How much a missed call really costs (2026)

1. Cost per missed call, by trade (2026)

Cost per missed call is calculated as booking_rate × avg_ticket. Annual mid-size column assumes ~500 calls/month and a 35% miss rate.

Sources: rate-sheet audits 60+ shops Q1 2026 (locksmith, garage door, towing, plumbing, HVAC); IBISWorld trade reports for restoration, pest control, appliance, electrician ticket medians; booking rates from platform aggregates where coverage exists, estimates elsewhere.

3. The four hidden costs (beyond the headline number)

The dollar amount of one job is the visible cost. There are four more, each meaningful.

5. The math (in plain notation)

Nothing here is hidden. The model has six inputs and produces one output.

Inputs:
  calls_per_day        (your inbound call count, daily average)
  work_days_per_month  (typical: 30 for locksmith / 22 for residential)
  miss_pct             (share of inbound calls unanswered, 0.0 - 1.0)
  booking_rate         (share of answered calls that book, 0.0 - 1.0)
  avg_ticket           (your trade and market median, USD)
  seasonal_uplift      (peak-season volume bump, e.g. 0.15 for summer)

Derived:
  monthly_calls        = calls_per_day x work_days_per_month x (1 + seasonal_uplift)
  missed_calls         = monthly_calls x miss_pct
  lost_bookings        = missed_calls x booking_rate
  lost_revenue_month   = lost_bookings x avg_ticket
  lost_revenue_year    = lost_revenue_month x 12

AI receptionist recovery model:
  recovered_monthly    = lost_revenue_month x 0.9     # 90% recovery factor
  net_monthly_gain     = recovered_monthly - 500      # TheKeyBot flat fee
  roi_multiple         = net_monthly_gain / 500

Where the 0.9 recovery factor comes from

Across the customer cohort we have measured at-scale (n > 200 shops, 12+ months of data each), AI-handled calls during the previously-missed window convert to booked jobs at a rate that recovers 88-92% of the previously-lost revenue. We use 0.90 as the conservative point estimate. The remaining 10% breaks down approximately:

• ~4% are spam, robo-dialers, or wrong-number calls (no recovery possible)
• ~3% are out-of-service-area (correctly declined by the AI)
• ~2% are price-shoppers who do not book any provider on that call
• ~1% are calls that fail mid-stream due to caller-side audio quality

Different vendors quote different recovery factors. Smith.ai, Ruby Receptionist, and AnswerConnect publish recovery numbers in the 60-75% range — accurate for their product class (message-takers and human answering services). AI booking agents that quote and book mid-call get the higher recovery because they close the loop on the same call.

7. What this research deliberately does not cover

For honesty: here are the holes we know about.

• Geographic granularity below metro level. Our platform aggregates have signal at the metro-area resolution. Shop-level variability inside a metro can be substantial (Houston west side vs Houston east side, Brooklyn vs Queens). We do not publish at sub-metro granularity to protect tenant identity.
• Per-trade booking-rate variance. The booking rates in the table are medians; actual variance per shop can be ±15 points, driven by pricing position, review score, and service-area definition.
• Multi-call buyer journeys. We model cost per missed call. Some buyers call several shops sequentially; the dollar weight of being the second or third unanswered call is different from the first. We have not modeled this rigorously.
• Cross-channel attribution. A caller who found you via Google, also saw your Yelp listing, and previously got an SMS from your remarketing has a different cost-of-miss than a fresh Google LSA caller. We treat all calls as fresh acquisition, which slightly understates the cost-of-miss for shops with strong organic presence.
• Trade-specific seasonal patterns. Locksmith demand is roughly flat year-round. HVAC, plumbing, garage door, and restoration are seasonal — the cost-of-miss model needs trade-specific seasonality factors which we have not published in this piece.

Related