Nobody prices a losing account deliberately. Programs go underwater gradually, through a series of small differences between what was assumed at pricing and what operations actually delivered.
That’s the pattern worth holding onto. Almost every leak below is the same shape: an assumption, a reality, and a gap that compounds quietly across a contract term.
Why These Leaks Stay Invisible
Your profit and loss statement shows salaries, technology, facilities and overhead. It does not contain a line called “shrinkage we underestimated by four points.”
Leaks hide in the difference between two numbers, only one of which appears anywhere. The pricing model lives in a spreadsheet somebody built during the bid; the operational reality lives in your platform reporting; and the two rarely get compared after the contract is signed. I’d argue that single omission explains more disappointing accounts than any commercial term does.
Organizational structure makes it worse. Whoever built the bid model has usually moved to the next opportunity by the time delivery begins, and whoever runs the account inherits targets without the reasoning behind them. Ask an operations manager why their program assumes 24% shrinkage and you will frequently get a shrug, which is not incompetence so much as a handover that never happened. Nobody owns the comparison because nobody owns both halves of it.
Shrinkage Assumed Rather Than Measured
The most common leak, and the one carrying the largest absolute value.
ACXPA describes shrinkage as the percentage of paid time agents are not available to handle contacts, covering breaks, meetings, training, leave, sickness, coaching, late starts and system downtime, and separates planned from unplanned components (ACXPA).
Price a program at 25% and run it at 32%, and you have committed to delivering coverage from roughly a tenth fewer productive hours than your model assumed. Every service level breach after that traces back to a spreadsheet cell.
Unplanned components cause most of the variance. Planned items such as leave and scheduled training are knowable in advance and generally modeled adequately; sickness, late starts, system downtime and unscheduled coaching are the ones that drift. Track the two separately, because a rising unplanned figure is an operational signal worth acting on, while a rising planned figure usually just reflects a decision somebody already made deliberately.
Attrition, Ramp and the Replacement Cycle
Turnover costs recruitment, training and lost output, all landing entirely on the provider. Clients feel the quality effect; you pay for it.
Ramp is the part that gets underpriced. New hires occupy a seat, draw a salary and produce very little billable output for weeks, so where a program grows or replaces staff continuously, some proportion of your headcount sits permanently in ramp. A program running high turnover is therefore paying twice: once for the departure, and again for the extended period before the replacement reaches full output. Report ramping headcount separately, or an account that is investing will read as an account that is failing.
One methodological warning worth carrying into any comparison: providers frequently report voluntary attrition only, which understates actual turnover considerably. When a client asks for your figure, or you assess an acquisition target, establish whether involuntary departures are included before treating two numbers as comparable. TabaTalk’s guide to reducing turnover covers the drivers.
Idle Capacity in Language-Specific Queues
Serving Arabic, English, Hindi, Urdu and Tagalog speakers means maintaining coverage in each, and language queues carry thinner volume than the general one. Those agents sit available more often, which depresses utilization through deliberate design rather than through performance.
Price multilingual coverage as the distinct cost it is. Where it sits inside a blended rate built on single-language assumptions, the shortfall lands on you every month of the term.
Cross-skilling helps at the edges, though less than people hope. An agent fluent in two of your languages can cover both queues, which raises their availability, but the scheduling complexity rises with each additional skill combination and the pool of genuinely bilingual candidates is smaller and more expensive than the general one.
Penalties Triggered by Definitional Gaps
This one genuinely irritates me, because it is entirely preventable.
Nearly every committed measure contains an unwritten choice: the short-abandon exclusion window, whether abandoned contacts count in speed of answer, whether automated acknowledgments stop a response clock, whether resolution is judged across channels. COPC consultants have reported clients using abandon exclusion windows of 5, 10, 20 or even 40 seconds (COPC).
Where the definition is undocumented, the client’s interpretation tends to prevail during a difficult quarter. Notice also that platform defaults usually produce the flattering number, which feels convenient until an analyst on the client side works out why, at which point a definitional difference starts looking like something less innocent. Our guide to BPO call center KPIs sets out the definitions annex worth attaching to every agreement.
Forecast Error You Did Not Cause
Where the client owns volume forecasting and you own staffing, their planning error becomes your cost. Overtime, agency cover, missed service levels and penalties all follow from numbers somebody else supplied.
Reasonable agreements suspend penalties when actual volume exceeds forecast beyond an agreed tolerance, commonly set around ten or fifteen percent. Getting that clause costs a conversation during contracting and saves considerably more later.
Measure the accuracy at interval level rather than daily. A forecast landing within a couple of percent across a day can still be badly wrong mid-morning and again after lunch, and those are precisely the intervals where accessibility targets get missed and overtime gets authorized.
