Stand at the dispatch board of any mid-sized last-mile delivery operation at 6:15 a.m. and you'll see the same picture in market after market. Eighty-three drivers are loading vans for the day's routes. The first-attempt delivery success (FADS) dashboard for yesterday is sitting at 87 percent — meaning roughly one in eight stops failed and is now back in the queue for today, taking a route slot away from a new delivery. Seven drivers are running 15 to 20 percent over the engineered hours on their routes because of mis-loaded vans or addresses the routing engine couldn't find. The customer service queue is showing 340 'where is my package' tickets from yesterday. The operations manager is in a meeting trying to explain to the regional VP why the cost per stop went up 11 percent year-over-year. Nobody is doing anything wrong. The system is just — fragile.
Last-mile delivery is one of the most operationally consequential and economically asymmetric processes in modern supply chain. The reason is the math: roughly 50 to 60 percent of total parcel delivery cost lives in the last mile, and the unit economics of every failed first attempt cascade through the rest of the operation. A failed first attempt costs the redelivery (roughly $7 to $12 per attempt fully loaded), the customer service contact (roughly $4 to $8 per ticket), the customer experience hit, and a meaningful share of the time the package becomes a return. Get FADS right and you simultaneously lift first-attempt success from 87 percent to 96 percent, cut cost per stop by 12 to 18 percent, reduce returns driven by delivery friction by 25 to 35 percent, and protect the customer experience scores that drive repeat purchases. The published case studies from the National Private Truck Council and the Postal Service's research on consumer delivery preferences consistently document these gains across structured improvement programs.
This article is the playbook. We'll walk through what last-mile underperformance really costs, how to size the prize before you commit a project team, the structured DMAIC approach that delivers durable improvement (and why route optimization software alone rarely does), the human factors that decide whether the gain holds, and the mistakes that quietly destroy the math after the project closes. By the end you'll have a clear view of what a credible last-mile initiative looks like — and a way to estimate the dollars before you commit a budget.
Why first-attempt success is the most important metric in last-mile
Most last-mile operators track three numbers: first-attempt delivery success (FADS), cost per stop, and on-time delivery against the customer-promised window. The benchmarks are well-published. Top-quartile residential parcel FADS runs 95 to 97 percent; the industry median sits at 86 to 91 percent. Top-quartile cost per stop runs $4.20 to $5.80 fully loaded; the median runs $6.80 to $9.40. Top-quartile on-time-window compliance runs 94 to 97 percent; the median runs 86 to 92. The gap between top-quartile and median is roughly the ROI of a structured Lean Six Sigma program — typically $4 to $12 million per market on a mid-sized regional operation.
Here's the math. A regional last-mile operator running 1.8 million stops per year at an average cost per stop of $7.10 and a FADS rate of 88 percent is spending $12.8 million per year on direct delivery cost. The 12 percent first-attempt failure rate generates roughly 216,000 redelivery attempts per year at a fully-loaded redelivery cost of roughly $9 per attempt — that's $1.94 million in pure redelivery cost. Add the customer service contact volume (typically 35 to 45 percent of failed first attempts generate a contact, at $6 per contact average — roughly $560,000 annually). Add the returns driven by delivery friction (the data is unambiguous: parcels that fail first attempt return at 1.4 to 1.8x the rate of successful first-attempts, costing reverse-logistics handling of $8 to $14 per return). The total annualized cost of the FADS gap on this single market lands in the $3 to $5 million range — recoverable through a structured project that lifts FADS to 95 percent.
The cost recovery is only half the story. The bigger strategic effect comes from what reliable delivery does for the e-commerce relationship. A retailer whose 3PL delivers at 96 percent first-attempt success has a measurably lower NPS impact from delivery, lower customer service contact volume, lower return rates, and higher repeat-purchase frequency. The downstream effect of a last-mile FADS project isn't just delivery cost — it's the e-commerce shipper retention and the new business wins that depend on operational reliability. We've seen 3PL operators win net-new contracts worth $20 to $50 million in annual revenue purely on the strength of a documented FADS turnaround.
