Is there a fixed sample size required by the QI Agreement?

No. Reviewers expect a risk-based, explainable sampling plan covering documentation and payments that fits the bank’s scale and complexity.

How do we choose documentation strata?

Stratify by form type and risk. Larger slices for W-8IMY and entity treaty claims; smaller for simple W-9 or W-8BEN individual files.

How should payment samples be stratified?

Use income codes, rate buckets (30%, treaty, 0%) and payee types. Include targeted top-ups for high-risk scenarios like treaty reductions or exemptions.

What evidence should we keep for the sampling?

A sampling plan memo, selection lists with attributes and timestamps, working papers for each sample, a findings log with remediation, and reconciliations to ledgers and returns.

Can we do the sampling fully remote?

Yes. With a structured evidence pack and secure dataroom access, reviewers can perform the work remotely. Local data-residency rules may require on-site viewing.

QI sampling sizes (documentation vs. payments) — what passes in practice

QI sampling sizes (documentation vs payments) — what passes in practice

There is no one-size-fits-all number. Reviewers and ROs expect a risk-based, explainable sampling plan that covers documentation and payments/withholding, fits the bank’s scale/complexity, and produces defensible evidence. This guide shows structures that typically pass scrutiny in QI periodic reviews.

Scope reminder: align sampling to the QI Agreement appendices. Keep a written plan (populations, strata, size rationale, randomization method) and map each sample to evidence and findings.

1) The two core populations

Documentation population: all valid W-8/W-9 (incl. documentary evidence) in scope period, by account/entity type.
Payments population: all withholdable U.S. source payments and reportable events (incl. 1446(f) if relevant), by income code and payee type.

For each population, define: period, size (N), strata, exclusions, sampling method, sample size (n).

2) Documentation sampling — a practical pattern

Stratify by form type & risk. Higher-risk forms (e.g., W-8IMY with underlying allocations, entity W-8BEN-E with treaty claims) get larger slices; low-risk (e.g., recent W-9, simple W-8BEN individuals) can be smaller.

Stratum (examples)	Risk signal	Typical sample intent
W-8IMY (QI/non-QI) with underlying docs	Flow-through, allocation logic, cascading withholding	Larger slice + targeted checks on pooling/allocations
W-8BEN-E claiming treaty (entities)	Treaty LOB, chapter 4 status & GIIN validity	Moderate slice; test LOB, chapter 4, renewal controls
W-8BEN (individuals)	ID match, capacity, change-in-circumstances	Smaller slice; focus on validation & expiry handling
W-9 (U.S. persons)	TIN match/B-Notice handling	Small slice; test TIN match & backup withholding logic

How big? Use a base size per stratum (e.g., a fixed minimum) and scale with risk/volume. Document your rationale (why this covers plausible error). Randomize within each stratum.

3) Payments sampling — balance breadth and depth

Create strata by income code, rate bucket (30% / treaty / 0%), and payee type (individual, entity, intermediary/QI).

Payment stratum (examples)	Key tests	Sampling note
Dividends (27), portfolio interest (01)	Rate justification, chapter 3/4 status, exemption checks	Ensure both full-rate and reduced-rate cases are covered
Treaty-reduced payments	Treaty article/LOB evidence, residency, beneficial owner	Targeted oversample; link to underlying documentation
0% rate (statutory exemptions)	Exemption basis, QI/non-QI pooling, reporting footprint	Check for misclassifications; include edge cases
1446(f) / PTP transfers (if applicable)	Broker logic, exceptions, certifications & timing	Even a few cases warrant inclusion due to risk

Reconciliation: each sampled payment should reconcile to general ledger and 1042-S/1042 totals (and where relevant, 1099/backup withholding).

4) Sizing logic you can defend

Base + risk uplift: pick a base per stratum (e.g., 10–25 files/payments) and uplift for higher risk/volume.
Representativeness: every major form type/income code/rate bucket appears in the sample at least a few times.
Randomization: document the method (e.g., RNG with seed, spreadsheet RAND) and re-run only if technical issues occur.
Targeted top-ups: add directed cases (exceptions, overrides, manual handling) in addition to the random picks.

5) Evidence pack — what reviewers expect to see

Sampling plan memo: populations (N), strata, sizing rationale, RNG details, date/time, responsible person.
Selection sheet: unique IDs, attributes (form type, income code, rate), picker, timestamp, exclusion reason (if any).
Working papers: one sheet per sample with tests performed and pass/fail.
Findings log & remediation: issue description, population impact, fix/owner/ETA.
Reconciliation excerpt: tie-out of sampled payments to ledgers and returns.

6) Common pitfalls (and easy fixes)

Only one random pull: Always top-up with risk-based items (treaty, exemptions, intermediaries).
No documentation-payment linkage: Build a map from payee to W-8/W-9 (or underlying docs) for every sampled payment.
Ignoring changes in circumstances: Include accounts with renewals, re-papering, or status changes in period.
Missing 1446(f): If applicable, include at least a handful of PTP transfers/manual exceptions.
Weak rationale: Write a short paragraph per stratum explaining why the size is sufficient.

7) Quick-start templates

Use a simple tracker to define populations, strata, and selections. Replace the URLs below with your uploaded files after adding them to Media.

Sampling Plan Workbook (.xlsx)
Populations, stratification, RNG log, selection sheet, findings register.

Download workbook

Need a reviewer-ready sampling plan?
We’ll size and draft a plan aligned to your populations and risk profile.

Talk to an expert