BPO vendor evaluation: the exact questions to ask before signing
Most BPO evaluations focus on price per document. The questions that actually predict a bad six months later are different — here they are, in order.
A BPO evaluation usually starts and ends with a price sheet: dollars per document, tiered by volume. That number matters, but it predicts almost nothing about whether the relationship works six months in. The questions that do predict it rarely make it onto a vendor comparison spreadsheet.
Start with accuracy, not price
Ask for the vendor's actual accuracy rate, broken down by document type and jurisdiction — not a single blended number. A vendor that reports a high accuracy figure on clean, single-currency domestic invoices and folds that into the same headline as multi-currency, multi-jurisdiction paperwork is describing two different products with one number. Ask how the rate is measured (against what ground truth, over what sample size) and whether it's independently audited or self-reported.
Ask what happens when volume spikes
A vendor that performs well at steady-state volume can behave very differently during a month-end crunch or a client's seasonal peak. Ask specifically: what is the largest single-day volume this vendor has handled without degrading turnaround, and what changes operationally when volume doubles unexpectedly — does processing slow down, does accuracy drop, does a human queue simply grow silently in the background? What actually breaks past a much larger volume threshold is worth reading in full before signing anything sized for real growth.
SLA and security deserve their own checklists, not a paragraph each
Turnaround guarantees and data-security posture are two of the highest-stakes questions in this whole evaluation, and each deserves more depth than a vendor call has time for — the specific SLA questions that separate a real guarantee from a marketing number and exactly what to ask about data access, retention, and model training are both worth working through in full before the accuracy and corrections questions below even come into play.
Ask about vendor concentration risk
A single BPO handling 100% of a company's document volume is a single point of failure — if that vendor has an outage, a security incident, or simply a bad quarter, the client absorbs the full impact with no fallback. Ask what a transition would actually look like if the relationship ended: is the client's historical data portable in a standard format, or does switching mean starting over. A vendor confident in their own service has a straightforward answer; one who gets evasive about portability is signaling something worth noting.
Ask how corrections are handled — and whether they're remembered
Every AP automation vendor makes mistakes; the honest ones say so upfront. The real differentiator is what happens after a correction: does the system remember it for that specific vendor going forward, or does the same coding error recur every month indefinitely? Ask to see the correction-to-recurrence rate, not just the initial accuracy rate — a vendor that can't answer this hasn't been asked before, which is itself informative.
The comparison table a vendor won't hand you
| Question | Weak answer | Strong answer |
|----------|-------------|----------------|
| Accuracy rate | One blended percentage | Broken down by document type and jurisdiction, with methodology |
| Volume spike behavior | "We scale seamlessly" | A specific largest-volume-handled figure and what changes under load |
| SLA | "Fast turnaround" | A percentile target with a named exception process and remedy |
| Data access | "SOC 2 compliant" | Named roles with access, retention period, no training use |
| Corrections | "We fix errors" | A measurable correction-to-recurrence rate over time |
Related reading
- What clients now expect from a BPO's data security: SOC 2, GDPR, and the questions that actually matter
- AP outsourcing accuracy benchmarks: what "best in class" actually means
- BPO pricing models: per-seat vs. per-document, compared
FAQ
Should price still factor into the decision?
Yes, but as a tiebreaker between vendors that have already passed the accuracy, SLA, security, and correction-handling checks — not as the first filter. A cheaper vendor that fails on any of those dimensions costs more in rework and risk than it saves on the invoice.
How long should a trial period run before committing?
Long enough to include at least one volume spike (a month-end close, a seasonal peak) — a trial confined to steady-state volume never tests the scenario most likely to cause problems later.
Is a written SLA legally enforceable, or just a target?
That depends entirely on the contract language — a "target" and a "guarantee with a remedy" are different legal commitments, and the difference is usually buried in a single clause. Read that clause before signing, not after the first missed deadline.
Should a reference check be part of this evaluation?
Yes, and specifically ask a reference about the failure scenarios above — a missed SLA, a volume spike, a correction that recurred — rather than general satisfaction, since a reference call that only confirms "they're happy" skips exactly the information a price sheet and a sales call already couldn't provide.