The methodology: DMAIC for last-mile delivery
DMAIC for last-mile has a structural twist that distinguishes it from warehouse or transportation projects: the variability is dominated by the consumer — addresses, gate codes, time-of-day availability, signature requirements, weather access — and the consumer interaction is mediated by a notification and self-service infrastructure that's almost always under-leveraged. A FADS project that focuses only on the driver and the route produces 30 to 40 percent of the available gain. A FADS project that includes the consumer notification and pre-delivery confirmation flow produces the full 8 to 10 point improvement. The methodology has to scope both halves of the system from week one.
Define: scope by market and by parcel type
Pick one market and one parcel type. The right starting market is your highest-volume metro with the lowest current FADS. The right starting parcel type is residential ground parcels under 25 pounds — the highest-volume, highest-leverage segment in most operations. Define the scope as 'FADS, cost per stop, and on-time-window compliance for this market and this parcel type.' Don't try to fix all markets and all parcel types at once.
The Define charter names the market, the parcel type, the baseline (FADS percentage with the variance, plus cost per stop and window compliance), the target (typically 6 to 9 points of FADS lift, 12 to 18 percent cost-per-stop reduction), the dollar value, the timeline (120 to 180 days for a Green Belt last-mile project), and the sponsor (typically the regional operations VP or the head of last-mile).
Measure: walk the day from load-out to proof of delivery
Ride along on 20 to 30 routes across multiple drivers, days, and conditions. Timestamp every category: drive time between stops, dwell time at each stop (curbside, walk to door, wait at door, signature capture, return to van), failed-attempt time, exception handling, and break time. Pull six months of failed-attempt data with the actual disposition codes. Validate the codes — most operations have a 'no one home' code that gets used as a catch-all and hides 40 to 60 percent of the actual failure modes. Re-walk the data with drivers to reclassify the actual failure modes.
Most operations discover that the FADS failures cluster in five categories: (1) consumer not home and no acceptable safe-drop location, (2) wrong or incomplete address, (3) gate or building access denied (especially apartments and gated communities), (4) signature required and consumer not present, and (5) driver-side errors (mis-load, route deviation, time-of-day mismatch). Each has a different intervention. Treating them as 'consumer not home' is what produces the failed FADS campaigns most operators have run for a decade.
Analyze: separate consumer-side, driver-side, and system-side failures
Pareto the failed attempts by reclassified disposition code. Separate the failures into consumer-side (not home, signature required, access), driver-side (mis-load, mis-route, time mismatch), and system-side (address quality, routing engine errors, ETA inaccuracy). Most operations discover that the consumer-side failures account for 50 to 65 percent of FADS misses — and that the consumer-side interventions (notification, ETA accuracy, safe-drop authorization, alternate-location flexibility) are the highest-leverage and lowest-cost fixes available. The driver-side failures are roughly 20 to 30 percent and respond to load-out discipline and route-engine tuning. The system-side failures are 10 to 20 percent and respond to address quality programs and ETA model improvement.
Improve: notify earlier, authorize broader, route smarter
The Improve phase has three reliable patterns. First: redesign the consumer notification flow to give the consumer a precise ETA window the morning of delivery (not a 12-hour window the day before), and prompt for safe-drop authorization, alternate-location preference, and access codes inside the same notification. Operations that move from morning-of 12-hour windows to morning-of 2-hour windows with embedded preference capture lift FADS by 4 to 6 points on consumer-side failures alone. Second: tighten the load-out discipline so drivers leave the depot with their van loaded in stop sequence, with exception alerts for mis-loads. Third: tune the routing engine so the suggested time-of-day for each stop reflects historical consumer availability for that address — apartment buildings and houses have very different optimal delivery windows, and the engine should learn that.
