The Hidden Cost of Running Clients in Separate Tools
Most BPOs run each client account in its own copy of the WFM software: a separate tenant, a separate login, a separate report pipeline. It looks tidy. It costs you about 12% of your operating margin.
Serge Belov
Founder, FrontLine · Published May 12, 2026
Most BPO (Business Process Outsourcing: a firm that runs contact-centre operations on behalf of other brands.)s run each client account in its own copy of the WFM (Workforce Management: forecasting, scheduling, and adherence.) software. One tenant for the bank, one for the airline, one for the SaaS company. It looks tidy on the org chart. The agents log in once and only ever see their own client. The reports go to the right contact. The auditor leaves happy.
It costs you about 12% of your operating margin and you'll never see it on a line item.
Where the cost actually lives
When the same agent is licensed across two tenants, you're paying for that seat twice, most WFM vendors charge per agent per tenant, and a multi-client agent shows up in both meters. That's the smallest piece.
The bigger costs are operational. A supervisor managing a cross-trained agent has to pull two reports and reconcile them by hand to see total productivity. A schedule conflict between Client A's Monday morning peak and Client B's Monday afternoon training is invisible until an agent calls out. QA (Quality Assurance: the program that scores and reviews agent interactions.) scores live in two systems and the cross-LOB (Line of Business: a distinct client program or queue within an operation.) performance review becomes spreadsheet work that doesn't get done.
Then there's the time the operations director spends explaining to a new sales prospect why the demo only shows one client at a time, because that's the only way the product was built to be shown.
The number, with the math
Take a 300-agent BPO running four client accounts in four separate WFM tenants. Annual revenue around $12M (Canadian, fully loaded, contact-centre-typical). Operating margin in the 10–15% band, so $1.2M–1.8M of margin to defend. The hidden cost shows up in five places.
Duplicate licensing. About 25% of agents in a BPO this size are cross-licensed across two accounts. That's 75 agents paying the per-seat fee twice. At a typical $50/agent/month for a mid-tier WFM platform, the duplicate spend runs about $45K a year. It's the easiest cost to see and the smallest piece.
Supervisor reconciliation. Supervisors who manage cross-trained agents lose three to five hours a week pulling reports from each tenant and combining them in Excel, total productivity, total adherence, total coaching minutes. Twelve of the thirty supervisors at this size do this work. At fully-loaded supervisor cost of around $60/hr, the annual bill is $112K–$187K. This is the largest line item and the one that's almost never on a budget anywhere.
Missed schedule conflicts. When the same agent is scheduled for Client A's Monday morning peak and Client B's Monday afternoon training, neither system catches it because neither knows about the other. The miss surfaces as a call-out, an SLA (Service Level Agreement: a contractual performance target, e.g. answering X% of calls within Y seconds.) breach, or a back-filled shift at overtime rates. One incident per month at $5K–$10K of avoidable cost is conservative for this size: $60K–$120K a year.
QA-coaching cycle drag. A QA flag in one tenant doesn't trigger a coaching session in the other. The cross-LOB performance review that should run weekly runs monthly because the data lives in two places. The cost lands as slower remediation of underperformance, which compounds into attrition the operator pays for elsewhere. Hard to monetize directly, but operators we've talked to estimate $40K–$80K a year of avoidable repeat issues per 300 agents.
Audit and RFP prep. Every SOC 2 or client-driven audit cycle runs once per tenant. For four tenants, that's four times the document prep, four times the access review, four times the evidence-pull. A two-week prep cycle at $5K of loaded analyst time runs $40K a year, more if a major client triggers an annual review.
Sum the conservative end of those ranges and you get $297K. Sum the higher end and you get $477K. For a BPO with $1.2M–1.8M of operating margin to defend, the gap eats 18–30% of margin on the high end and 16–25% on the low end. The 12% figure in the excerpt sits below that range deliberately; it's what an operator can recover net of any consolidation cost. The actual gross hit is larger.
These are operator-side estimates, not published benchmarks. Industry attrition and integration benchmarks (ContactBabel 2024 , MuleSoft 2024 ) put cross-system reconciliation cost at $200K–$400K per 1,000 agents annually, which scales reasonably to the figures above. Treat these as the shape of the number, not the exact number for your operation. Your supervisor-to-agent ratio, your cross-licensing rate, and your incumbent WFM price card all move it.
What multi-client native actually looks like
The short definition: one tenant in the platform, many client accounts inside it, with access scoped at the data layer to `client_account_id` and `client_lob_id`. The platform itself understands that an agent can belong to more than one client, that a supervisor's view is bounded by their assignment, and that a client logging into the portal can structurally never see another client's data.
The hard part isn't the definition. It's the consequences the data model has to support without the operator wiring them up themselves.
