Tiers vs. swarming: what the evidence says

The ticket that's just... waiting

Open your queue and find the one marked "escalated to tier 2." It has been sitting there since Tuesday. The customer has explained the problem twice — once to the tier-1 rep who couldn't fix it, and once again, badly, in the handoff notes. Nobody is technically working it. It's in a holding pattern, waiting on a specialist who is heads-down on three other escalations. This is the tiered support model doing exactly what it was designed to do, and it's the reason a lot of teams are quietly taking it apart.

Tiers and swarming are the two dominant ways to route work a frontline rep can't finish alone. The tiered model is a factory line: tier 1 triages and clears the easy stuff, tier 2 takes the hard stuff, tier 3 is the engineers of last resort. Swarming throws out the ladder. Instead of passing a case up, you pull the right people to the case — a rep, a specialist, sometimes a product engineer — who work it together and then disband. The Consortium for Service Innovation calls the disciplined version "intelligent swarming."

What the swarming data actually shows

The cleanest before-and-after comes from Ping Identity, an identity-software company that replaced tiered escalation with intelligent swarming and measured the first ten weeks.

That isn't a rounding-error improvement; it's the gap between a customer waiting hours and waiting minutes. The gains landed hardest exactly where tiers hurt most — the gnarly, multi-team cases that used to bounce between queues.

And the backlog, the graveyard where escalated tickets go to wait, drained too.

The mechanism is not mysterious. In a tiered model every escalation is a handoff, and every handoff is a queue, a context reload, and a customer repeating themselves. Swarming deletes the handoff.

Where tiers still earn their keep

None of this makes swarming universally correct, and the case studies come from organisations that stuck with it — survivorship you should read with one eyebrow raised. Tiers exist for a reason: they are a load-management tool. A clean tier-1 layer lets you staff predictable, high-volume, low-complexity contacts with a large pool of generalists, shield expensive specialists from routine questions, and keep occupancy in a sane band.

That last point is the one the swarming evangelists skip. Contact-centre staffing lives and dies on occupancy — the share of paid time an agent spends actually handling contacts.

Push past that and you buy a different problem.

Swarming done carelessly is an occupancy nightmare. Pull three people into every hard case and you either have specialists idling between swarms or generalists yanked off their own queues to firefight someone else's. Tiers, for all their friction, make capacity legible: you know who handles what, you can forecast it, and you can staff to it. In a large operation running mostly repeatable problems, that legibility is worth real money.

The tiebreakers: complexity, expectations, churn

So how do you choose? Two questions settle most of it. First, how variable is your work? If most cases are repeatable and self-similar, tiers scale cheaply. If your hard cases are genuinely novel and cross-functional — the Ping Identity profile — the handoff tax of the tiered model exceeds the coordination cost of a swarm.

Second, what will your customers tolerate? The pressure runs one direction. In Zendesk's 2026 research, 88% of customers expect faster responses than they did a year ago, and 85% of CX leaders say customers will abandon a brand over an issue that isn't resolved — even on first contact. A tiered model optimises for throughput and is structurally bad at first-contact resolution on hard problems, because "hard" means "escalate," which means "not resolved today."

There is a people cost buried in the tiers, too. The tier-1 seat is often a churn machine — high volume, low autonomy, a script and a transfer button — and North American call-centre turnover hit 38% in 2022, which SQM ties directly to falling resolution and satisfaction. Swarming's implicit promise is that a junior rep learns by sitting inside the swarm instead of blindly kicking cases upstairs. That is a better job and a faster ramp.

The honest answer is that most mature teams run a hybrid: a tier-1 front door for volume, and swarming in place of a tier-2/tier-3 ladder for the cases that would otherwise rot in a queue. Kill the escalation chain; keep the triage layer. If you're redrawing these lines, the org chart after AI is the natural next read, and it's worth pressure-testing any resolution-time claim — including these — against your own numbers in the benchmarks before you reorganise around it.