Module 6, lesson 1 · 45 minutes

RSET, ASET and pre-movement time

Two clocks decide whether everyone gets out of a burning building, and the longest part of the second one is usually the time before anyone moves.

You’ll be able to

define RSET and its four components, ASET, and the safety margin between them;
read the PD 7974-6 pre-movement percentiles for an occupancy and a standard of management;
explain why pre-movement so often dominates the total, and what shortens it.

An alarm sounds in an office. Nobody runs. People look up, look at each other, and someone asks whether it is a drill. One person saves a file; another goes back for a coat. A minute passes, then two. Only when someone in authority says “out, now” does the floor start to empty.

That pause is not foolishness and it is not panic. It is the normal human response to an ambiguous signal, and it was measured on the worst day imaginable. In the World Trade Center on 11 September 2001, NIST found that occupants of WTC 1, the North Tower, on floors up to 76 took a median of 3 minutes to start evacuating; 70% talked to others and 46% gathered personal items before they left. Fire engineers have a name for this interval, pre-movement time, and this lesson is about why it is so often the biggest number on the clock.

Two clocks

Every evacuation is a race between two clocks that start at the same moment, ignition. The first belongs to the fire. Smoke, heat and toxic gases build up until, at some point, conditions where people are become untenable. The available safe escape time, ASET, is the time from ignition until conditions in a place stop being survivable. Working it out is the fire engineer’s half of the problem: smoke spread, ventilation, the size of the space.

The second clock belongs to the occupants. The required safe escape time, RSET, is the time from ignition until the last occupant reaches a place of safety. British practice, set out in PD 7974-6:2019 and reproduced by the CFPA-E, splits it into four parts:

$\text{RSET} = t_{\text{detection}} + t_{\text{alarm}} + \left(t_{\text{pre-movement}} + t_{\text{travel}}\right) \qquad \text{margin} = \text{ASET} - \text{RSET}$

RSET = t(detection) + t(alarm) + (t(pre-movement) + t(travel)); margin = ASET − RSET

detection: until a detector, or a person, notices the fire
alarm: until occupants are warned
pre-movement: recognition plus response: until a person starts to leave
travel: walking to safety, including waiting at doors and stairs

PD 7974-6 splits pre-movement itself into recognition (realising something is wrong) and response (deciding what to do and doing the things people do before they leave). The design requirement is simple to state: ASET must exceed RSET by an appropriate safety margin. The document does not put a number on “appropriate”; that judgement is left to the engineer and the approving authority.

The two clocks. The building is safe if the last occupant is out before conditions become untenable, with a margin to spare. The block sizes are schematic, but the proportions are typical of what this lesson finds: pre-movement is long and travel is short.

Everything a designer does either lengthens ASET (smoke control, compartmentation, sprinklers) or shortens RSET (faster detection, clearer warnings, better management, more exit width). The margin is what that work buys.

Pre-movement: the numbers

Pre-movement time is the interval between the warning and the moment a person starts to move towards an exit: recognition plus response. It is not one number. In any group some people go at once and a few take much longer, so PD 7974-6 gives it as a distribution, summarised by its 1st and 99th percentiles: the quickest one in a hundred and the slowest one in a hundred.

The distributions depend on two things. The first is the occupancy: whether people are awake or asleep, and whether they know the building. The second is the management level, M1 to M3, which describes the standard of the building’s fire safety management, M1 being the highest.

Pre-movement time, 1st / 99th percentile (minutes)PD 7974-6:2019 as reproduced by CFPA-E Guideline 19:2023. “>” means the document gives only a lower bound.
Occupancy	M1	M2	M3
A: awake and familiar	0.5 / 1.5	1 / 3	>15 / >30
B: awake and unfamiliar	0.5 / 2.5	1 / 4	>15 / >30
Ci: sleeping, individual dwellings	not given	5 / 10	10 / >40
Cii: managed sleeping occupancy	10 / 30	15 / 40	>20 / >40

Read across a row and the effect of management is startling. For awake, unfamiliar occupants (category B), the slowest one in a hundred starts within 2.5 minutes under M1, 4 minutes under M2, and more than 30 minutes under M3. For the same people in the same building, the standard of management changes the slowest start by an order of magnitude. Complex buildings add more: the table is extended by a wayfinding allowance of 0.5–1.0 minutes for buildings at the higher complexity levels, B2 and B3.

The table is secondary evidence in one sense: PD 7974-6 is a paid British Standards document, and these values are as reproduced by the Confederation of Fire Protection Associations Europe. They are design values, built from experience and from studies of real evacuations, not laws of nature.

Why pre-movement dominates

David Purser’s analysis of how RSET is used in practice made the point directly: pre-movement “often represents a greater component of the total escape time”, and it depends on the occupancy type, the warnings given, the occupants’ characteristics, the building’s complexity and its management. Travel times are usually short because buildings are designed to make them short, with limited travel distances. Nothing in the building regulations limits how long people take to decide to go.

