Systems Thinking with AI

Chapter 17

Cohorts, Aging Chains, and Coflows

People are not interchangeable units, and the experience they hold moves when they move.

Four demonstrations built on the chapter's three-band workforce. Check that experience leaves with the people who leave, run a hiring surge and read three measures, tell four workforce stories from two curves, and see why the capability gained per head hired can fall short of one.

Every example in these readers is a constructed teaching example built from the chapter's own numbers. Chapters 36 to 39 start from published public records; their fitted values are inferred from those records, and nothing here is a forecast.

1Demonstration 1 of 4

Experience cannot be left behind

When people leave a band, what happens to the average experience of the people who remain?

In the chapter's model, leavers take the band's own average with them, so the ratio moves only by aging. If experience is a separate stock that ignores departures, attrition shrinks the people and not the years, and the average rises every time someone quits.

Equation: a junior band of 20 people holding 40 person years

Equation: after a short step the junior band holds 19.9 people

The junior band holds 20 people and 40 person-years, so 2.0 years each. Attrition is the fraction leaving per quarter; dt is the step length in quarters. Aging adds 0.25 x dt years for everyone present.

Predict first. With attrition 0.5 and dt 0.01, does the average experience in the junior band stay close to 2.0 when experience is carried?

Choose an example

Figure: Experience cannot be left behind. Two panels for the junior band: people fall from 20 to 19.90, and average experience goes from 2.000 to 2.003 years with experience carried by the people.
Attrition fraction per quarter: 0.5, Step length dt (quarters): 0.01, How experience is tracked: Carried by the people
Constructed example: the chapter's starting band of 20 people and 40 person-years, advanced with the chapter's advance function.

Calculated values

People in the band after
19.90
Total experience after (person-years)
39.85
Average experience after (years)
2.003
Change in the average (years)
0.003
Aging alone would add (years)
0.0025

Leavers = 0.01 x 0.5 x 20 = 0.10, so people = 20 - 0.10 = 19.90. Experience = 40.0 + 0.050 - 0.10 x 2.0 = 39.850. Average = 39.850/19.90 = 2.003, a change of 0.003 years against 0.0025 from aging. Leavers took their own share, so the average moved by about the aging alone.

Use the idea

Plot the ratio of experience to people, not only the totals; the parallel-stock bug hides in the totals.

Where the conclusion applies

One band, no hires and no maturation, short steps. The parallel version is the bug the chapter describes, computed here by plain arithmetic and not by the chapter's code. Aging and departing experience use the start-of-step state, which is why the chapter's test uses a short step.

Check your understanding: With dt = 0.1 and attrition 0.2, how many people leave the junior band in one step?
0.1 x 0.2 x 20 = 0.4 people, leaving 19.6.

Chapter 17 source: section "Experience cannot be left behind".

2Demonstration 2 of 4

What the surge does

If headcount nearly doubles, how much does capacity rise?

Hires add heads at once and capability slowly, because they arrive with none. With enough hires the junior band also gets greener, since newcomers arrive faster than the juniors mature; with few hires it does not.

Equation: the headcount across all bands

Equation: effective capacity of the bands on a four year ramp

Equation: average years of experience in the junior band

Three bands start with 20, 30 and 50 people holding 40, 150 and 500 person-years. Hires enter the junior band with no experience. Effective capacity counts each person in proportion to their experience up to the ramp, and fully after it.

Predict first. With 40 hires per quarter and a four-year ramp, does capacity rise by more or less than headcount after 3 quarters?

Choose an example

Figure: What the surge does. Left: headcount goes from 100 to 193 and effective capacity on a 4-year ramp from 90 to 96. Right: the junior band's average experience goes from 2.00 to 0.46 years.
Hires per quarter: 40, Ramp to full capacity (years): 4
Constructed example: the chapter's three-band workforce and its 40 hires per quarter, with other hire sizes and a two-year ramp defined for this reader.

Calculated values

Headcount
100 to 193.4
Headcount rise (percent)
93
Effective capacity
90.0 to 96.2
Capacity rise (percent)
7
Junior average experience (years)
2.00 to 0.46

40 hires x 3 quarters = 120 people entered; headcount rose by 193.4 - 100 = 93.4, so 120 - 93.4 = 26.6 left. Headcount rise = 93.4/100 = 0.934. Capacity on a 4-year ramp changed by 96.2 - 90.0 = 6.2, which is 6.2/90.0 = 0.069. The gap between the two rises is the dilution a headcount model reports as zero.

