Control tower – Lastmily

Lastmily · Platform · Control tower

Control tower

One event log. Every block. Every parcel.

The control tower is not a dashboard bolted on at the end. It is the read side of the same record every block on the line writes to – so the live map, the ETA, the KPI and the alert are all looking at the same object, at the same moment.

Book a walkthrough The whole platform

Why it is one layer, not one screen

Most operations have three versions of the truth.

The dispatcher’s screen, the shipper’s report and the driver’s app usually run on three databases, refreshed at three different times. By lunchtime they disagree, and the argument about which one is right costs more than the problem they were describing.

One record

Every block on the line – commercial, first mile, hub, middle mile, last mile, networks – writes to the same shipment record. The control tower reads it. There is no export, no nightly job, no second warehouse to reconcile.

One clock

A scan on the hub floor, a status change in the driver app and a recipient redirecting from their phone all land on the same timeline, in the order they actually happened.

One definition

A failed delivery means the same thing in the live map, in the KPI, and in the report you send your shipper at the end of the month. Definitions live with the platform, not in each person’s spreadsheet.

The control tower: live map, routes, driver rail, stop detail and the KPI bar across the top

Execution & live monitoring

The day as it is happening.

The live map is the whole operation on one canvas: every round drawn, every driver where they actually are, every stop in its current state. The driver rail down the side is the fleet as a list, so a dispatcher can work either way round – from the map to the person, or from the person to the map.

Click a stop and you get its whole history: when it was routed, who it was assigned to, what the recipient did with their notification, what happened at the door. Not a status field – the sequence of events that produced it.

Lastmily · Control tower

The reports

The reports that prove it.

A delivery platform earns its place on the month-end call, not on the live map. Every report below is the same event log, filtered – so the number in the report and the number on the screen are the same number, and a disputed one can be opened down to the scan that made it.

Operating

For the person running tomorrow morning.

Round reportOne round, stop by stop: planned against driven, the sequence, the deviations, the PoD on every completed stop and the reason code on every one that failed.

Driver reportPer driver, per day or per period: stops, service time against predicted, adherence, exceptions, self-debrief. The same view the driver can be shown, so the conversation is about a record, not a recollection.

Hub reportInbound, weighed and measured, exported, loaded, dispatched. What arrived against what was expected, and what is still on the floor at closing.

Line-haul & closingsTransport sessions, containers, seals and custody – what left in which container, who sealed it and who broke it.

Exceptions registerEvery stop that did not go to plan, with the reason the driver selected at the door and what happened to the parcel next.

Commercial

For the person who has to send the invoice.

Invoice backupThe line-by-line evidence behind a charge: the shipment, the agreement, the price list that priced it and the events that make it billable. A disputed invoice is opened down to the scan.

COD settlementCollected, by driver, by round, by method – cash, Soft PoS, NFC – against what has been handed over and what is still outstanding.

Cost centre reportCost allocated where it was actually incurred, by centre, lane, vehicle and round – rather than averaged across a month.

Weight & volume varianceDeclared against measured at the gate. This is the report that finds weight-revenue leakage, and it is a report because it has to be arguable.

Shipper statementThe same numbers, in the shipper’s own view, in the shipper app – so they do not have to ask you for them.

Promise

For the person who sold the delivery.

SLA & window complianceDeliveries inside the window promised at checkout, by shipper, area and service – not inside the window the route happened to produce.

First attempt & failure analysisFirst-attempt success and failed deliveries broken down by reason, area, hour and driver, so the fix has an address.

Notification logWhat was sent, on which channel, when, and whether it arrived – before anyone asks why the recipient did not know.

Rating, NPS & reviewsCollected at the door, attributable to the round, the driver and the area, and reportable back to the shipper whose brand was on the parcel.

Self-service activityRedirects, reschedules and PUDO switches the recipient made themselves – the volume that never became a phone call.

Network

For the person deciding who carries the next one.

