relevance of the options, accessibility, unclear deadlines, or a policy concern. An unusually long time-to-choice may be harmless for a flexible occasion and problematic for an expiring seasonal program. Each hypothesis needs evidence beyond the metric itself.
For every cohort, review counts before rates. Ten completions out of twelve mature invitations and ten out of a thousand invite very different decisions, even if a narrow segment’s rate looks better. Check the underlying program size and the number censored. Keep a record of when the extract was taken; late-arriving events can revise a historical cohort. Publish or circulate a refreshed rate only with its new cutoff. Do not copy a percentage from a prior slide and silently pair it with a new cost total.
The cohort summary gives a place to enter up to six stable cohort IDs and compare eligible and completed counts. If one group has no mature invitations, the rate displays “n.a.” rather than zero. That distinction matters: zero completions among a real mature cohort is a result, while no observable denominator is a missing comparison. The sensitivity sheet changes the window without rewriting source rows. The time-to-choice view estimates elapsed days among valid completed rows. It does not claim that an average or median describes the uncompleted invitations; those are censored or unresolved observations.
The follow-up view is descriptive. People who receive a reminder are often selected because they have not yet completed, so a comparison between reminded and unreminded groups is confounded by design. The workbook counts reminder exposure and a choice observed after a reminder, but it does not label that difference a causal lift. To test whether a reminder helps, plan an authorized experiment or another comparison with a credible assignment rule, record timing and exclusions before outcomes are seen, and have a privacy and communications owner review the contact cadence. Do not infer consent to future marketing from an operational reminder.
Treat cost per completion as an accounting question first
Cost per completed choice is a useful operating measure only when its numerator and denominator describe the same population and period. Decide whether cost includes invitation production, platform fees, gift value, delivery, support, returns, and staff time. Not every organization can allocate every shared cost to a single invitation. State the boundary instead of filling a blank with a guessed figure. If a program pays for unredeemed gifts differently from redeemed gifts, finance must decide whether a committed amount, actual charge, refund, or accrual belongs in the report. This workbook’s input columns are intentionally limited; expand them with an approved ledger mapping when the decision requires richer accounting.
The cost sheet sums the entered invitation and fulfillment costs for valid, eligible, mature rows and divides by in-window completions. When no completion exists, it shows “n.a.” rather than an infinite or falsely zero unit cost. A zero entered cost remains valid, but a negative value is flagged for review rather than treated as a quiet credit. Credits and refunds should have a separate, documented sign and reconciliation rule in the source. If cohorts use different currencies, convert them under a recorded rate date and policy before combining; otherwise compare within one currency. A blended number without a common currency is not a benchmark.
Cost and completion can move in opposite directions. A cheaper invitation may produce fewer genuine choices. A more expensive option mix may improve completion but violate the budget or send the wrong signal to a recipient. Evaluate the decision against the program’s purpose and constraints, not against a universal target. The recipient experience guide maps choice to address, delivery, and support events; measure those downstream obligations separately. A selection that cannot be fulfilled is not a successful recipient experience simply because the choice cell is populated.
Worked decision 1: a hypothetical employee milestone cohort
Consider a hypothetical organization inviting employees in two regions to choose a service-anniversary gift. Its owner wants to know whether a confusing invitation caused a weak response in one region. The inputs are two month-based cohorts, stable invitation IDs, sent and delivered dates, choice dates, reminder dates, an eligibility rule approved by People Operations, and a fixed seven-day window. The communications owner supplies the actual invitation text and language. Operations supplies the status definitions. Privacy reviews the extract. Finance supplies the bounded cost fields. No individual address or gift preference is exported to the workbook.
The initial aggregate may show a lower completion rate for one region, but the analyst first checks delivery. If many invitations were not confirmed delivered, rewriting the choice page is premature. The team compares cohorts using the same cutoff and looks for source-system changes, holiday timing, and small sample sizes. It inspects the count censored because the second region’s batch went out later. It then reviews the invitation copy with local colleagues: was the sender recognizable, was the deadline stated correctly, and were the available options genuinely usable? These are questions for a controlled content review, not conclusions that the spreadsheet can prove.
The decision has at least three paths. First, if delivery is the main issue, repair the sending route and verify a fresh delivery event before changing gift choices. Second, if delivery is sound but the copy is unclear, test a revised invitation in a new cohort with equivalent eligibility and timing. Third, if people hesitate because the options are not suitable, adjust the catalog or provide a decline route before increasing reminders. A single reminder may be appropriate under the organization’s communications policy, but it is not a substitute for accessible, culturally relevant choice. The tradeoff is that changing several elements at once may improve the experience yet make it impossible to attribute which change mattered.