Scope Creep Between Statements of Work
Small additions accumulate: an extra report, a new contact type, weekend coverage for a campaign, a system your team now updates manually because an integration was never built.
None of these individually justifies reopening commercial terms, which is exactly why they accumulate. TabaTalk recommends logging every out-of-scope request with an estimated effort figure, even when you absorb it. At renewal, that log is the strongest negotiating document you will have.
Per-Client Technology Overhead
Separate instances per client multiply licensing, patching, administration and configuration drift. Two years in, eleven dedicated deployments are eleven variants, each needing knowledge that transfers poorly between staff.
Shared infrastructure with genuine client-level separation removes most of that overhead, though it carries its own compliance obligations. Our comparison of multi-tenant and single-tenant architecture covers the trade properly.
The drift is the part that surprises people. Each instance gets patched on its own schedule and customized by whoever handled that particular onboarding, so the operational knowledge required stops being transferable between staff. That cost never appears in a business case and is frequently what makes providers regret the model.
Rework, Repeat Contacts and the Pricing Model Inversion
Here is a genuinely counterintuitive one, and it deserves more attention than it gets.
Unresolved issues return as additional contacts. Under per-contact pricing, you get paid for handling the same problem twice, which means poor resolution quietly increases revenue. Under FTE or per-seat pricing, that repeat volume consumes capacity you already sold, so the identical operational failure destroys profit instead.
Same failure, opposite commercial consequence, determined entirely by the pricing model. Worth knowing which incentive your contracts create, because CEB’s research, published in Harvard Business Review, found customers wanted quick resolution rather than exceptional gestures (Harvard Business Review), and per-contact pricing rewards the opposite behavior structurally.
Reporting labor belongs in the same category. Every client wants their pack in their format on their cadence, and where that assembly happens manually across several accounts, an analyst’s week disappears into spreadsheet work producing nothing billable. Configurable per-account reporting turns that week back into capacity, which is an unglamorous saving that shows up reliably and requires no commercial conversation with anyone.
The Outbound Window in Gulf Operations
Outbound programs face a structural constraint imported playbooks miss entirely. UAE Cabinet Resolution No. 56 of 2024 requires prior regulatory approval for telemarketing and restricts calling to a window running from 9:00 am to 6:00 pm (UAE Legislation portal; Morgan Lewis).
Nine hours caps how much of a paid shift can produce live outbound contact. Where a rate card was built on productivity assumptions from a market with evening calling, the shortfall is structural and permanent. Price it in, or decline the target.
Where Pricing Models Sit in All This
Deloitte’s Global Outsourcing Survey reports that outcome-based delivery models have grown in adoption as buyers pursue results-driven relationships, with skilled talent and agility joining cost reduction as key drivers, and 80% of executives planning to maintain or increase third-party investment (Deloitte).
Outcome-based pricing transfers risk toward the provider. That can work well where you control the variables driving the outcome, and badly where you do not. Before accepting one, list what actually determines the result: if the client’s product, policy or forecast dominates, you are pricing exposure to somebody else’s decisions.
| Leak | Where it hides | First diagnostic |
| Shrinkage gap | Pricing model vs actual | Compare assumed against measured, by month |
| Ramp | Blended headcount | Separate ramping agents from tenured |
| Attrition | HR reporting | Check voluntary versus total methodology |
| Language idle time | Blended utilization | Segment by skill group |
| Penalty exposure | Undocumented definitions | Audit every committed measure |
| Forecast error | Client-owned planning | Compare forecast against actual by interval |
| Scope creep | Nowhere | Start a log this week |
| Technology overhead | Per-instance costs | Count variants you maintain |
| Repeat contacts | Volume reporting | Cross-channel repeat rate per issue |
| Reporting labor | Analyst time | Hours per client per month |
A Diagnostic Sequence Worth Running
Work in this order. The early steps cost nothing and usually find the largest gaps.
- Retrieve the pricing model for your three largest accounts. Somebody has the spreadsheet.
- List every assumption in it: shrinkage, handle time, occupancy, ramp weeks, attrition, seat fill.
- Pull the actual figure for each, by month, across the last four quarters.
- Rank the gaps by monetary value, not by percentage. A two-point shrinkage miss on a large account beats a ten-point miss on a small one.
- Audit committed measures against your definitions annex, or write one if it doesn’t exist.
- Segment utilization by skill group before concluding anything about performance.
- Count your reporting hours for one month, honestly.
Most operations find two or three leaks carrying nearly all the value, which is encouraging, since fixing three things is achievable and fixing eleven is not.
One caution about what happens next. Several of these fixes require a client conversation, and the instinct is usually to bundle them into one renewal discussion. I’d separate them instead: raise definitional and forecast-tolerance issues during routine governance, where they read as housekeeping, and save genuine commercial asks for renewal. Arriving at a renewal meeting with six grievances at once tends to produce a defensive client, whereas the same six raised individually across a year mostly get settled without anyone treating them as a dispute.