Carrier scorecardPer carrier, per lane: SLA met, failed deliveries, cost, recipient rating. It is the same data the orchestration rule reads, so the report and the routing decision do not disagree.

PUDO & locker utilisationVolume and occupancy per point, and how much door volume moved out of home delivery – the number that justifies the network.

Lane cost reportCost per lane and per load, against the alternatives the engine considered at the time.

Partner performanceFor a network you do not own outright: the same measures, per partner, on your own event log rather than on theirs.

CO₂ reportFrom distance actually driven and the vehicle profile – the same source as the routing decision that produced it.

One source, four audiences

Operations, finance, the shipper and the network are looking at different reports built from one log. That is why the ops number and the invoiced number do not drift – there is no second system for either of them to drift towards.

Scheduled, exported or queried

Any of them on a schedule to an inbox, exported to XLS or CSV, or pulled over the API into the warehouse you already report from. The platform does not insist on being where you read your numbers.

Traceable, and therefore arguable

Every figure resolves back to the events underneath it: the scan, the seal, the gate measurement, the PoD, the notification. A report nobody can open is a report nobody believes.

The live map tells you about this afternoon. The reports are what you take to the month-end call, the shipper review and the carrier negotiation.

Lastmily · Reporting

Reporting & KPIs

What gets measured, and where the number comes from.

Every KPI below is computed from the event log, not typed in. Each one names the events it is built from, so a disputed number can be traced to the scans that made it.

Execution

First-attempt success rateDeliveries completed on the first attempt, over deliveries attempted. Built from the PoD event and the attempt events before it. The single number most operations are actually judged on.

Failed deliveriesAttempts that ended without a handover, split by reason code the driver selected at the door – not by a category someone assigned afterwards.

Route adherenceHow much of the planned sequence the driver actually drove. B2C typically sits around 35% – an industry baseline, not a Lastmily result – which is why it is a KPI here and not a footnote. A plan the driver ignores is not a saving.

Stops per hour · per driver · per roundThroughput at three levels, so an outlier can be traced from the fleet to the person to the round that caused it.

Service timeDoor-to-door time per stop, against the service time the engine predicted. The gap is the model’s error, and it is visible.

Promise

Time windows metDeliveries inside the window that was promised at checkout, not inside the window the route happened to produce.

ETA accuracyPredicted arrival against actual arrival, per stop. Tracked as a distribution, because an average ETA error of zero can hide a large error in each direction.

Notification delivery & openWhat was sent, on which channel, and whether it arrived – SMS, Viber, email – before anyone asks why the recipient did not know.

Self-service change rateHow many recipients changed the delivery themselves rather than calling. Together with proactive notifications – not on its own – this is what moves inbound call volume.

Rating & NPS at the doorCollected at the point of delivery, attributable to the round, the driver and the area.

Cost & network

Cost per deliveryRolled up from the round, the vehicle and the lane – not allocated evenly across a month.

Vehicles on the road · capacity usedPlanned against actual. Fleet capacity that never left the depot is visible the same day, not at the end of the quarter.

Billable weight vs measured weightWhat was declared against what the gate measured. This is where weight-revenue leakage shows up – Lastmily estimate: up to 20% of billable weight bypasses the sorter.

Carrier & partner performancePer carrier, per lane: SLA met, failed deliveries, cost. Feeds the orchestration rule that picks who carries the next one.

CO₂ per deliveryFrom distance actually driven and the vehicle profile – the same source as the routing decision that produced it.

Bring the number you are judged on.

We will show you where it lives in the event log, how it is computed, and what moves it – before you commit to anything.

Book a walkthrough The whole platform

Lastmily · Reporting

Anomaly detection

The report tells you what happened. The alert tells you now.

A KPI is a summary of a day that is already over. Anomaly detection is the same event log read forwards: the platform compares what is happening against what it predicted would happen, and raises the gap while there is still time to act on it.

A baseline, not a threshold

A fixed threshold fires every Monday and never on the day it matters. The models already predict service time, preferable windows and area-point matching. An anomaly is a departure from that prediction, for this round, this driver, this area, this day of the week.