Use the idea

Report capacity beside headcount whenever the workforce is changing fast.

Where the conclusion applies

Three bands, the chapter's maturation and attrition rates per quarter and no training cost. The model steps one quarter at a time and everyone present accrues 0.25 years of experience per quarter, so the ramp is in years.

Check your understanding: If 120 people entered over three quarters and headcount rose by 93.4, how many left?
120 - 93.4 = 26.6 people left through attrition.

Chapter 17 source: section "What the surge does".

3Demonstration 3 of 4

Read headcount and average experience together

What does it tell you when headcount and average experience move in different directions?

Each pairing of directions is a different story: growing and diluting, aging in place, or the experienced people leaving. Aging pushes the average up over time, so the dilution shows most clearly early in a surge.

Equation: the headcount across all bands

Equation: average years of experience in the junior band

Headcount is people across all bands. Average experience is total person-years divided by headcount. The three scenarios differ only in hiring and attrition.

Predict first. Under the senior band leaving, do headcount and average experience both fall after one quarter?

Choose an example

Figure: Read headcount and average experience together. Two panels over 3 quarter(s) for the scenario hiring surge: headcount moves from 100 to 193 and average experience from 6.90 to 3.59 years.
Scenario: Hiring surge, Quarters shown: 3
Constructed example: the chapter's three-band workforce under three scenarios defined for this reader, advanced with the chapter's advance function.

Calculated values

Scenario
Hiring surge
Headcount
100 to 193.4 (rising)
Average experience (years)
6.90 to 3.59 (falling)
Reading
growing and diluting at once

Start: 690 person-years / 100 people = 6.90 years. After 3 quarter(s): 693.6 / 193.4 = 3.59 years, a change of (-3.31). Headcount changed by 93.4. Headcount rising with average experience falling reads as growing and diluting at once.

Use the idea

Plot both curves and read them as a pair, not one at a time.

Where the conclusion applies

Constructed scenarios in which the senior band loses 0.8 of its people per quarter in the third; real attrition is rarely uniform across bands, which a single rate cannot show.

Check your understanding: Start with 690 person-years among 100 people. What is the average experience?
690 / 100 = 6.9 years per person.

Chapter 17 source: section "Three curves, read together".

4Demonstration 4 of 4

The recommendation that reverses

Inside a short horizon, how much capability does each person hired actually deliver?

Hires arrive with no experience, so within a short horizon with a long ramp they add heads and little capability. The ratio climbs toward one as the hires mature.

Equation: the headcount across all bands

Equation: effective capacity of the bands on a four year ramp

Heads added is headcount with the surge minus headcount without it after the same number of quarters. Capacity per head is the capacity gained divided by heads added. A headcount model assumes 1.00.

Predict first. After 1 quarter on a four-year ramp, how much capacity does a head added by hiring deliver?

Choose an example

Figure: The recommendation that reverses. Capacity gained per head added over a horizon of 2 quarter(s) on a 4-year ramp, reading 0.09 at the horizon against a dashed line at 1.00.
Horizon (quarters): 2, Ramp to full capacity (years): 4
Constructed example: the chapter's surge of 40 hires per quarter against a no-hiring run, compared with the chapter's effective_capacity.

Calculated values

Heads added by hiring
76.0
Capacity gained
6.53
Capacity per head added
0.09
Shortfall against 1.00 per head
0.91

With 40 hires per quarter against none, 76.0 more people remain after 2 quarter(s). On a 4-year ramp they add 6.53 of capacity. Capacity per head = 6.53/76.0 = 0.09, so the shortfall against a headcount model is 1.00 - 0.09 = 0.91. Each added head delivers a fraction of a head inside this horizon, and the training cost, which this model omits, would lower it further.

Use the idea

State the horizon next to a hiring recommendation; the case for hiring has to be made on the capability curve.

Where the conclusion applies

No training cost, which the chapter says its model omits and which would lower the ratio. Beyond the horizon where hires mature the headcount and chain models agree.

Check your understanding: If 100 heads are added and they deliver 29 units of capacity, what is the capacity per head?
29 / 100 = 0.29, so each head delivers 0.29 of a head inside the horizon.

Chapter 17 source: section "The recommendation that reverses".