How grocery delivery firms set customer slots by how fast the roads move
Set delivery slots by traffic: measure road speed in each zone through the day, fit fewer drops into slow hours, and warn dispatch when a zone jams.
Part of the guide Workflow automation
To set delivery slots by traffic, measure how fast the roads move in each delivery zone at every half hour you deliver, for a few weeks, then size each slot to the speed you saw: fewer drops per van in the hours when a zone crawls, more when it flows. Keep measuring once the slots are live, so dispatch hears when a zone jams on the day and can stretch the next windows before customers are left waiting.
This is for operations and planning teams at grocery and quick commerce firms that sell timed delivery windows. It explains what a slot really depends on, how to collect the speeds, a made-up example of the arithmetic, where teams go wrong, and an AI employee that does the measuring and the warning.
What a delivery slot really depends on
A slot is a promise that a van or rider will reach a set of homes inside a window. How many homes fit depends on three things: the time spent at each door, the distance between doors, and how fast the vehicle moves between them. The first two barely change from one Tuesday to the next. The third changes by the hour, and differently in each part of the city: a zone around a ring road can be quick at 11:00 and slow at 17:30, while a zone of narrow residential streets is slow all day but never stops.
Firms that offer the same number of slots in every zone at every hour end up either late in the busy zones or half empty in the quiet ones.
Collecting speeds zone by zone
The Traffic Flow block measures the roads around a point at the moment the flow runs: the current speed, the free-flow speed, a jam factor from 0 (free) to 10 (standstill), the congestion as a percentage and the busiest road. It does not count vehicles and it has no history of its own, so the history comes from running it on a timetable and keeping what it returns.
- Pick a few points per zone. One point in the middle of a zone describes the middle. Three or four, on the roads vans actually use, describe the zone. Places within 300 m of each other share a measurement, so spread them out.
- Run it every 30 minutes across your delivery hours, every day, and write each reading to a sheet.
- Keep it going for three or four weeks so that each weekday has several readings at each half hour, and one odd day does not set the pattern.
- Average by zone, weekday and half hour, then look at the congestion percentage rather than the raw speed: a zone whose free-flow speed is low is not slow because of traffic.
The arithmetic, with made-up numbers
Say a van in a suburban zone of Johannesburg spends 6 minutes at each door, and the drops in a slot are spread over about 12 km of driving. At the zone's free-flow speed of 40 km/h, the driving takes 18 minutes. In a two-hour slot that leaves 102 minutes for doors, so 17 drops fit with no slack.
Now say the readings show this zone at 55% congestion on weekdays between 16:30 and 18:30, so the van moves at about 18 km/h. The same 12 km takes 40 minutes, which leaves 80 minutes for doors and 13 drops. The planner offers 13 drops per van in that evening slot, or keeps 17 and stretches the window to two and a half hours.
None of those figures come from real deliveries: they show how a speed reading turns into a slot size. Your own door times come from your own records.
Where teams go wrong
- Measuring once, at one hour. A single morning reading tells you about mornings. Slots are set across the whole day.
- Treating speed as volume. Live traffic data measures how fast traffic moves, not how many vehicles there are. That is the right measure for a slot, but it will not tell you where demand is.
- Ignoring closures. A road closed for works can make a zone slow for weeks. Traffic incidents lists closures and roadworks in each zone, so a sudden change in the averages has an explanation.
- Letting the plan go stale. School holidays, a new bridge or a stadium season change the pattern. Keep the measuring flow running and look at the averages every month.
An AI employee that measures and warns
Two flows do the job: one builds the history quietly, the other watches the day. Both are the kind of routine work described on the AI employee for field operations page.
The history flow
- Schedule: every 30 minutes, with active hours set to your delivery hours.
- My data: a table of zone points, each with its zone name.
- Traffic Flow: roads within 500 m of each point.
- A Google Sheets app block: one row per point per run, with the time, zone, speed, free-flow speed, jam factor and congestion.
The day-of watch
- Schedule: every 15 minutes in delivery hours.
- My data and Traffic Flow, as above.
- Filter: jam factor above 6.
- Only new, so a zone that stays jammed is reported once, not every quarter of an hour.
- Email or a Slack message to dispatch: which zones are jammed now, the busiest road in each, and the slots starting in the next hour.
Dispatch gets a short alert like this: zone 7 at jam factor 7.4, busiest road named, three slots starting before 18:00. They can add a van, move drops, or text customers before the window is missed.
Questions
Can live traffic data tell me how many cars are on a road?
No. It gives speeds, the free-flow speed and a jam factor. Car counts by hour are only available for roads in the United Arab Emirates, as a separate regional dataset.
How long should I collect speeds before changing slots?
Three or four weeks gives each weekday several readings at every half hour. Keep collecting afterwards so the averages follow changes in the city.
How many points do I need per zone?
Enough to cover the main roads vans use, often three or four. Points within 300 m of each other share one measurement, so closer points add nothing.
Does this replace route planning?
No. Slot sizing decides how many drops to sell; route planning decides the order of the drops once they are sold. The speeds help both.