Control: hold FADS through peak and through driver turnover
Last-mile control plans face two unique stresses: peak season (volume doubles or triples in November and December for many operators) and driver turnover (industry annual turnover runs 35 to 70 percent in many markets). The control plan names the daily metrics (FADS by route, cost per stop, exception code distribution), the owner (the dispatch supervisor and the route auditor), the cadence (a 5-minute morning huddle and a 5-minute afternoon huddle reviewing the day's FADS performance), and the escalation. The plan also has to include a structured driver onboarding that bakes the new standard work into the first week of any new driver — without that, the gain decays as drivers turn over.
What a real last-mile project looks like, week by week
Weeks 1–3: Define and charter
Regional VP sponsors. Project leader is typically a market operations manager. Team includes drivers from each tenure tier, a dispatcher, a router, a customer service partner, an address-quality analyst, an IT partner for the notification flow, and a finance partner.
Weeks 4–8: Measure
Ride along on 20–30 routes. Reclassify six months of failed-attempt data. Pull customer service contact volume tied to delivery. Lock the baseline by route and disposition code.
Weeks 9–12: Analyze
Pareto failures by reclassified code. Identify the top three to five interventions across consumer-side, driver-side, and system-side. Validate against data. Sign the Analyze tollgate.
Weeks 13–18: Improve
Run three Kaizen sessions: notification redesign, load-out discipline, route engine tuning. Pilot interventions on two to four routes for three to four weeks. Refine. Roll across the market.
Weeks 19–24: Control
Run the new operation across the full market for six weeks. Hold the daily huddles. Validate impact with finance. Write the control plan and the driver onboarding update. Hand off with named accountability.
The mistakes that destroy the math
Mistake 1: Treating FADS as a driver performance problem
Most operators reflexively attribute FADS misses to driver effort. The data, when honestly reclassified, almost always shows 50–65 percent are consumer-side and another 10–20 percent are system-side. Driver-side is real but it's not the dominant cause.
Mistake 2: Skipping the disposition code reclassification
If 'consumer not home' is your top failure code, you don't actually know your top failure mode. Reclassify the codes with drivers in the room. The reclassified data is unrecognizable from the original — and it's what the entire improvement program rests on.
Mistake 3: Investing in route optimization without redesigning the consumer notification
Route optimization software produces a 2–4 percent FADS lift on consumer-side failures. Notification redesign produces a 4–6 point lift on the same failures, at a fraction of the cost. Sequence the notification work first.
Mistake 4: Closing the project before peak
The real test of a FADS project is whether the gain holds at 2–3x daily volume. Hold the project open through one full peak. Hand off only after the new operation has held through the heaviest two-week stretch.
Mistake 5: Not building the driver onboarding update
A FADS gain that requires the experienced drivers to maintain it will decay through the natural turnover cycle. Bake the new standard work into the first-week driver onboarding. The gain has to survive a 50 percent driver workforce refresh inside 18 months.
How to size the prize for your last-mile operation
Pull your last 12 months of stops by market, your FADS rate by market, your cost per stop, your customer service contact volume tied to delivery, and your return rate. Calculate the redelivery cost recovery (current redelivery volume × redelivery cost × estimated 60 percent reduction). Add the customer service contact reduction (current delivery contacts × $6 average × 35 percent reduction). Add the return reduction (current delivery-driven returns × handling cost × 30 percent reduction). Discount by 50 percent for realism. If the discounted number is more than $2 million on one market, you have a project worth chartering. Most last-mile operations running more than 1 million annual stops in a market are sitting on $3 to $7 million of opportunity per market.
If you'd like to walk through the math on your specific last-mile operation — confidentially, with a Master Black Belt who has run these projects in parcel, e-commerce 3PL, and retail home-delivery operations — book a free 30-minute consultation. We'll size the prize and tell you honestly whether a Lean Six Sigma project is the right next move.
