How to monitor BPO quality on one independent scorecard

• Own the scorecard: one set of criteria for every BPO vendor and your in-house team.
• Grade independently of the vendors, using the conversations themselves.
• Compare teams like for like, by contact reason and line of business.
• Flag small samples so nobody is judged on a handful of conversations.
• Review evidence-linked results with each vendor and track the actions agreed.

When each business process outsourcing (BPO) vendor grades its own agents on its own form, you get numbers that can’t be compared and a vendor reporting on its own performance. Each vendor’s QA team works for the vendor. Their scorecards differ in wording, weighting and what counts as a fail, so a 90 from one vendor and an 85 from another tell you little about which team served customers better.

That gap has a cost. Gartner research on customer service BPO partnerships warns that “without effective vendor management, 75% of outsourcing savings can disappear within 18 months” ¹. Quality monitoring is the part of vendor management that tells you whether the service you’re paying for is the service customers get.

Write one scorecard that you own and apply it to every human team, outsourced or in-house. Vendors can keep running their own internal QA, but your scorecard is the one that decides how quality is reported to you.

A practical BPO scorecard covers four groups of checks:

• Competency. Accuracy of the answer, clarity, tone and empathy.
• Process. Required steps for each contact type, such as identity checks or required disclosures.
• Compliance. Privacy handling and any regulated language your business requires.
• Outcome. Whether the issue was resolved correctly under your policy.

Weight the groups to match what matters to your business, and write every check in plain language with examples of a pass and a fail. Our guide to building support QA scorecards that agents trust covers how to write checks that hold up when agents challenge them.

Every team should be graded on the same checks, because comparisons fall apart the moment one vendor gets a different form. Different contact types still need different steps, so handle that with “not applicable” rules: a check that doesn’t apply to a conversation is left out of that conversation’s score, so it counts as neither a pass nor a fail.

For example, an identity-verification check applies to account changes. Marking it not applicable on general product questions keeps a vendor that mostly handles product questions from getting credit for checks it never had to pass.

Monitor BPO quality from the conversations themselves, pulled from your own support platform, so you aren’t relying on exports the vendor prepares. If your outsourced teams work inside your helpdesk or a platform you can connect, the transcripts are already yours.

Write data access into the contract: every conversation a vendor handles should be available to you, with the vendor and agent identified, so you can grade it on your scorecard.

A single average per vendor hides more than it shows. Break scores down by the dimensions that explain differences:

Compare vendors on the same contact reasons. If one vendor handles most of your billing disputes and another handles password resets, their overall averages aren’t comparable. Their scores on the contact reasons they both handle are.

Flag small samples. A score built on a handful of conversations shouldn’t drive a coaching plan or a vendor penalty.

Before you hold a vendor accountable for a low score, check whether the miss was the agent’s, the process’s or the product’s. If every team fails the same check on the same contact reason, the cause is more likely a confusing policy, a broken tool or a product issue than a vendor problem.

Look for checks that fail across all teams at once. Those belong to your policy, operations or product owners. Checks that fail for one vendor only belong in that vendor’s review.

Review results with each vendor on a fixed cadence, such as monthly, and bring example conversations alongside the numbers. A review built on linked examples is harder to argue with and easier to act on.

A practical vendor review covers:

1. Overall score and each check group, against last period and against your in-house team.
2. The two or three checks with the biggest gaps, by contact reason.
3. Example conversations for each gap, so the vendor sees exactly what a fail looks like.
4. Coaching actions agreed for specific agents and team leads, with a date to check progress.
5. Any issues that belong to you, such as a policy that confuses every team.

Track the coaching that follows. A record of which agents were coached on which skill, and whether their scores moved on that skill, closes the loop between one review and the next.

If AI agents handle part of your volume alongside BPO teams, grade them too, side by side with your human teams. AI agents fail in different ways, such as loops and confident wrong answers, so give them their own bot rubric instead of forcing them onto the human scorecard. Viewing both together shows where work moves between the bot and outsourced teams, and where customers fall through the gap. Our guide on what AI agent QA is covers what that bot rubric should include.

Consider a hypothetical retailer with two BPO vendors and an in-house team. Vendor A’s overall score is noticeably lower than Vendor B’s, and the first instinct is to move volume to Vendor B.

Broken down by contact reason, the picture changes. Vendor A handles most refund requests, which have the strictest process checks. On the contact reasons both vendors handle, their scores are close. On refunds, every team, including in-house, misses the same step: confirming the original payment method before promising a refund.

The fix belongs to the retailer. It rewrites the refund procedure, adds a clearer example to the check, and reviews refund conversations again next month across all three teams.

An independent scorecard isn’t always worth the setup. Relying on your vendor’s QA fits when:

• You work with one BPO, for one simple contact type, and quality has been stable.
• Volume is small enough that your own team can read a meaningful share of the vendor’s conversations.
• You’re in a short trial with a vendor and will set up your own scorecard if you sign a longer contract.

An independent scorecard fits teams with more than one vendor, teams preparing for a renewal, consolidation or vendor switch, teams in regulated lines of business, and teams running AI agents alongside outsourced and in-house agents.

unitQ Support QA gives you an independent scorecard for human agents across in-house and BPO teams, plus a separate bot audit (configured for your workspace during setup), independent of the vendor, platform or labor being graded.

• One scorecard you own. Define each quality check in plain language: what it means, which answers count as a fail, and worked examples that show the edge cases. Your team owns and edits the rubric in Tag Studio in unitQ Support QA.
• Weighting and not-applicable checks. Weight rubric groups to match how you score quality. Checks that don’t apply to a conversation are left out of the score.
• Breakdowns that compare like with like. Break scores down by contact reason, line of business, BPO vendor vs in-house team, region, manager and agent, all on the same scorecard. Small samples are flagged. BPO vendor, manager and agent breakdowns use the team and vendor mapping set up for your workspace.
• One quality score with context. The Quality report in unitQ Support QA gives one quality score built from your rubric, with competency, outcome, operations and compliance views beside it.
• Coaching with a record. Coach in unitQ Support QA gives managers a roster of each rep’s strengths and gaps by rubric skill, a coaching brief per agent built from that agent’s conversations with linked evidence, a way to log the session with a goal per area, open coaching tasks that close when the session is logged, and coaching history by manager. Coaching impact compares coached agents’ score movement with a baseline of agents who weren’t coached.
• Evidence and coverage. Every scored miss links back to the conversation behind it, and unitQ Support QA scores the conversations you connect, not a sample.
• Your existing stack and metrics. unitQ integrates with support platforms such as Zendesk, Salesforce Service Cloud, Intercom and Freshdesk ³. Track quality alongside outcome metrics you already measure, such as CSAT, where that data is connected.