FrontLine for QA & Coaching — scorecards, evaluations, calibration, coaching that closes the loop.
QA isn't a quality gate. It's the input layer for coaching, training, PIPs, and skill certifications. FrontLine connects them — so a finding in QA Friday morning lands in the coaching plan Monday afternoon, not in a forgotten spreadsheet.
QA & Coaching dashboard
Acme BPO · 247 agents · 4 clients · 84% avg QA score
Avg QA score
84%
Evals this month
338
Active coaching
14
KB articles
147
QA volume by client · this month
RetailCo
142
FinCo
98
TelcoPro
76
Other
22
Recent QA activity
Coaching plan triggered · Tyler B. · 3 evals below threshold
5m ago
Calibration session scheduled · TelcoPro evaluators · 4 calls · Fri
1h ago
Agent dispute submitted · Maya P. · disputes score on eval #2891
3h ago
KB article published · FinCo · refund policy v3.2 · EN + FR
Yesterday
Active calibration sessions
4 active
Greeting Protocol
4 evaluators
ScoringPCI Handling
6
ResolvedRefund Flow
3
ScoringEmpathy
5
DraftWhat's hard about QA at a multi-client BPO
Generic QA tools (NICE Engage, Verint Quality Monitoring, Calabrio Quality) were built for one workforce, one scorecard, one calibration model. Multi-client BPO breaks every one of those assumptions.
Per-client scorecards — not one-size-fits-all
RetailCo wants 5 items scored. FinCo (regulated) wants 12 including disclosures and identity-verification. TelcoPro wants script adherence with 3-strikes rules. Generic QA tools force one scorecard per workforce. You end up running QA across three different spreadsheets.
Calibration drift across clients and evaluators
TelcoPro evaluators score harshly. FinCo evaluators are lenient. Aggregated workforce-wide, the QA number looks precise but means nothing. Per-client calibration is the only way the reports tell the truth — and most QA tools treat it as an afterthought.
Coaching plans that don't follow the agent
Mike's FinCo coaching plan flags disclosure-script gaps. He rotates to RetailCo; the plan stays in FinCo. The same gap surfaces in RetailCo evals 3 months later because nobody connected the dots.
Training assignments scattered across 3 systems
LMS has the courses. QA tools have the gaps. Manager has the PIP. The agent gets 3 assignments from 3 sources; nobody knows what was actually completed. Coaching becomes a "I told her to do that course" game of telephone.
Knowledge base fragmentation per client
RetailCo SOPs, FinCo SOPs, and shared policies all live in one KB. Mike searches "refund policy" on RetailCo's queue and gets all three. He reads the wrong one to a caller — SLA lost, calibration session opens Monday.
Reporting the COO and account managers both want
The COO wants enterprise-wide QA scores and evaluator productivity. The account manager wants RetailCo-only scores for the QBR. Same data, two roll-ups. Generic QA tools give one view; you rebuild the other in Excel every month.
What FrontLine does for your QA & Coaching team
Seven QA-and-coaching surfaces designed for multi-client operations from the ground up. The scorecard builder, evaluation workflow, and calibration sessions plus the coaching loop they feed into.
Feature 01
Scorecard builder — per-LOB scorecards, weighted sections, auto-routing
Build one scorecard per client + LOB. Sections, items, weights, scoring rubrics — all configurable from the UI. Bilingual labels. Auto-routes evaluations to the right evaluators based on LOB and skill match.
- Per-LOB templates. Different scorecards per client + LOB. FinCo gets the regulated 12-item form; RetailCo gets the 5-item conversation form. The data model carries the scorecard ID per evaluation; reporting respects the structure.
- Weighted sections. Greeting 10%, problem-solving 30%, closing 20%, etc. Per-section weights tunable by QA lead. Item-level weights inside each section. Final scores reflect the priority your client cares about.
- Per-item rubric. Yes / No, 1–5 scale, free-text reasoning per item. Add custom evidence requirements (call timestamp, quote, attached transcript).
- Auto-routing. Evaluations dispatch to the right evaluator based on LOB certification + workload. FinCo evaluations don't accidentally land with a RetailCo-only evaluator.
- Bilingual platform UI. Every QA surface — scorecard editor, evaluation form, agent acknowledgement — renders in EN or FR per the user's locale preference. Scorecard items are authored once in the language your QA team writes in today; per-item EN ↔ FR sibling translation is on the roadmap.