Frequently Asked Questions
Which leak is usually the largest?
Shrinkage variance, in most programs, because it applies to every hour of every day across the whole account rather than to a subset of contacts. A gap of a few points between the assumed figure and the delivered one compounds across thousands of scheduled hours. Ramp and attrition typically rank next, though their relative size depends heavily on growth rate and tenure profile. Rank by monetary value rather than by percentage before deciding where to spend effort.
How often should pricing assumptions be revisited?
Quarterly for active accounts, and always before renewal. Assumptions built during a bid reflect conditions at that moment, and volume mix, contact complexity, absence patterns and wage levels all move afterward. Where a contract runs multiple years without a review mechanism, the gap widens silently. Building a scheduled comparison into your governance cycle costs an afternoon per quarter and surfaces problems while there is still contractual time to address them.
Should clients see the provider’s internal efficiency figures?
Generally no, and offering them invites questions about staffing decisions that belong to you. Occupancy, utilization and cost per contact are commercial measures rather than service measures. Clients legitimately monitor accessibility, quality and outcomes. Where a client requests efficiency data, ask what decision it would inform, since the answer usually reveals a concern about service that is better addressed directly through the measures already committed.
Does automation improve or damage program economics?
Both, depending on the pricing model. Under per-contact pricing, deflecting simple contacts removes billable volume while leaving a harder residual mix, so revenue falls and cost per remaining contact rises. Under FTE pricing, the same deflection frees capacity you continue to be paid for. Model the commercial effect before deployment, and renegotiate targets alongside it, since handle time and satisfaction both move when easy contacts leave the queue.
What is the fastest leak to fix?
Reporting labor, usually within a month. Manual assembly of client packs consumes analyst time that produces nothing billable, and configurable per-account reporting removes most of it without renegotiating anything. Scope creep logging is nearly as quick to start, though the benefit arrives at renewal rather than immediately. Both cost little to address, which makes them sensible first moves while larger structural gaps get worked through.
How should multilingual coverage be priced?
As a distinct cost rather than inside a blended rate. Language-specific queues carry lower volume, so agents dedicated to them are available more often by design, which permanently depresses efficiency figures built on single-language assumptions. Either price each language stream separately with its own utilization expectation, or set a blended rate that explicitly accounts for the coverage requirement. Clients generally accept this when the reasoning is explained during contracting rather than raised afterward.
What should be logged as out-of-scope work?
Anything not named in the statement of work: additional reports, new contact types, extra system updates, coverage outside agreed hours, ad hoc analysis requests. Record the request, the requester, the date and an estimated effort figure, even where you absorb it willingly. The log rarely triggers an immediate commercial conversation, and that’s fine. Its value appears at renewal, when accumulated evidence supports a rate discussion far better than recollection does.
Can a program be profitable while missing its service targets?
Yes, temporarily, and it is a dangerous position. Understaffing improves short-term cost while breaching accessibility commitments, so profit rises until penalties and renewal risk arrive. The reverse also occurs: programs comfortably exceeding targets are frequently overstaffed, since COPC has noted that over-servicing raises cost and lowers occupancy without necessarily improving satisfaction. Both states indicate a mismatch between commitment and staffing rather than genuine performance.
How do I raise a leak with a client without damaging the relationship?
Bring evidence and a proposal rather than a complaint. Show the assumption, the actual figure, the period covered and the effect, then propose a specific remedy: a revised target, a forecast tolerance clause, a scope adjustment. Clients respond considerably better to a documented gap with a suggested fix than to a general statement that an account is difficult. Raise it during scheduled governance rather than during a month when penalties have already applied.
Want to find out which leaks are draining your programs?
Start with the pricing spreadsheet for your largest account and the actual figures for the same period. That comparison alone usually explains more than a quarter of review meetings have managed.
TabaTalk provides cloud contact center software built for Gulf operations, with client-level account separation, configurable per-account reporting, real-time supervisor visibility and per-account numbering and dialing rules. Contact our sales team to review your BPO setup, or ask us where reporting effort is currently consuming capacity you could be selling.
Based in Sofia, Bulgaria, a city where history meets innovation, Aleksandar is part of the Product Marketing team at Voiso. After completing his studies in Cologne, Germany, he built a strong background with 5 years in telecommunications and 2 years in and contact center solutions (CCaaS). He thrives on closing the gap between customers and product teams, turning insights into strategies that drive clarity, alignment, and growth. He believes great Product Marketing is about creating understanding, momentum, and belief, both inside and outside the company. Outside of work, Aleksandar channels his energy into sports and the mountains, where each hike offers space to organize his thoughts and the quiet needed to return renewed, much like a trail that leads to a higher and clearer view.
