Analytics · Cohorts
A day is not
a demand segment.
Calendar labels are convenient reporting buckets. They are not explanations. The people using the curb, the reasons they parked, and the clocks governing their departures determine the operating shape.
The mistake
An average across unlike guests describes nobody.
A Tuesday of business checkouts may peak before breakfast. A Wednesday banquet may release hundreds of people at once after dessert. Their total vehicle counts can match while their staffing needs point in opposite directions.
Forecast the cohort and its timing pattern, then roll those forecasts into the day. Do not start with the day and hope the mix repeats.
The cohorts
Different clocks create different curbs
Name the demand source
Hotel
Business transient
Arrivals cluster around flights and meetings; departures often gather around morning checkout and transport schedules.
Venue
Event guest
Vehicles arrive across an opening window and may all be requested within minutes of the same ending.
Residential
Resident
Repeat behavior creates recognizable weekday rhythms, but individual routines and service expectations matter.
Restaurant
Diner
Table turns, meal length, weather, and the check process shape a rolling departure wave rather than a fixed checkout.
Compare shapes, not totals
Same daily count
MISLEADING- Hides how tightly departures cluster
- Hides parking distance and service mix
- Assumes yesterday's mix returns
Cohort forecast
ACTIONABLE- Estimates each demand source separately
- Carries its own timing and uncertainty
- Recombines into one staffing picture
Build segments from facts the operation can know
A useful cohort should change a decision and be reliably observable. Avoid elaborate profiles that require guesses about guests.
Source
Reservation or contract type
Hotel stay, resident permit, event ticket, restaurant validation, or transient parking can be operationally meaningful.
Window
Known time anchor
Checkout, event end, shift change, reservation time, or permit pattern can explain when demand may emerge.
Geometry
Parking and stand
The same cohort behaves differently when cars are remote, elevators are slow, or one lane serves several entrances.
Learning
Observed departures
Compare the expected cohort curve with what actually happened and update the forecast without rewriting the story after the fact.
→Only collect segmentation data with a legitimate operational purpose, appropriate access, and a defined retention policy.
The planning ritual
Ask what kind of Tuesday is coming.
Review reservations, resident rhythm, event calendars, restaurant covers, weather, and property constraints. Then compare this Tuesday with prior shifts that had the same mix, not merely the same weekday name.
The takeaway: the daily total is the sum of several demand systems. Forecast those systems before averaging them away.