Scorecard · RetailCo CSR · v3
RetailCo · CSR conversation scorecard
5 items · 4 sections · weighted · active for RetailCo EN + FR queues
Opening
15%Greeting includes brand + name
Yes / NoAcknowledges customer reason for call
Yes / NoProblem solving
40%Identifies issue correctly
1–5 scale + quoteOffers correct resolution path
1–5 scale + quoteCustomer experience
30%Empathy + tone appropriate
1–5 scale + reasoningClosing
15%Confirms resolution + next steps
Yes / NoAsks if anything else
Yes / NoAuto-routes to RetailCo-certified evaluators · agent acknowledgement bilingual
Feature 02
Evaluation workflow — score, dispute, calibrate, close
Pick a call (random, sampled by score band, or supervisor-selected). Score against the rubric. Submit for agent acknowledgement. Disputes route back to a second evaluator. Calibration sessions run on disputed evaluations.
- Call selection modes. Random sampling, score-band sampling (e.g., score below 70 always reviewed), supervisor-pick, or escalation from a customer complaint. Quotas configurable per client.
- Interaction reference + evidence capture. Each evaluation links to the source interaction (call, chat, ticket) by reference. Evaluators capture quotes and timestamps inline as evidence on each scoring item. Native waveform-and-transcript playback inside FrontLine is on the roadmap (depends on the CTI / recording integration in spec 44); for now evaluators play the recording in your existing CCaaS / recording tool alongside the scorecard.
- Inline scoring. Rubric on the right pane, audio on the left. Score each item with a quote or timestamp as evidence. Free-text reasoning per item if needed.
- Agent dispute workflow. Agent can dispute the score with reason. The dispute routes to a second evaluator (calibration check). Both evaluators' scores stay on record; the dispute outcome is logged.
- Calibration loops on disputes. Disputed evaluations surface in the calibration queue — both evaluators' scores stay on record, the QA lead reviews variance per item, and the calibration outcome feeds back into evaluator drift reporting.
- Final score audit-logged. Every score change captured: who, when, why. When a client asks "what evidence backs this 72?", the answer is one screen.
Evaluation · EV-2891 · RetailCo · Maya P.
Inbound · RetailCo support · 5:42 duration
Recorded May 17, 14:08 · transcript synced
Audio · transcript synced · click to jump
Thank you for calling RetailCo, my name's Maya. How can I help?
Hi Maya, I'm trying to return something but the website isn't working.
I understand, let me look that up for you right away.
The order number is 41827.
Got it. I see the issue — let me file the return for you.
Scoring · 4 of 7 items completed
Greeting · brand + name
0:08 · "my name's Maya"
Acknowledges reason
0:22 · empathetic
Identifies issue correctly
0:38 · correct return path
Empathy + tone
neutral tone in middle
Resolution path
—
Provisional score: 86 / 100 · final after 3 remaining items + agent acknowledgement
Feature 03
Calibration sessions — multi-evaluator agreement, per client
Periodic sessions where multiple evaluators score the same calls. Variance surfaces drift; coaching plans for evaluators who drift. Per-client calibration so the scores you publish to RetailCo aren't compromised by TelcoPro-evaluator harshness.
- Multi-evaluator score capture. Same call, N evaluators score independently. Scores reveal automatically once all submit; no "I'll match yours" anchoring.
- Variance surfaced. Inter-evaluator variance computed per item and per section. High variance → calibration session item-by-item discussion → consensus.
- Anchor calls. Pre-scored exemplars used to train new evaluators. New hires score the anchor calls; their drift from the consensus is measured before they go live.
- Per-client calibration sessions. RetailCo evaluators calibrate together; FinCo evaluators calibrate together. Per-client scoring discipline maintained — the client's reports stay accurate.
- Evaluator drift visible to the QA lead. Per-evaluator variance against the calibrated consensus surfaces in the QA lead's view. Persistent drift is the trigger for a manually-assigned coaching plan — same product mechanic as agent coaching, opened by the QA lead based on the drift report.
- Calibration audit chain. Every calibration session logged: attendees, calls reviewed, variance scores, outcomes. Audit-grade evidence of QA discipline.
Calibration · RetailCo evaluators · May 17
Same call, 4 evaluators
RetailCo CSR · eval #2891 · variance detected on "Empathy + tone"
Pat R. (Lead)
RetailCo · 6 yrs
Sam C.
RetailCo · 2 yrs
Devon R.
Cross-client
Carlos D.
RetailCo · new
Variance: 13 pts on Empathy + tone (item 4). Group discussion → consensus 84.
