Service Level Target Selection

From WFM Labs

Service Level Target Selection is the process of choosing a Service Level target that optimizes the economic trade-off between staffing cost and customer value — as opposed to adopting the conventional 80/20 standard by default. The 80/20 target is not optimal for most operations. It's a convention from the 1980s Bell System that persists because it's familiar, not because anyone proved it maximizes value.

Overview

Most contact centers run at 80/20 because that's what "everyone else does." When asked why, the answers are circular: "It's the industry standard." Why is it the standard? "Because everyone uses it." The actual origin: AT&T's internal engineering standards from the early 1980s, designed for a regulated monopoly telephone network where callers had no alternative. The assumption was that 80% of callers answered within 20 seconds provided acceptable service for a general-purpose telephone exchange.

That assumption may not apply to your business in 2026. Your customers have different patience, different value, different alternatives, and different expectations than AT&T's captive subscribers in 1983.

The right approach: derive the SL target from your economics — what does it cost to staff to a given SL, and what revenue/value do you protect by doing so? The optimal target is where total cost (staffing + customer loss) is minimized.

The Economic Optimization Framework

The total cost of operating a contact center at a given service level is the sum of two cost curves:

Staffing cost: increases as SL target becomes more aggressive. Going from 80/20 to 90/20 requires significantly more agents due to Erlang C's nonlinearity. The staffing cost curve is convex — each additional percentage point of SL costs more than the last.

Customer loss cost: decreases as SL target becomes more aggressive. Better service means less abandonment, less churn, higher satisfaction, and protected revenue. This curve is concave — the biggest gains come from moving off very poor SL, with diminishing returns as SL improves.

The sum of these two curves has a minimum — that's your optimal SL target.

Building the Staffing Cost Curve

For a given demand profile (volume and AHT), calculate required staffing at each SL target:

SL Target Required Agents (1,500 calls/hr, 5-min AHT) Annual Staffing Cost ($60K loaded)
70/20 134 $8,040,000
75/20 136 $8,160,000
80/20 139 $8,340,000
85/20 142 $8,520,000
90/20 147 $8,820,000
95/20 155 $9,300,000

The cost difference between 70/20 and 80/20 is $300K (5 agents). Between 80/20 and 90/20, it's $480K (8 agents). Between 90/20 and 95/20, it's another $480K (8 agents). The curve steepens dramatically above 90/20.

Building the Customer Loss Cost Curve

This requires three inputs:

  1. Abandonment rate at each SL level — from Erlang-A with empirical patience data
  2. Revenue impact per abandoned contact — what fraction of abandoners don't call back, and what's the revenue consequence?
  3. Churn impact — what fraction of customers who experience poor service will churn, and what's their CLV?

A worked example for a B2C operation with 2 million customers, $800 average CLV, and $15 average revenue per contact:

SL Target Abandon Rate Annual Abandoned Contacts Non-Callback Rate Lost Revenue Incremental Churn CLV Erosion Total Customer Loss
70/20 7.2% 224,640 25% $843K 0.35% $5,600K $6,443K
75/20 5.1% 159,120 25% $597K 0.22% $3,520K $4,117K
80/20 3.4% 106,080 25% $398K 0.14% $2,240K $2,638K
85/20 2.0% 62,400 25% $234K 0.08% $1,280K $1,514K
90/20 1.0% 31,200 25% $117K 0.04% $640K $757K
95/20 0.3% 9,360 25% $35K 0.01% $160K $195K

Note: These numbers illustrate the method. Your operation's numbers will differ based on your patience distribution, CLV, and churn sensitivity. The method is what matters, not the specific figures.

Finding the Optimum

Add the two cost curves:

SL Target Staffing Cost Customer Loss Cost Total Cost
70/20 $8,040K $6,443K $14,483K
75/20 $8,160K $4,117K $12,277K
80/20 $8,340K $2,638K $10,978K
85/20 $8,520K $1,514K $10,034K
90/20 $8,820K $757K $9,577K
95/20 $9,300K $195K $9,495K

In this example, the optimal target is approximately 90–95/20 — well above the conventional 80/20. The total cost is minimized near 95/20 because the CLV is high enough that customer loss dominates staffing cost.

This is the critical finding: for high-CLV operations, the conventional 80/20 target is too low. The operation is under-investing in service relative to the value at stake.

For a low-CLV operation (e.g., $150 CLV, $3 revenue per contact), the customer loss curve is much flatter, and the optimum shifts to 75/20 or even 70/20.

Beyond 80/20: Choosing the Threshold

The discussion above varies the percentage (70%, 80%, 90%) while holding the threshold at 20 seconds. But the threshold matters too:

Target What It Measures When Appropriate
X/10 Fraction answered within 10 seconds Premium, emergency, high-value B2B
X/20 Fraction answered within 20 seconds Standard inbound voice
X/30 Fraction answered within 30 seconds Moderate-patience queues
X/60 Fraction answered within 1 minute Long-AHT or patient populations
X/120 Fraction answered within 2 minutes Low-priority or cost-optimized

The threshold should reflect customer expectations in your channel and industry:

  • Emergency services: 95/10 or better — callers expect near-instant answer
  • Financial services: 80/20 to 90/15 — high expectations, regulatory pressure
  • Retail/e-commerce: 80/20 to 80/30 — moderate patience
  • Government services: 70/30 to 80/60 — callers have no alternative
  • Internal helpdesk: 80/30 to 80/60 — employees more patient, lower value per contact
  • Email/async channels: measured in hours or days (90% within 4 hours, 95% within 24 hours)

Differentiated SL by Queue Value

Applying the same SL target to every queue is the single most common waste in contact center planning. A retention queue handling customers threatening to cancel ($2,000 CLV at risk per call) and a balance inquiry queue ($0 direct revenue) should not have the same service level.