Recovery is explicit. If the analyst finds duplicate IDs, the cohort result is paused and the source owner resolves the mapping. If a translated message misstated the deadline, the communications owner corrects it, records the affected invitation set, and begins a new measurement period rather than reclassifying old outcomes. Acceptance evidence is a reconciled row count, a documented window and cutoff, a reviewed invitation, a small-group privacy check, and a decision memo naming the owner of the next test. This hypothetical case provides a way to reason, not a customer result or industry statistic.
Worked decision 2: a hypothetical customer thank-you program
Imagine a hypothetical customer-success team offering a choice of physical gift, digital reward, or donation after a completed onboarding milestone. It wants to know whether a planned follow-up is worth sending. The program owner defines the occasion and eligibility. Customer success confirms which customer contacts may be invited. Legal and procurement review gift and conflict policies. Operations validates the invitation and choice events. Finance chooses whether to include committed gift value, redeemed value, delivery, and returns in cost. The team avoids a universal “best” completion target because the invitation is voluntary and the relationship context differs from an employee program.
The analyst separates invitations from reminders. A follow-up sent to nonresponders cannot be judged by comparing their crude completion rate with people who never needed a reminder. Instead, the team first verifies that the email was deliverable, the invitation link remained usable, and the follow-up did not contradict the recipient’s decline or contact preference. The source extract records a reminder timestamp without copying the message body or recipient identity. The analyst then examines choices completed after the reminder as an operational count, with an explicit note that it is not causal evidence.
One path is to keep the existing reminder if the content is clear, the cadence is approved, and support cases show recipients needed a simple nudge. Another is to change the invitation itself if recipients cannot identify the sender or understand what happens after choosing. A third is to skip reminders and offer a clearer opt-out if the audience has a higher risk of feeling pressured. Each path trades reach, workload, recipient autonomy, and measurement clarity differently. For a defensible causal estimate, the team would need a properly approved comparison design and sufficient observations; it should not advertise “reminder lift” from this descriptive workbook alone.
Failure recovery matters here because a gift selection is upstream of fulfillment. If donation selection records exist but the fulfillment status is missing, operations reconciles the downstream event before reporting program success. If a physical-gift route collected an address without the appropriate notice, privacy owners handle the incident and suspend the affected workflow. The benchmark is not a privacy assessment or a compliance approval. Acceptance evidence includes an authorized contact list, delivery and choice event definitions, a reconciled source extract, reviewed reminder language, a cost boundary, and a documented send/defer/stop decision. Again, every number in this case would come from the organization’s own records; the scenario itself supplies no invented outcomes.
Use the workbook without turning a template into a benchmark claim
Download the English XLSX workbook and matching CSV import schema. Both files are in English; choose the corresponding localized file if your team works in Traditional Chinese, Japanese, or Korean. The CSV is a header-only preparation template, not a data export. Enter one invitation per row in the workbook or prepare a governed import with the same fields. Record the cohort ID, unique invitation ID, dates, eligibility flag, choice state, and bounded costs. On Sensitivity, set the observation date and completion window for the decision. Inspect QA before reading Funnel, then enter cohort IDs in the summary. Keep a dated copy of the source query and any exclusion decisions in a restricted workspace.
Do not publish the input workbook if it contains recipient-level rows. Share a sufficiently aggregated result with a date, geography, cohort rules, numerator, denominator, observation window, cost boundary, and known gaps. A chart without those details is not a benchmark. If another team cites the method, it should cite this version and describe its own data period and exclusions. The workbook is version 1.0, dated October 18, 2026; its public method can be reviewed quarterly for schema changes and annually for a fuller methodology update. A changed invitation state, new fulfillment route, or new privacy requirement warrants an earlier review.
The model is intentionally conservative. It does not infer statistical significance, causal reminder effect, market average, legal permission, or platform performance from empty cells. It does not calculate a unique-person rate across campaigns or automatically exchange currencies. It accepts zero as a genuine cost value when the owner confirms its meaning. It flags invalid rows rather than burying them in a denominator. Where the source cannot distinguish sent, delivered, and opened, mark the missing event and avoid labeling a proxy as a verified stage. An honest “unknown” often protects a better decision than a precise-looking percentage.
Conclusion: make the next program decision traceable
A useful recipient-choice benchmark is a repeatable local measurement contract: one real invitation opportunity per key, stable cohorts, an explicit eligibility rule, a mature observation window, a clear completion event, and costs on the same basis. Read counts before rates and examine delivery, choice, follow-up, and fulfillment as different events. Use sensitivity windows to learn how timing changes the answer, not to select the most flattering percentage. Two hypothetical programs can use the same workbook yet make different decisions because their buyers, policy constraints, recipient relationships, and failure modes differ. Save the rules and the cutoff beside every result so a later reviewer can reproduce the decision.
For teams that have already made the necessary program, privacy, policy, and finance decisions, Giftpack may help execute a recipient-choice gifting flow and track its operational stages. Confirm the current product fit for the specific route; Giftpack does not replace the organization’s legal, tax, payroll, privacy, employer, or statistical judgment.