Anchor call · pre-scored exemplar · new evaluator drift measured against this
Carlos D. trending +8 pts above consensus over 6 sessions → personal calibration coaching
Feature 04
Agent QA portal — agents see their own evaluations
Agents see their evaluations, coaching plans, training assignments, and skill certifications in one self-serve portal. Acknowledge evaluations, dispute scores, mark training complete — without supervisor mediation. Transparency builds trust.
- Self-view. Every evaluation in chronological order. Score breakdown by item. Quotes and timestamps attached as evidence. Agent sees what the evaluator saw.
- Acknowledgement workflow. Agent confirms they've reviewed. The acknowledgement is the audit anchor — when HR asks "did the agent see this finding," the answer is the timestamp.
- Dispute flow. Agent disputes with reason. Routes to calibration. Same workflow as the evaluator's view, just role-reversed.
- Coaching plan visibility. Agent sees their own coaching plan, milestones, evidence required. No "surprise coaching" — the plan is theirs to engage with.
- Training assignment list. Courses assigned (from QA findings, coaching plans, skill gaps) all in one list with progress. Agent doesn't get 3 emails from 3 systems.
- Bilingual interface. Agent sees the portal in their preferred language. Quebec agents see FR; Ontario agents see EN. Same data, two views.
Agent QA portal · Maya P.
Maya Patel · CSR · RetailCo EN support
Recent evaluations
Awaiting your acknowledgement
ActionAcknowledged
✓Acknowledged · disputed → 85
✓Acknowledged
✓Acknowledged
✓Your coaching plan
Empathy + tone skill build · 30-day plan started May 10 · M1 closed May 17
Training assigned to you
Empathy in customer recovery
Completed May 14
Advanced de-escalation
Due May 24
RetailCo refund policy v3.2
Due May 31
Feature 05
Coaching plan management — milestone-driven, evidence-backed
Coaches build plans with milestones, evidence requirements, and outcomes. Plans link to specific QA evaluations as starting evidence. Closing requires measurable improvement, not a manager check-mark. PIP linkage if coaching fails.
- Milestone-driven. Configurable intervals (typical 30 / 60 / 90 but tunable). Each milestone is a tracked review with required evidence types.
- Evidence inheritance. When QA findings trigger a coaching plan, the failing evaluations attach as starting evidence automatically. Coach starts from data, not memory.
- Manager + skip-level visibility. Same pattern as PIPs in HR: the agent's manager AND the skip-level see the coaching plan. Kills the "I didn't know my report was on a coaching plan" failure.
- Decision gating at milestones. At each milestone, the coach decides: closed (gap met) / continue / escalate to PIP. The decision is gated; the plan can't quietly expire.
- Escalation to PIP with evidence carried forward. At the final coaching milestone, if the gap isn't closed, the manager escalates to HR with one action — the coaching evidence stays attached so the PIP starts from data, not memory. Native PIP linkage (the bridge from coaching to PIP) is on the Atlas roadmap; today the escalation crosses to spec 28A's PIP workflow with evidence references.
- Audit chain across QA → Coaching → PIP. One record from the QA finding to the final decision. Audit-grade traceability for labour-relations defence.
Coaching plan · CP-128
Tyler Brennan · CSR · RetailCo
Coach: Pat Rivera · Manager: Sam Chen · Started Mar 18
Apr 17
May 17
Jun 16
Evidence inherited from QA + new
3 evals < 70 · inherited from QA findings
Coaching trigger
2 evals · avg 72 · 1 coaching session attended
Up from 64 baseline
4 evals · avg 78 · 1 escalation
On track
Awaiting final reviews · decision Jun 16
Confirm or → PIP
If M3 doesn't close: PIP auto-spawns in HR with all coaching evidence inherited.
Feature 06
Training assignment + LMS — courses, paths, certifications
A native LMS with course catalog, learning paths, completion tracking, certification issuance. Auto-assignments triggered from QA findings, coaching plans, or skill gaps. Compliance reporting for regulated certs.
- Course catalog. Per-client courses (FinCo-specific regulatory training) plus general (cross-client) courses. In-product authoring with structured content blocks; SCORM / xAPI ingestion is on the roadmap.
- Learning paths. Sequenced courses for role onboarding, skill development, regulatory recertification. Path assignment by role family or individually.
- Auto-assignments. QA finding (e.g., "missed disclosure") triggers the recommended course. Skill gap analysis from WFM triggers a path. Coaching plan adds courses as evidence requirements.
