B2B data sample evaluation checklist

A delivered sample gives you something concrete to review. Use this checklist to document which requested values are present, where review keys repeat, and whether each row fits your agreed scope. Keep the findings tied to the sample you received so another reviewer can reproduce them.

These checks describe the sample. They do not establish accuracy, email deliverability, permission to use the data, or quality across a larger delivery.

Start with a review record

Keep the original sample unchanged and work on a review copy under your organization's authorized handling rules. Record the sample ID and version, received date, review date, reviewer, evidence location and agreed brief version. Note what one row represents: an organization, a person or another defined unit.

List the requested fields and scope criteria before counting. Record how the sample was selected, if known. If selection is unknown, say so; you cannot assume the rows represent the full delivery. Define your acceptance criteria with the decision owner rather than choosing a threshold after seeing the results.

1. Count present values

For each requested field, divide rows with a present value by applicable rows. For overall completeness, divide present requested cells by all applicable requested cells.

Trim surrounding whitespace and count empty values as missing. Agree and record any additional missing-value tokens before counting. Exclude an inapplicable cell only with a recorded reason. If a denominator is zero, report not applicable rather than 100%.

A filled cell can still contain an incorrect value. Label this measure as presence, and keep factual verification separate.

2. Review repeated keys

Choose a key that suits the row unit and record the normalization rule. In the example below, normalization only trims whitespace and lowercases key text.

For each repeated-key group, count rows beyond the first: group size minus one. Add these excess rows and divide by all rows with usable keys. Report missing-key rows separately, alongside the number of distinct usable keys.

Also record every row in each flagged group. A group of two contributes one excess row but two rows to investigate. A repeated key is a review flag; it does not prove that two rows describe the same identity. Resolve the reason before considering any merge or deletion.

3. Check scope without hiding unknowns

Evaluate each row against every required scope criterion in the agreed brief:

• In scope: every required criterion has supporting evidence.
• Outside scope: at least one known value fails a required criterion.
• Unknown: no known criterion fails, but evidence for at least one required criterion is missing.

Report all three counts over the raw row count. Retain repeated rows for this check and state that choice. If a row has both a known failure and a missing value, classify it outside scope and preserve the missing-value note.

Copy this blank review worksheet

Use the register and tables below in your own authorized review document. Empty cells are for your findings.

Add a separate ledger row for each requested field or scope criterion where needed.

For status, use meets criterion, does not meet, unknown or not tested. Unknown means the evidence cannot resolve the check; not tested means you have not performed it. Use not applicable for a zero-denominator measure and record why. Leave the criterion and decision empty until the responsible buyer defines them.

Worked example: five fictional organization rows

Entirely synthetic records for demonstrating calculations. These are not AvalenData records, available fields or a benchmark. No personal or contact records are used.

For this fictional exercise, the buyer requests company label, country and sector. Scope requires country US and sector Manufacturing. The reviewer_key is a synthetic review helper, excluded from completeness calculations. “[blank]” marks an empty cell.

Presence: 14 of 15 requested cells

Five rows multiplied by three applicable requested fields gives 15 cells. Company label is present in 5/5 rows, country in 4/5, and sector in 5/5. Overall presence is (5 + 4 + 5) / 15 = 93.3%, rounded to one decimal. S4 supplies the one missing country value. No correctness test has been performed.

Repeated keys: one excess row among five usable keys

All five rows have usable keys, with four distinct keys and zero missing keys. The org-a group contains S1 and S2. Its size is two, so it contributes 2 − 1 = one excess repeated-key row. The excess-row rate is 1/5 = 20%. Two rows belong to the flagged group; that is a different count from one excess row. Ask whether the repeat is intentional before taking action.

Scope: three in, one outside, one unknown

S1, S2 and S5 meet both stated criteria: 3/5, or 60%, in scope. S3 has sector Software: 1/5, or 20%, outside scope. S4 has no country value: 1/5, or 20%, unknown. The counts include the repeated row and total all five raw rows.

These are observations, with no overall pass/fail verdict. Acceptance criteria have not been supplied. Accuracy, freshness, deliverability, permitted use and representativeness remain untested; they cannot be inferred from these counts.

Turn findings into specific questions

Attach row references to each question. For this example: What explains S4's missing country? Do S1 and S2 intentionally repeat? Why does S3's sector differ from the agreed scope?

Record the evidence needed, who will obtain it, and who owns the decision. Recheck affected findings against the same rules when a revised sample arrives, preserving both versions. A small or selected sample cannot establish full-population quality.

Explore data planning tools for related planning exercises. For the commercial context, discuss B2B lead requirements.

Discuss your sample review requirements

Bring your agreed brief, review definitions and unresolved questions to the conversation. Share sample material only through an authorized channel.

Discuss your sample review requirements