Here’s a scenario we sometimes encounter in our consulting projects.
The business hands us a number: “Keep SL at 80/20.” We build the model, staff it, and hit it. Then, in the weekly review, someone asks why ASA is creeping up. We check. ASA is sitting at 30 seconds, against an unstated target of 20. Nobody mentioned an ASA target. “But shouldn’t 80/20 mean ASA is under 20 seconds? If 80% of calls are handled within 20 seconds, why is the average that high?”
This sounds intuitive. But it doesn’t always hold true.
ASA, Service Level, and Answer Rate sound like they’re all measuring how fast you’re answering calls. They’re not. They are not interchangeable. Each one measures something different, responds differently to staffing changes, and will lead you to a different headcount recommendation. Sometimes by a meaningful margin.
Let’s take a closer look.
The Three Metrics, Defined Precisely
Before anything else, let’s define them precisely to avoid confusion.
Service Level (SL)
The percentage of contacts answered within a defined threshold time.
Service Level = (Calls answered within X seconds) / (Total calls offered) x 100
Example: 850 out of 1,000 calls answered within 20 seconds = SL of 85%.
The denominator is total calls offered, including abandoned calls. This is the standard Erlang/industry definition. Some platforms calculate SL using only calls that reached the queue (excluding pre-queue abandons), so always confirm what your ACD is counting before benchmarking.
Average Speed of Answer (ASA)
The average wait time across all calls that were answered.
ASA = Total wait time for answered calls / Number of calls answered
Example: 500 calls answered, total cumulative wait of 12,500 seconds = ASA of 25 seconds.
ASA is an average that can be heavily influenced by outliers. A small number of callers who waited a very long time will pull it up significantly, even if most callers were answered quickly. Two queues can have identical ASA values with completely different distributions of wait times. And in a high-abandon queue, ASA is essentially lying to you: callers who gave up after three minutes aren’t in the denominator.
Answer Rate
The percentage of callers who were successfully answered before hanging up. Some contact centers call it access rate or performance rate. Answer Rate = 1 minus Abandonment Rate.
Answer Rate = Calls answered / Total calls offered x 100
Example: 1,000 calls offered, 60 abandoned = Abandonment Rate of 6%, Answer Rate of 94%.
Answer Rate, like SL, is not a pure staffing metric. It is also a customer behaviour metric. It is shaped by staffing (longer waits mean more abandonments) but also by caller patience, which varies by contact reason, channel availability, and time of day. You cannot control the answer rate directly. You can only influence it by changing wait times. Framing it as Answer Rate rather than Abandonment Rate matters in stakeholder conversations: “94% of customers reached an agent” lands very differently than “6% abandoned.”
Same Queue, Three Different Verdicts
Let’s go back to the example that trips people up. Take one 30-minute interval: 10 calls, 5-minute AHT, and 5-minute patience. Same staffing, three different stories across SL, ASA, and Answer Rate.
| Metric | Result | vs. Target |
|---|---|---|
| Service Level (20 sec threshold) | ~80% | ✅ Hit |
| ASA | 32 sec | ❌ Miss (target: 20 sec) |
| Answer Rate | ~93% | ❌ Miss (target: 95%) |
Same staffing. Three different answers. If you were only watching SL, you would call it a good interval. ASA and Answer Rate tell a different story.
When to Use Each One
Service Level is your planning metric. The Erlang formula is built around it. It captures the distribution of wait times, not just the average, which makes it far more reliable for staffing models. If you’re doing capacity planning, start here. Using our Erlang X formulas also ensures you account for the role of caller patience.
ASA is your communication metric. “Average wait was 30 seconds” lands with a non-technical audience in a way that “80% in 20 seconds” doesn’t. Use it for trend-spotting and stakeholder updates. Just never use it as your primary planning input in a queue with meaningful abandonment.
Answer Rate is your customer impact metric. If your business thinks in terms of lost contacts and lost revenue, Answer Rate (“94% of customers reached an agent”) is the number that connects staffing directly to outcomes.