- Completion + assessment. Pass/fail per course with assessment. Score capture for compliance evidence.
- Certification issuance. Pass a certification course → cert issued, recorded in the skills matrix (WFM), tracked for expiry. Re-certification reminders 30/60/90 days before expiry.
- Compliance reporting. Per-client compliance status: which agents have which certs current. The report you hand to a client auditor without rebuilding it in Excel.
Training catalog + assignments
Course catalog · per-client + general
247 enrolled across 22 active coursesRetailCo refund policy v3.2
RetailCo · 25 min
FinCo regulated disclosures
FinCo · 1 hr
Empathy + de-escalation
General · 45 min
Bilingual customer interactions
General · 30 min
Learning path · CSR onboarding · 5 days
Welcome + platform basics
Multi-client confidentiality
Client-specific systems · RetailCo
Empathy + de-escalation
Day-1 readiness check
Completion → cert added to skills matrix · routing engine sees new eligibility within 1 hr
Feature 07
Knowledge management — per-client SOPs, audience-scoped
Articles scoped per client + LOB via database-enforced audience subqueries. Article freshness tracked; stale articles flagged for review. Article gap detection (recurring agent questions surface as KB needs).
- Per-client + LOB scoping. RetailCo SOPs scoped to RetailCo audiences. FinCo SOPs scoped to FinCo. Shared general policies tenant-scoped. Audience scope runs as an EXISTS subquery on every list, search, and detail endpoint.
- Bilingual platform UI today. Every KB surface — author, editor, agent reader, client portal — renders in EN or FR per the user's locale. Articles can be authored in either language; the structural EN ↔ FR sibling-row pairing (so a French article links to its English source and stale-flags when the source changes) is on the Atlas roadmap.
- Article freshness. Last-reviewed timestamp per article. Articles past their review interval flag stale; the owner gets a notification, then the reviewer is escalated 30 days later if no action.
- Article gap detection. Recurring agent searches that return no good answer flag as KB gaps. The KM lead sees a queue of "articles needed" derived from real usage, not anecdotes.
- Audit trail. Who edited what, when, why. When a client asks "when did your SOP change," the answer is the article history.
Knowledge base search · context-scoped
“refund policy”
EN · FRAgent context: Maya P. · RetailCo EN support · results scoped to RetailCo + general
Results · 4 articles
RetailCo refund policy v3.2
RetailCo · updated May 12
Refund timelines + customer notice
RetailCo · updated Apr 28
Exception handling — VIP customers
RetailCo · updated Mar 04
General refund principles + escalation
General · updated Feb 14
12 agent searches in past 7 days returned no result for "FinCo dispute resolution" → KB gap detected
The performance loop — QA → Coaching → PIP → resolution
When QA evaluations surface a recurring issue, the QA lead opens a coaching plan with the failing evaluations attached as starting evidence — one click, no rebuilding the case in email. The coaching plan runs through configurable milestones; at the final milestone the manager closes the plan or escalates to HR with the coaching evidence carried into the PIP. The whole chain is one workflow with role-tuned views: QA sees evaluations, coaching team sees the plan, manager and HR see the PIP, agent sees their portal.
- QA → Coaching with one-click open. Recurring findings surface in the QA lead's queue. Open a coaching plan and the failing evaluations attach as starting evidence — the coach starts from data, not memory.
- Evidence inheritance. The failing evaluations attach as starting evidence on the coaching plan. The coach doesn't rebuild the case; the data is already there.
- Manager + skip-level visibility. Same pattern as PIPs in HR. Both the manager and their manager see the coaching plan from the start.
- Escalation to PIP at the final milestone. If the agent doesn't meet the final milestone target, the manager escalates to HR with one action. The coaching evidence carries into the PIP — no case rebuilding. Native auto-spawn of the PIP from coaching is on the Atlas roadmap (spec 28A pip_support_bridge); today escalation is manager-initiated.
- Cross-team status visible to all. QA sees coaching progress; coaching sees PIP status; manager and HR see the whole chain; agent sees their own view in the portal. One source of truth, four role-tuned views.
- One audit chain spans QA → Coaching → PIP. From the first failing evaluation to the final decision — one record. Labour-relations defence is forensic-grade because the entire performance path is in one place.
Performance loop · Tyler B.
