Scaling a new support channel without sacrificing quality
SaaS / Conversational commerce · Designing and scaling a new customer-support channel · Jul 2022 — pilot handoff · via Hugo
A brand-new support channel had to launch across dozens of brands on unfamiliar technology, with an aggressive ramp — and speed could not be bought by dropping quality.
Onboarding, not agent capability, was the constraint: new starters were being asked to learn tooling, brand voice and throughput expectations at the same time.
Redesigned onboarding and performance management for the channel: staged certification, brand-voice standards, throughput coaching and quality flag monitoring.
Onboarding pass rate rose from 67% to 90% against a 60% target, throughput reached 33 sends/hour against a 28 target, with quality flags at 0.09% against a 1% ceiling.
Overview. The client is a leading conversational-commerce platform, reinventing business-to-consumer communication through SMS. Hugo ran its outsourced customer support pilot, launched 18 July 2022, handling inquiries and approved troubleshooting for dozens of the platform's own retail and e-commerce brands through its CRM.
I served as Team Lead through the pilot with 10+ agents across two onboarding batches, owning training, quality and performance — then stepped up as Operations Manager on the standing post-pilot team once the account converted.
- Led the pilot as Team Lead over 10+ agents across two onboarding batches, reporting to both Hugo and client leadership.
- Built and delivered the agent training curriculum, including brand-research “scavenger hunt” assignments across a dozen real client brands.
- Took over training mid-pilot after an initial 67% pass rate and pushed it to 90% against a 60% target.
- Owned an 8-category QA taxonomy and the weekly feedback loop that held quality flags at 0.09%.
- Authored the pilot's Weekly Business Reviews and the final client-facing Post-Pilot Report.
- Transitioned the account into steady state and continued as Operations Manager on the standing post-pilot team.
This was a from-zero pilot: a brand-new CRM agents had never touched, and a client base that wasn't one brand but dozens — apparel, footwear, photography gear, hair extensions, electronics, furniture — each with its own notes, templates and escalation rules.
The success criteria were explicit and aggressive: sends-per-day and sends-per-hour targets that climbed every week, a hard ceiling on quality flags, and an onboarding pass-rate bar the client had historically struggled to hit even at 60%. Speed and accuracy were both being tested at once, on unfamiliar ground, from day one.
A 5-day, brand-immersion onboarding curriculum
Day 1 covered the Concierge platform and compliance fundamentals through guided tutorials and knowledge checks. Days 2–3 put agents through seven brand-specific research assignments on real client brands, building the research muscle before anything went live, plus hands-on demos of restocks, promo codes, order edits and cart-link features. Day 4 moved into trainer and peer shadowing plus a Slack crash course. Day 5 onward was nesting — live conversations worked side-by-side with a trainer before flying solo.
The client's own pass rate for this process had been 60%; my pilot team hit 67% on that model, and once I took over training directly, the pass rate reached 90%.
| Flag category | What it catches |
|---|---|
| Abandoned Customer | Skipping a conversation, or asking questions that should have been obvious |
| Acknowledgement / Empathy | Failing to acknowledge what the customer is experiencing |
| Critical Thinking | Misreading the issue, or missing nuance that should have been clear |
| Notes / Templates | Not using the brand's own notes and templates correctly |
| Research | Insufficient research into the brand, product or order before replying |
| Workflow / Process | Deviating from the required escalation or handling process |
| English / Soft Skills | Grammar, structure, tone-matching and personalisation quality |
| Missed Conversion | Missing a clear opportunity to recommend a product or drive a sale |
Closing the quality loop weekly
Every week, regardless of which agent was flagged, the whole team reviewed the flags together, plus best-practice syncs with Hugo's top performers and the client's own team. After the first sync, average 'Needs Work' flags per agent dropped 71% week-over-week (2 in Week 3 vs 8 in Week 2), and a later push on note-taking habits and pre-shift brand-update review drove a further 33% drop (14 in Week 6 vs 21 in Week 5).
| Week | Sends/day target | Actual | Sends/hr target | Actual |
|---|---|---|---|---|
| Wk 1 | 100–120 | 133 | 12.5–15+ | 14 |
| Wk 2 | 120–140 | 166 | 15–18+ | 16 |
| Wk 3 | 160–180 | 208 | 20–23+ | 26 |
| Wk 4 | 180–200 | 216 | 23–25+ | 27 |
| Wk 5 | 200–220 | 247 | 26–28+ | 31 |
| Wk 6 | 220–240 | 263 | 28–30+ | 33 |
Training a second cohort that outperformed the first
By the final week, average sends/hour had grown 136% from where the team started, and total pilot sends reached 57,051 — 13,649 in the final week alone — while the skip rate held at 4% against a 15% ceiling.
Partway through, I onboarded a second batch onto the now-refined training model. Despite starting a week later, Batch B held its own on total output, and its top performer led all agents with 6,207 total sends, finishing ranked #10 on the client's own company-wide leaderboard at 41 sends/hour. A clean signal that the curriculum, not just tenure, was driving performance.
| Metric | Result |
|---|---|
| Onboarding pass rate | 90% vs 60% target (from a 67% baseline) |
| Sends/hour, final week | 33 vs 28 target (+18%) |
| Sends/day, final week | 263 vs 220–240 target |
| Quality flags | 0.09% average vs <1% target |
| Skip rate | 4% average vs ≤15% target |
| Total pilot sends | 57,051 (13,649 in the final week) |
| Post-pilot steady-state volume | 308,204 outbound messages; 103,228 closed chats |
- Onboarding pass rate60% cap → 90%
- Quality flag rate against ceiling0.09% vs <1%
- Productivity targets met or beaten6 of 6 weeks
- Sends/hour growth across pilot+136%
- Turned a training process the client had capped at 60% pass rate into one that reliably hit 90%, without lowering the bar.
- Every productivity target in the ramp plan was met or beaten while quality flags stayed roughly 10x under the ceiling.
- Handed the QA taxonomy, curriculum and reporting cadence to the standing team intact, rather than leaving it in my own head.
More case studies
Full index- 01US digital pharmacy clientDiagnosing the root causes of customer and compliance failures, separating agent performance from systemic issues, and building the governance to improve both safely at scale93% CSAT vs 90% target
- 02EdTech learning marketplaceReducing recurring customer dissatisfaction through frontline capability84% drop in issue-recognition DSATs
- 04Clinical decision-support platformRedesigning support operations to achieve 100% SLA compliance100% SLA compliance, Q1–Q3