A composite performance view that weights an agent's QA across Client A's scorecard and Client B's scorecard with their respective contractual definitions of "meets expectations." A supervisor's queue that filters down to only the agents she actually supervises, even when those agents work for different clients on different shifts. A schedule that prevents the system from booking the same agent into two clients at the same hour without an explicit override. A reporting layer that can roll up across clients (for the BPO's own COO dashboard) and slice down per client (for the client's QBR) using the same underlying data.
Generic HRMS and WFM tools break in three predictable places when asked to do this. Role-based access control assumes one organisational hierarchy; you get either one with the client structure compiled in (which means every new client needs the access matrix re-engineered) or one without it (which means you filter in application code and pray every screen got it right). Reporting joins assume one set of dimensions; cross-client roll-ups become an ETL project. Schedule conflict detection assumes one queue per agent; multi-client agents either get over-booked or get manually scheduled outside the optimiser.
The fix is structural. The Atlas is the easiest way to see what each piece looks like in a platform built around the multi-client model from the start: Recruiting and Onboarding under one workflow, WFM with concurrent client assignments as a first-class concept, QA that applies different scorecards per client to the same agent's calls, and a client portal that the database itself prevents from returning the wrong rows.
If the product was built for one tenant and patched into many, you'll find the seams. The seams cost money.
The exit move
Consolidation looks scary on a Gantt chart and is usually less disruptive than the chart suggests. The reason: most of the cost above is recurring, so the first quarter post-consolidation pays back most of the migration cost, and the second quarter starts banking margin. The sequence below is the one that's worked for the BPOs I've watched go through it.
Quarter 1: data export and shape audit. Before you migrate anything, pull a clean export from each tenant. Map every field to your target schema. The fields you'll find don't line up are the ones that matter; they're where the old tools were quietly representing different things by the same name. Common offenders: "client" (some tools mean the BPO's customer, some mean the end-consumer the agent is helping), "shift code" (often per-tenant proprietary), and "QA score" (the weighting model differs per scorecard).
Quarter 1–2: pick the cutover order. Migrate your smallest, lowest-SLA-risk client first. The goal is to learn the migration before doing it on the client that would notice a hiccup. Save the largest or most sensitive client for last. If you have a client mid-renewal, that's the one to move second, the renewal window is the natural cover for any small reporting cadence changes.
Quarter 2: re-licensing and supervisor retraining. Cross-licensed agents move to single records with explicit multi-client assignment, which usually reduces the seat-count line on your invoice by 20–30%. Supervisors get a half-day session on the composite view: how to read a cross-client performance summary, how to spot a scheduling conflict before the system warns about it, how to action a QA flag that came from a client they don't directly manage.
Quarter 2–3: the client conversations. Every client gets a heads-up note: the report cadence stays the same, the contact stays the same, the SLA stays the same, the underlying platform changes. Most clients don't care; a few will ask security and audit questions, and the only real preparation needed is having the SOC 2 and PIPEDA documentation ready for the new tenant. Two clients will want to test the new portal access. Schedule that into the rollout plan up front.
MSA review, in parallel. Almost every existing MSA was written assuming a specific reporting tool or tenant arrangement. The change of underlying platform usually doesn't require an amendment, but the more careful clients will ask, and having a one-paragraph addendum ready (same controls, same reports, same SLAs, different infrastructure) keeps the legal review from becoming the long pole.
The whole sequence runs about six to nine months for a 300-agent operation. The recurring cost savings show up in the first full quarter of consolidated operation; the operational improvements (faster QA cycle, fewer missed schedule conflicts, lighter audit-prep load) compound over the following year. For the operators who've done it, the most common comment a year later is some version of "I can't believe we paid for the duplication that long."
Sources
The published industry benchmarks referenced in the math section are listed below. The dollar figures attached to each line item in the breakdown are operator-side estimates rather than published benchmarks, and the article flags them as such.
Contact-centre cost benchmarks. ContactBabel, US Contact Center Decision-Makers' Guide, 2024 . Annual benchmark on US contact-centre operating costs, agent metrics, and attrition; the closest published reference for the supervisor-reconciliation and duplicate-licensing line items.
Cross-system integration cost. MuleSoft, Connectivity Benchmark Report . Annual enterprise integration benchmark. The 2024 edition documents an average $4.7M of annual integration spend and 39% of IT-team time spent on integrations across enterprise IT; the per-1,000-agent reconciliation cost in this essay is derived by scaling those figures to a BPO operation.
All other figures (per-seat WFM pricing, supervisor hourly cost, the 25% cross-licensing assumption) are based on operator-typical numbers from the BPOs we've talked to. Where a figure is an estimate rather than a benchmark, the article flags it as such.
Serge Belov
Founder, FrontLine
Three decades building software for BPOs. FrontLine is the workforce platform BPO leaders kept asking for and never quite got.