How to Run a BPO Vendor RFP: 18 Questions That Actually Matter

The standard BPO RFP template has not meaningfully changed in 15 years. It asks about agent counts, languages supported, hours of operation, certifications held, and references available. The vendors all answer roughly the same way. The procurement team makes a decision based mostly on price. The program goes live. Six months later, the quality numbers tell a story no one expected. The RFP did not surface the real differences. Here are the questions that do.

Questions about the operating model (not the org chart)

The operating model is the single biggest predictor of program performance, and almost no RFP asks about it directly.

1. Who owns the program day-to-day, and what is their background? "Account manager with operations escalation" is a different operating model from "operator leading the program directly." The first one introduces a translation layer between you and the people running the floor.
2. How often does program leadership change hands? The honest answer in most BPOs is every 6-9 months. That is the source of most quality drift.
3. Walk us through a calibration session. Not the description on the slide. The actual cadence, the participants, the artifacts produced. Vendors who run real calibration can describe it in three minutes. Vendors who do not give you marketing language.
4. What percentage of calls do you actually QA, and how do you decide which ones? The legacy answer is 3-5% sampled, often biased toward the easiest calls. The right answer in 2026 is closer to 100% via AI-assisted scoring, with human calibration on a stratified sample.

Questions about technology (not the platform name)

Asking which CCaaS platform a vendor uses tells you almost nothing. The real questions are about what the vendor does with the data the platform produces.

5. What signals from the call do you score, and which do you act on in real time? "We have AI" is not an answer. "We score sentiment per utterance, flag compliance language in real time, and route churn intent to a save desk within 30 seconds of detection" is.
6. Can you show us live dashboards from a current client operation? Vendors who can show their work do. Vendors who cannot will offer a sales engineer demo on a sandbox.
7. What is your data ownership and portability story? If your call recordings, transcripts, and intelligence objects are locked in their platform, you have a switching cost problem starting day one.
8. How does your platform handle prediction, not just scoring? Churn prediction, burnout prediction, escalation prediction. Scoring is rear-view. Prediction is operational.

Questions about the people (not just the headcount)

Most RFP responses lead with agent counts. Agent counts predict almost nothing about program quality.

9. What is your annual agent attrition rate, and how is it calculated? Industry norms vary wildly by region. Get the methodology. Attrition above 60% is a quality death spiral.
10. What is your average agent tenure at program go-live versus 12 months in? This number tells you whether their hiring is selective or volume-based.
11. What does your training look like for industry-specific workflows? Generic "communications and customer service" training is the tell. Real training is industry-specific, workflow-specific, and refreshed quarterly.
12. How do you handle agent burnout detection and intervention? Most vendors detect burnout when the agent quits. The right answer involves conversational signal monitoring and intervention before attrition.

Questions about commercial structure (not just the rate card)

13. What does your pricing model look like beyond the per-FTE rate? Setup fees, training pass-through, technology licensing, calibration overhead, reporting fees. The "loaded rate" is the answer that matters.
14. What are the unwind terms? If we need to exit the contract in month 8, what does it look like? Vendors who make unwinding hard are betting against you.
15. How is performance tied to commercial terms? SLA penalties are the legacy answer. XLA-tied commercial structure is the operator answer.
16. What does a 60-day challenger pilot look like? Any vendor unwilling to run a defined-slice pilot against an incumbent is telling you something. Listen.

Two final questions that surface the real culture

17. Tell us about a program you walked away from in the last 12 months. Why? Vendors who never walk away from anything are running everything for the revenue, including programs they cannot deliver well. The honest answer is more reassuring than the perfect one.
18. Who runs our CX review? Will we talk to them again after the contract is signed? The CEO at the sales meeting, the account manager after the SOW. That gap predicts where the program goes.

An RFP run on these 18 questions will surface 80% of the differences between BPO vendors. The other 20% only shows up in the first 90 days of the program.