Second-Order Effects of a Policy Change
Why this matters
A change to how techs are scheduled or paid always does what you designed it to do, at first. The trouble is what it does next. Send techs straight from home to the first job and you save the morning drive, and then the trucks stop getting restocked, the huddle where juniors learned from the lead tech disappears, and by month three the supply house runs have eaten the savings. First-order effects are the ones you planned. Second-order effects are the reactions to those, and third-order effects are the reactions to the reactions. Owners who trace two steps out before announcing a change save themselves the reversal, which costs more than the change ever saved because a reversed policy also costs trust.
The one question that does the work
For every effect you expect, ask "and then what?" Then ask it again about the answer. Howard Marks, writing about investing, called this second-level thinking: the first-level thinker sees the obvious consequence, the second-level thinker asks what everyone's response to that consequence will be. In a shop, "everyone" is a small, known list of people, which makes the tracing very doable.
The case used throughout
An HVAC and plumbing shop with 8 techs, 2 of them juniors. Today everyone reports to the shop at 7:00, loads parts, has a 15-minute huddle, and leaves around 7:40. The average drive from the shop to the first job is about 35 minutes. The owner wants techs to take their trucks home and drive straight to the first job, which averages about 25 minutes from their homes.
Step 1: State the first-order effect as a number
Write down what the change is supposed to achieve, in a unit you will measure.
The case: each tech skips about 40 minutes at the shop and about 10 minutes of drive (35 minus 25), so about 50 minutes per tech per day. Across 8 techs that is 400 minutes a day, about 6.7 hours, or about 33 hours a week (6.7 times 5) of capacity freed for billable work.
Why a number: at the end you will compare the second-order costs against this, and "more efficient mornings" cannot be compared against anything.
Step 2: List everyone the change touches
Name every role that does something differently because of the change, including people who are not the target.
The case: the techs, the two juniors in particular, the lead tech who mentors them, the dispatcher, whoever restocks trucks, the customers, and the owner's insurance and vehicle policies.
What breaks if you skip it: the effects you miss are almost always on the people the policy was not aimed at. The parts person and the dispatcher do not appear in "techs start from home," and that is exactly where the cost lands.
Step 3: For each one, ask "and then what?" twice
Trace each role out to a second and third step. Write it down; this is where a sketch earns its place.
POLICY: techs start from home, trucks go home
first order second order third order
----------- ------------ -----------
no shop stop -> no morning restock -> stockouts, midday
supply house runs
-> no huddle -> juniors lose the
lead tech's time,
callbacks drift up
-> dispatcher phones -> changes reach
each tech techs late
truck at home -> techs treat it as -> retention up,
a perk easier hiring
-> personal use, -> insurance and
parking at homes policy questions
Notice that one of the five chains ends somewhere good. Second-order thinking is not a search for reasons to say no. A take-home truck is one of the more valued perks in the trades, and a policy that makes hiring easier may be worth some inefficiency.
Step 4: Look in the four places second-order effects hide
The tracing in step 3 finds the obvious chains. These four places find the rest.
Work nobody wrote down. The morning at the shop was not wasted time. It was when trucks got restocked, when the dispatcher reshuffled the day face to face, and when the juniors asked the lead tech about yesterday's problem job. None of it appeared in any job description, so none of it was in the owner's picture of what the change would remove.
Work that moves to someone else's desk. Restocking does not vanish; it moves to a weekly session or to the tech's evening. Schedule changes move to the dispatcher's phone.
How people adapt to the new measure. If the policy is judged on "time to first job," expect techs to optimize that number, for instance by leaving the first job's paperwork for later.
What changes over months, not weeks. Truck stock accuracy decays slowly. A junior who misses mentoring does not get worse in week one; he plateaus over a season. These effects appear long after anyone is watching.
Step 5: Size each effect in the same unit as the first-order gain
Put a rough number on each second-order effect, in the same hours-a-week unit as step 1, so you can net them.
The case, with each assumption stated:
- Supply house runs. With restocking moved to Friday afternoons, the owner expects about 3 midday runs per tech a week at about 45 minutes each. That is 8 techs times 3 runs times 0.75 hours, about 18 hours a week.
