Tens of millions of Americans rely on SNAP and other social services, where call centers are the gateway to interviews, recertification, case updates, and application help.
When call centers are overloaded, legally required steps are practically impossible.
In Missouri, nearly half of SNAP denials were procedural - applicants couldn't complete application steps.
Why current call-center guidance is not enough:
Standard queueing and call-center guidance (Erlang-A) misses the key urgency of social service: callers need to get through, so they'll call back if they can't get through.
We develop a queueing model and performance evaluation framework that includes key SNAP and social service issues:
Callers who can't get through will call back for essential services.
Even successful callers may need to call back for follow-ups.
Same staffing, different dynamics
Abandonment is not always an exit.
Erlang-A treats callers who abandon as leaving the system. In SNAP access, many callers still need the interview, so abandonment and follow-ups can become new demand on the same staffed queue.
Calibration cue from the paper: call center #2 has c=117.8 staff, AHT=52.0 min, observed arrivals 2931.5 calls/day; staffing-curve examples use (θA, θS, θL) = (4, 3, 3).
Same daily demand reaches the same staffed queue.
Takeaway: current staffing guidance that omits endogenous congestion is too optimistic because it treats abandoned calls as gone, not recycled demand.
Displayed quantities are illustrative, time-averaged outcomes for the same normalized demand. The moving circles show flow, not instantaneous rates.
new demandrecycled redials/follow-upsabandonmentable to enroll
92Erlang-A expected enrolled
92SNAP feedback expected enrolled
8Erlang-A expected abandonment
8With redials expected abandonment
0Expected redials + follow-ups
Erlang-A mean effective load
Feedback mean effective load
Performance evaluation framework
Service design changes can improve access directly and through the feedback loop.
The model lets agencies compare staffing and process-design bundles against the same queue outcomes: wait, abandonment, repeat demand, and enrollment.
Capacity lever: staff c
More agents move callers out of wait.
directindirect
More staffed capacity directly increases the rate at which callers connect. The downstream effect is smaller abandonment piles and fewer redials feeding back into tomorrow's queue.