With the reason attached

An alert arrives with what it is based on – the events, the prediction it departed from, and the size of the gap. You can disagree with the reason rather than only with the number.

Acted on where policy allows

Inside the boundary you set, Mily – the agent that acts on the record – re-sequences, re-assigns and notifies. Outside it, it asks. Everything it does lands in the same event log as everything a person does.

What it watches for

Round running lateAccumulated delay against the predicted service time for the stops already done, projected onto the stops remaining. Raised while the round can still be changed, not when it is finished.

A driver off the planRoute adherence collapsing mid-round. Sometimes the driver is right and the plan was wrong – that is the signal the territory model reads. Sometimes it is a problem. Either way it should not be a surprise at debrief.

Time window at riskA promised window that the current position and remaining sequence will not meet – early enough to re-sequence, re-assign, or tell the recipient before they notice.

Failed-delivery clusterThe same reason code repeating in one area, one building or one round. One failure is noise; a cluster at the same postcode in the same hour is an address problem, an access problem or a driver who needs help.

Weight & volumetric mismatchDeclared against measured at the gate, flagged per shipper and per lane. This is the revenue-protection model doing its job in the open.

Notification not deliveredA message that failed on its channel, before the recipient is blamed for not knowing.

COD not reconciledCash or card taken at the door that has not been settled back against the shipment.

Carrier SLA driftA partner whose performance on a lane has moved away from what the orchestration rule assumed when it chose them.

An alert that cannot be traced to the events that caused it is an opinion. Every one of these can be opened down to the scan.

Lastmily · Anomalies

Reporting

The same record, read at a different speed.

The live map answers “what is happening right now”. The dashboard answers “what has been happening”. Both read the same event log, so the number in the report is the number that was on the screen – not a version of it that was exported, transformed and rounded somewhere in between.

The activity feed is the operational narrative: what changed, who changed it, what the models recommended, and which recommendations were accepted. When Mily acts inside your policy, it appears here in the same list as a person.

Reporting: KPI dashboard and the activity feed, with Mily in it

Slice it the way the argument is

By shipper, by area, by round, by driver, by carrier, by lane, by day of week. The dispute is usually about a slice, so the slice is the unit.

Compare against the prediction

Every operational number has a predicted twin – service time, window, capacity. Showing the pair is more useful than showing either alone.

Export or read it live

The reports your shipper wants monthly and the view your dispatcher wants hourly are the same query at two cadences.

Lastmily · Reporting

What the numbers are worth

Every figure with the label it was measured under.

Two come off comparable implementations, two are modelled, two are a Lastmily estimate and two are reported by the shipper with no control group. The tag under each one says which, because a figure without its provenance is a slogan. None of them is a forecast for your operation.

84 → 90%

first-attempt success rate

modelled on a 7,000-stop day

−35%

inbound calls, with proactive notifications and self-service

comparable implementations

−19%

failed deliveries

comparable implementations; −9% at Public

6-10%

of fleet capacity – adherence, prediction, first attempt

modelled · the three effects overlap; do not add them up

up to 80%

of weight-revenue leakage recovered

Lastmily estimate · up to 20% of billable weight bypasses the sorter

10-15%

of line-haul transportation cost, from lane and load optimisation

Lastmily estimate

+8 pts

brand NPS for the shipper

Public case study; shipper-reported, no control group

+19%

second purchase from new customers

Public case study; shipper-reported, no control group

None of these is a promise. Each one is a number we can baseline on your data before you commit to anything.

Lastmily · The platform in numbers

Keep reading

Routing AIThe engine the control tower is watching. Last mileThe block where most of the events are written. Hub & middle mileWhere weight and custody are measured. DevelopersReading the event log over the API.

Lastmily · Control tower

The close

Pick the number we should be measured on.

Then let us measure it before you commit to anything.

Book a walkthrough See the whole platform