The Staffing Curve That Changes Everything
Although these three metrics are not interchangeable, they are related. Here’s what happens across a staffing curve: 50 calls, 30-minute interval, AHT 5 minutes, patience 5 minutes.
All results are calculated via our Erlang X calculator and rounded up to the nearest integer.
| Agents | SL (20 sec) | ASA | Answer Rate |
|---|---|---|---|
| 8 | 48% | 60 sec | 84% |
| 9 | 62% | 38 sec | 90% |
| 10 | 74% | 23 sec | 94% |
| 11 | 83% | 13 sec | 96% |
| 12 | 90% | 7 sec | 98% |
Two things to notice.
First: the jump from 8 to 9 agents gains 14 SL points. From 11 to 12, you gain 7. The Erlang X curve flattens fast, which is exactly why overstaffing above roughly 85% SL is expensive and why every resourcing conversation should reference this curve, not just a target number.
Second: at 11 agents, SL hits 83% (above an 80/20 target) and Answer Rate is 95%. If you’re managing both targets, you’re fine. But if the Answer Rate target were set at 98%, you’d need 12 agents. That’s one more than the SL target requires, with real headcount cost implications.
Same SL, Very Different Reality: A Tale of Two Volumes
Here’s where it gets really interesting, and where over-reliance on SL alone can quietly mislead you.
Take two queues. Same AHT (5 minutes). Same 30-minute interval. Both are hitting roughly 80% SL in 20 seconds. The only difference is volume.
| Scenario A: Low Volume | Scenario B: High Volume | |
|---|---|---|
| Calls offered | 80 | 300 |
| AHT | 5 min | 5 min |
| Patience | 5 min | 5 min |
| Agents | 16 | 53 |
| SL (80/20) | ~80% ✅ | ~80% ✅ |
| ASA | 14 sec | 10 sec |
| Abandonment Rate | 4% | 3% |
| Answer Rate | 96% | 97% |
| Occupancy | 80% | 91% |
Both queues pass the SL check. But look at what SL is hiding.
ASA diverges. The low-volume queue has a noticeably longer ASA (14s vs 10s) despite a near-identical SL. With a smaller agent pool, callers who do queue wait longer on average since there’s less capacity to absorb them quickly.
Answer Rate diverges. The high-volume queue delivers a better Answer Rate (97% vs 96%), even though both hover around the same SL. This is the Erlang pooling effect: a larger agent group handles demand more efficiently at scale, reducing the proportion of callers who hang up.
Occupancy diverges significantly. 80% vs 91%. The high-volume queue is running much hotter. Sustainable in short bursts, but across a full day, it’s a burnout and attrition risk that SL alone will never flag.
The takeaway: SL tells you whether you hit the threshold. It doesn’t tell you what the experience was like for callers who didn’t make it, or how hard your agents are working to deliver it. In a low-volume operation, good SL can coexist with relatively high abandonment and slow average speeds. In a high-volume operation, the same SL target is being delivered under very different and much more demanding conditions.
This is exactly why WFM analysts should never report SL in isolation, and why ASA and Answer Rate exist as complementary lenses on the same queue.
What To Do When Targets Conflict
You’re running with 80/20 SL and a 95% Answer Rate target. On a good day, both are satisfied at the same staffing level. On a rough day, high volume, short interval, spike in calls, they diverge.
Ask the business one question: if you had to choose on a bad day, which one do we defend?
- SL wins: staff to SL, track Answer Rate as a secondary signal.
- Answer Rate wins: staff to Answer Rate, SL becomes context.
- Neither is negotiable: that’s a capacity conversation, not a modelling problem. Escalate it clearly.
Most of the time, the business hasn’t actually thought about this. Asking the question is itself valuable.
The One-Line Summary
Service Level for planning. ASA for communication. Answer Rate for business impact.
Know which one you’re optimising, and make sure the business knows too. The model is only as good as the target it’s built around. In CCmath’s solutions, all three metrics, including occupancy, are built into our modules, because the full picture only emerges when you track them together.
Which metric does your operation live and die by? And does the rest of the business know that’s what you’re measuring?
That gap, more often than not, is where the real problems hide. Want to go deeper? Contact us or explore our Erlang calculators.