- Dispatcher time. About 30 extra minutes a day phoning changes, 2.5 hours a week.
- Junior callbacks. Without the huddle, the owner expects the two juniors' callbacks to rise by about 2 a week between them, at about 2.5 hours each, 5 hours a week.
Total second-order cost: 18 plus 2.5 plus 5, about 25.5 hours a week, against a first-order gain of about 33. The net gain is about 7.5 hours a week, under a quarter of what the policy promised on paper.
Step 6: Redesign before you decide
Use the trace to change the policy, not just to judge it. Most second-order costs can be cut with a small design change once they are visible.
The case: the owner adjusts three things.
- A parts runner restocks trucks Tuesday and Friday afternoons, about 4 hours a week. The owner expects that to cut supply house runs to about 1 per tech a week: 8 times 1 times 0.75, 6 hours.
- The juniors keep reporting to the shop two mornings a week to ride with the lead tech, which gives back part of the saving: 2 juniors times 2 days times 50 minutes is 200 minutes, about 3.3 hours a week of capacity not freed. The owner assumes that keeps junior callbacks at today's rate.
- The dispatcher's phoning stays at 2.5 hours a week, because no change addresses it.
Redesigned: the gain is about 33 minus 3.3, so about 29.7 hours. The costs are 4 (runner) plus 6 (runs) plus 2.5 (dispatcher), about 12.5 hours. Net about 17 hours a week, a bit over twice the 7.5 of the original design, and the juniors keep their mentoring.
Step 7: Check the legal second-order effect before you announce
Changing where the workday starts changes what counts as paid time, and this is the second-order effect most owners never trace. This is not legal advice.
At the federal level, under the Portal-to-Portal Act and its Employee Commuter Flexibility Act amendment, ordinary travel from home to the first job is generally not paid working time, and that holds when a nonexempt employee drives a company vehicle within the normal commuting area under an agreement with the employer (29 CFR 785.35 covers home-to-work travel). But the workday begins with the first principal activity, under the continuous workday rule at 29 CFR 790.6, which the Department of Labor's travel-time guidance also walks through. If a tech has to load parts at home, take calls from the dispatcher or review the day's tickets before he drives, the clock may start in his driveway, and that drive becomes paid time. That would cut straight into the first-order gain.
The state layer can go further. California measures hours worked by whether the employee is under the employer's control, and its courts have treated required travel in employer-provided transport as paid time, so a mandatory take-home program there needs review before launch. Most states follow the federal approach, but several have their own wage-and-hour definitions. Before announcing, take the draft policy, a description of what techs will do before the first job, and whether participation is voluntary to an employment attorney in your state.
Step 8: Announce with the measures you will watch, then check at 30 and 90 days
Tell the crew what the policy is for and which numbers you will watch: supply house runs, callbacks by tech, and arrival time at first jobs. Pilot it where you can, with half the techs for a month.
The case, at 90 days: supply house runs averaged about 1.4 per tech a week, more than the 1 expected, mainly on the two trucks with the most varied work. Junior callbacks held flat. Two techs mentioned the take-home truck when a candidate asked why they liked working there. The owner adds a Wednesday restock for the two high-variety trucks and leaves the rest alone.
A second, quicker trace: the callback penalty
The same method works on pay. An owner proposes docking a tech's bonus for every callback. First order: techs take more care. Second order: techs steer away from intermittent, hard-to-diagnose faults that are most likely to come back, and some callbacks get logged as new service calls. Third order: the hardest jobs pile up on the one tech who still takes them, and the callback rate becomes a number nobody trusts. Ten minutes of tracing finds all three. How pay plans shape what techs recommend is its own subject. See related: Incentives Explain Behavior Better Than Character.
References
- Marks, H., The Most Important Thing (2011), on second-level thinking
- Portal-to-Portal Act, 29 U.S.C. 254, as amended by the Employee Commuter Flexibility Act of 1996; 29 CFR 785.35 and 29 CFR 790.6
- US Department of Labor, Wage and Hour Division, guidance on travel time under the FLSA
- See related: Incentives Explain Behavior Better Than Character; The Tech Kept Recommending Replacement and It Was the Pay Plan; Running a Small Experiment Instead of Arguing; Reversible and Irreversible Decisions in a Small Shop