The Value-Based Planning Model formalizes this:

Queue Customer Value Recommended SL Rationale
Retention/saves Very high (CLV at risk) 90/15 – 95/10 Every extra second of wait increases churn probability
Sales/upsell High (revenue opportunity) 85/20 – 90/15 Abandoned sales call = lost revenue
Technical support Medium-high 80/20 – 85/20 Complexity warrants standard target; callers relatively patient
General inquiry Medium 80/20 – 80/30 Standard service
Balance/account status Low 70/30 – 80/60 Migrating to self-service; callers have IVR alternative
Outbound callback N/A Schedule adherence, not SL Outbound doesn't have inbound SL dynamics

The savings from differentiating are significant. Consider a 500-agent operation with 5 queues. If all run at 80/20, total staffing requirement is X. If the low-value queues run at 70/30 and the high-value queues run at 90/15, total staffing might be X−5 to X+5 depending on the mix — but the allocation of those agents shifts to where they protect the most value.

Regulatory and Contractual Constraints

Some SL targets aren't optional:

  • Emergency services (911/999/112): Regulatory mandates, typically 90/10 or 95/15
  • Financial services: Some jurisdictions mandate maximum wait times for specific transaction types
  • BPO contracts: SLAs define specific SL targets with financial penalties for misses — these are negotiated, not optimized
  • Healthcare: HIPAA doesn't mandate SL, but accreditation bodies may

For regulated queues, the economic optimization still applies — but the constraint floor is the regulatory minimum, and you optimize above it.

Channel-Specific Considerations

Voice, chat, email, and messaging have fundamentally different patience curves and customer expectations:

Voice: SL measured in seconds. Customer tied to the phone — cannot multitask. Patience typically 30–120 seconds for most queues. 80/20 to 80/30 is conventional.

Chat: SL measured in seconds to minutes. Customer can multitask while waiting. Patience longer than voice (60–300 seconds typical). 80/30 to 80/60 appropriate. But concurrent chat handling means the SL calculation is different — see Multi-Channel and Blended Operations.

Email: SL measured in hours. Customer expects asynchronous response. "90% within 4 hours" or "95% within 24 hours" typical. The economics are different — email staffing is driven by throughput and backlog management, not real-time queuing.

Messaging (WhatsApp, SMS, social): Hybrid — somewhere between chat and email. Customer expects relatively quick response but not real-time. "80% first response within 5 minutes" is a reasonable starting point. Patience varies enormously by platform and context.

Applying voice SL conventions (80/20) to email is absurd — it would require massive overstaffing. Applying email conventions ("respond within 4 hours") to voice is unacceptable. Each channel needs its own target derived from channel-specific patience data and customer expectations.

How to Implement Economic SL Target Selection

Step-by-step:

  1. Gather your inputs: offered volume, AHT, and mean caller patience per queue (from ACD data). Customer CLV per queue/segment. Revenue per contact. Loaded agent cost.
  2. Build the staffing cost curve: Use Erlang C or Erlang-A to calculate required agents at each SL from 60/20 to 99/20 (or your relevant threshold). Multiply by loaded cost.
  3. Build the customer loss curve: At each SL, calculate abandonment (from Erlang-A). Estimate non-callback rate and revenue loss. Estimate churn probability and CLV impact. Sum to total customer loss cost.
  4. Add the curves: Find the minimum of total cost. That's your economically optimal SL.
  5. Apply constraints: Regulatory minimums, competitive benchmarks, brand positioning. The target is max(economic optimum, constraint floor).
  6. Differentiate by queue: Run the optimization separately for each queue or queue group. High-value queues will have aggressive targets; low-value queues will have relaxed targets.
  7. Review annually: CLV changes, competitive dynamics shift, channel preferences evolve. The optimal target is not static.

Maturity Model Position

  • Level 1 — Initial (Emerging Operations). SL target inherited or arbitrary. "80/20 because that's what everyone uses." No economic analysis of the target. Same target applied to all queues.
  • Level 2 — Foundational (Traditional WFM Excellence). 80/20 understood as convention. WFM can explain why 80/20 is harder to achieve than 80/30. Some discussion of whether the target is appropriate. All queues still at same SL.
  • Level 3 — Progressive (Breaking the Monolith). SL targets differentiated by queue based on customer value. Economic analysis performed — staffing cost vs customer loss cost — even if simplified. Finance and WFM share a framework for discussing SL trade-offs. Channel-specific targets in place.
  • Level 4 — Advanced (The Ecosystem Emerges). Full economic optimization with Erlang-A, empirical patience data, and CLV integration. SL targets reviewed quarterly. The Value-Based Planning Model drives differentiation. Simulation validates the Erlang-based optimization for complex queues.
  • Level 5 — Pioneering (Enterprise-Wide Intelligence). SL targets adjust dynamically based on real-time demand, staffing availability, and customer-segment value. The economic optimization runs continuously. Channel-agnostic "customer wait tolerance" models replace fixed SL thresholds.

See Also

References

Retrieved from "https://wiki.wfmlabs.org/w/index.php?title=Service_Level_Target_Selection&oldid=2785"