Real evacuations agree. The NIST median start time in WTC 1 was 3 minutes on floors up to 76 and 5 minutes on floors 77–91, near the impact region. Among 119 survivors in the UK’s HEED study whose response times could be estimated, 13% responded in under a minute, 78% took between 1 and 8 minutes, and 9% took more than 8. Lesson 4 of this module looks at what they were doing in that time.

A worked example

A two-floor shop (a fictional example) is occupied by customers who do not know it well: category B. Its management is assessed as M2. The fire engineer’s smoke model gives an ASET of 10 minutes on the shop floor.

Detection. An automatic system is assumed to detect the fire in 1 minute.
Alarm. The warning reaches everyone 0.5 minutes later.
Pre-movement. For the design case, the slowest one in a hundred, the table gives 4 minutes.
Travel. The longest route is 45 m. At 1 p/m² the SFPE relation gives a walking speed of 1.40 − 0.266 × 1.40 × 1.0 ≈ 1.03 m/s, so the walk takes 45 ÷ 1.03 ≈ 44 s, or 0.73 minutes.
RSET = 1 + 0.5 + 4 + 0.73 ≈ 6.2 minutes. Margin = 10 − 6.2 ≈ 3.8 minutes.

Pre-movement is about two-thirds of the total. Now suppose the same shop is badly run: alarms are often false, staff are untrained, nobody takes charge. Under M3 the 99th-percentile pre-movement is more than 30 minutes, RSET exceeds 32 minutes, and the design fails by more than 22 minutes. No amount of exit width rescues it, because the slowest occupants have not yet started walking.

Start with the defaults: the shop from the worked example (B, M2, 99th percentile, ASET 10 minutes). Check the bar against the arithmetic above.
Change management to M3. Then raise ASET to 30 minutes. Does the design pass?
Back at M2, drag the percentile down to 50. How long does the median occupant take to start, and why is that the wrong number to design with?
Set the density on the route to 3 p/m². Walking slows to under 0.3 m/s. Compare that change in RSET with the change from M2 to M3.
Choose Cii, a managed sleeping occupancy, and look at the lower three bars. What would a designer have to do with ASET?

For awake, unfamiliar occupants under the lowest standard of management, the 99th-percentile pre-movement alone is more than 30 minutes: longer than a 30-minute ASET before anyone has walked a step. Faster walking cannot fix that. Earlier, clearer information and better management can.

Pre-movement anchors are PD 7974-6:2019 values as reproduced by CFPA-E; where the table gives only a lower bound, the bound is used and the bar ends in a ragged edge. Between the 1st and 99th percentiles the interactive assumes a lognormal shape, which is our interpolation, not the standard’s. Travel is walking time only, with speed from the SFPE relation; queuing at doors and stairs, the subject of the next lesson, would add to it. The amber line at a 2-minute margin is illustrative: PD 7974-6 sets no number.

A managed sleeping occupancy (Cii) under M2 has detection in 2 minutes and alarm 1 minute later. What is RSET for the slowest one in a hundred before anyone has walked anywhere, and what does that imply for the design?

2 + 1 + 40 = 43 minutes before travel even begins (the 99th-percentile pre-movement for Cii at M2 is 40 minutes). A design cannot rely on everyone leaving promptly, so it must either keep conditions survivable where people are for a long time, a very long ASET, or improve management so that pre-movement falls.

Fixed times: the Green Guide’s shortcut

Sports grounds use a different, older logic. The Green Guide does not ask how long spectators take to start moving. It sets fixed times and sizes exits to meet them: a normal maximum egress time of 8 minutes (the 6th edition calls it the “Zone 2 travel time”) and an emergency evacuation time of 2.5 to 8 minutes, no more than 2.5 minutes where the fire risk is high. Capacity then follows from width × time × rate, which the Green Guide capacity lesson turns into a calculation.

The two approaches answer different questions. The fixed-time method is prescriptive: follow the rule and the ground is deemed safe. RSET/ASET is performance-based: work out what will actually happen and show the margin. A stadium crowd at the final whistle has already decided to leave, which is one reason fixed egress times work there. An office or a hotel at night has not, which is why pre-movement must be modelled.

Notice which way the evidence points. The danger in a building evacuation is rarely that people flee too fast. It is that they start too late. Module 3’s panic myth lesson makes the same point from the psychology side, and collective resilience explains why clear, honest information is the strongest lever on pre-movement.

Why is the 99th percentile, not the median, used as the design pre-movement time?

RSET is the time for the last occupant to reach safety. A design based on the median would leave about half the occupants still inside when the margin is calculated. The 99th percentile stands in for “nearly everyone, including the slow starters”.

“Evacuation time is mostly walking time.” In most buildings, the minutes before people start moving are longer than the minutes they spend walking. Purser calls pre-movement often the greater component.
“People die in fires because they panic and run.” The common failure is the opposite: delay, information-seeking and finishing tasks before leaving.
“RSET is one number for a building.” It is a distribution over people. The design case is the slow tail, around the 99th percentile.
“The Green Guide’s 8 minutes is an RSET.” It is a fixed design time used to size exits, not a prediction of how long a crowd takes to get out.