Tyler Brennan · CSR · RetailCo
QA → Coaching → PIP chain · single record
QA score trend · last 6 evaluations
3 consecutive evals < 70 → coaching plan auto-triggered Mar 18 · all 3 evaluations attached as evidence
Coaching plan CP-128 · M2 of 3 in progress · avg score now 78 · on track
M1 · Apr 17 · closed
M2 · May 17 · in review
M3 · Jun 16 · decision gate
PIP status
Pending M3Not triggered · M3 will decide (Jun 16). If gap not closed → PIP auto-spawns with all coaching evidence inherited.
Audit chain: QA finding (Mar 04) → coaching trigger (Mar 18) → M1 closed (Apr 17) → M2 in review (May 17) → M3 decision (Jun 16). One record, four cross-team views.
Business outcomes for the people running the BPO
What this looks like in margin terms for the COO or owner.
Coaching cycle compresses — fewer escalations, fewer wash-outs
One-click coaching open + evidence inheritance compresses the typical coaching cycle from QA finding to plan close. The 90-day washout curve flattens because the right finding lands in the right plan within hours, not weeks.
QA scores trend up when coaching closes the loop
Coaching plans that link back to specific QA findings have measurable closure. Trend direction shows up in our reference design partners; magnitude depends on your scorecard structure, calibration discipline, and starting baseline.
Every QA finding traceable through coaching to outcome
From the first failing evaluation to the final disposition (closed / extended / terminated), one record. Labour-relations and client-audit defence baked into the workflow.
Same QA team handles multi-client coverage without dilution
Per-client scorecards, calibration sessions, and routing mean adding a new client doesn't double your QA headcount. The team scales linearly with seats, not exponentially with clients.
What QA & Coaching actually gets back
Directional outcomes — real magnitude depends on your scorecard structure, calibration discipline, and starting baseline.
Faster from QA finding to plan close
Evidence inheritance + one-click open compress the typical cycle. Magnitude depends on your prior workflow.
Three systems → one platform
QA tool + LMS + coaching tracker collapse into one platform with one source of truth and one audit chain.
Zero lost coaching plans
Plans open from the QA queue; the plan can't fall through. The "I forgot to follow up" failure mode goes away.
Self-serve, no supervisor chase
Was: manual supervisor chase. Now: agent self-acknowledges in portal. Disputes route to calibration, not arguments.
Regulatory posture
FrontLine QA & Coaching vs. a generic QA tool
Six things BPO QA teams notice in their first month away from NICE Engage / Verint Quality Monitoring / Calabrio Quality.
| FrontLine QA & Coaching | Generic QA tool | |
|---|---|---|
Per-client scorecards at the data layer (not workforce-wide) Generic QA forces one scorecard for the workforce; you need different forms per client + LOB. | ||
Native coaching loop with one-click open from QA findings Generic QA stops at the evaluation; coaching + PIP live in separate tools. The handoff happens in email. Auto-spawn of PIP from coaching is on the FrontLine roadmap; manager escalation today carries evidence forward. | ||
Bilingual platform UI (EN ↔ FR) across QA, agent portal, and KB Every UI surface renders in the user's locale today. Per-article and per-scorecard-item EN ↔ FR sibling-row content pairing is on the FrontLine roadmap. | ||
Knowledge base scoped per client + LOB Generic KBs are workforce-wide; agents see all clients' SOPs. Scoping is manual. | ||
Multi-evaluator calibration sessions per client Some QA tools support calibration; few do it per client with per-client drift detection. | ||
Scoring + coaching + training + PIP in one platform Generic stack: scoring in QA tool, coaching in LMS, PIP in HRMS — three systems, two handoffs. (Recordings continue to live in your CCaaS and are referenced by ID; in-platform playback is on the Atlas roadmap — spec 44.) |
In production today
Quality Assurance (Module 30) is shipped and running on customer tenants. Training & Coaching (Module 31) and Knowledge Management (Module 32) are shipped. See the technical status, full feature list, and weekly progress on the Atlas.
Questions QA managers actually ask before signing
Pulled from real fit calls. Short, direct answers.
How do scorecards relate to clients? Can we run different scorecards for RetailCo vs FinCo?+
Can we calibrate per client? Our TelcoPro evaluators score harshly and our FinCo evaluators are lenient.+
How does coaching escalate to a PIP? We don't want auto-PIP for one bad call.+
Can FrontLine ingest recordings from our CCaaS for QA evaluation?+
Can agents see their evaluations? Can they dispute scores?+
Ready to see it in your environment?
Two ways to take the next step.
Other teams using FrontLine
QA doesn't live alone. Here's what the rest of FrontLine looks like from the team next door.