The Tech Kept Recommending Replacement and It Was the Pay Plan

Why this matters

An HVAC owner got a call from a homeowner who had been told her system needed replacing, got a second opinion, and had it repaired for a small fraction of the replacement quote. The tech who made the first recommendation was one of the shop's best: well liked, good reviews, sharp diagnostics. A quick look at the numbers showed he recommended replacement on repair-or-replace calls about twice as often as the rest of the crew. The owner's options as he saw them were a retraining talk or a much harder conversation about honesty. The review he ran instead tested four explanations in order, from cheapest to check to most awkward, and the answer was none of the two he had in mind. This is how it was reasoned.

The general principle, that a system's rewards explain a pattern of behavior better than character does, has its own card. See related: Incentives Explain Behavior Better Than Character. This is the case.

The signal

The shop defines a repair-or-replace call as any service call where the tech finds a major failure (compressor, heat exchanger, coil leak, failed blower motor on an older unit) and the customer is presented a choice. Over the last two quarters:

• Marcus: 90 such calls, replacement recommended on 34, which is 34 of 90, about 37.8%.
• The other five techs combined: 340 such calls, replacement recommended on 61, which is 61 of 340, about 17.9%.

Roughly double. On two quarters of calls, with 90 on his side, that is too large a gap to wave off as noise.

Hypothesis 1: his diagnosis is wrong

If Marcus were misreading failures, calling a leaking valve core a failed coil, say, he would recommend replacement on systems that only needed a repair, and he would believe he was right.

The check. The shop's senior tech reviewed 12 of Marcus's replacement recommendations from the job records: photos, readings, his notes, and in four cases a revisit to the house during a maintenance call. The question was narrow: was the failed component the one Marcus said had failed?

The result. On 11 of the 12, yes. The twelfth was a judgment call the senior tech said he might have made the same way. The homeowner's second-opinion case was among the 12, and Marcus's diagnosis was correct: a leaking evaporator coil on a 13-year-old system. The other company had recommended replacing the coil; Marcus had recommended replacing the system.

Ruled out: a skill problem. His diagnoses held up. Whatever was happening was happening after the diagnosis, at the recommendation.

Hypothesis 2: he is misrepresenting the condition

The harder version: Marcus is overstating problems to push sales.

The check. The same 12 reviews, read for whether his notes and what he told customers matched the evidence. Plus the shop's complaint log and reviews for his customers over the two quarters.

The result. No misstatements of condition in any of the 12. His notes were accurate and more detailed than most. Complaints and reviews were in line with the rest of the crew. In the second-opinion case, both recommendations were defensible on a 13-year-old system with a leaking coil, and plenty of good techs would have made either one.

Ruled out: fabrication. That does not settle whether his judgment was tilting, but it removes the explanation the owner most dreaded, and it changes the conversation that comes next.

Hypothesis 3: he sees older equipment

Marcus covers the north side of the service area, which has more homes from the 1980s and 1990s. Older systems legitimately get replacement recommendations more often. If his calls skew old, a higher rate is expected.

The check. Split every repair-or-replace call by system age, for Marcus and for the other five:

His calls do skew older: 48 of his 90 (about 53%) were on systems 15 years or older, against 120 of 340 (about 35%) for the others.

How much does the mix explain? Apply the other techs' rate in each age band to Marcus's calls: 48 times 0.417 is about 20.0, 30 times 0.069 is about 2.1, 12 times 0.022 is about 0.3, for a total of about 22.4. So if Marcus recommended replacement at the crew's rate in each band, his older territory alone would produce about 22 replacement recommendations out of 90, about 25%, not the crew's 18%.

Partly explained. Mix accounts for a real piece of the gap, moving the expected rate from about 18% to about 25%. It leaves about 34 minus 22 = 12 recommendations unexplained. And the gap is not evenly spread: it is widest in the 10 to 14 year band, about 23% against 7%, which is exactly where repair and replacement are both defensible and the recommendation is a judgment call.

This step mattered twice. It kept the owner from blaming Marcus for his territory, and it located where the remaining tilt lived.

Hypothesis 4: what he is paid for

Seven months before the review, the owner had moved Marcus onto a pilot "comfort advisor" plan, intended to reward his strong customer skills. Nobody else was on it. The two plans, as written:

SERVICE TECH PLAN (standard, 5 techs)
  Hourly pay per wage schedule
  Replacement lead that closes in 30 days:
    small flat spiff, same for any system
  Approved major repair: no extra pay

COMFORT ADVISOR PILOT (Marcus, from 05/01)
  Hourly pay per wage schedule
  Equipment sale he originates:
    set percentage of the sale's gross margin
  Approved major repair: no extra pay

Under the pilot, a replacement Marcus recommended and the customer accepted could pay him many times what the standard spiff paid. A major repair, the other honest answer on the same call, paid him his hourly rate and nothing more.

The check. If the plan is the cause, his rate should have shifted when the plan started, and the shift should show up in the judgment-call band. The office pulled his repair-or-replace calls for the two quarters before 05/01: replacement recommended on 19 of 88, about 21.6%. That sits a little below the roughly 25% his age mix predicts. Then, in the two quarters on the pilot, 34 of 90, about 37.8%.

Confirmed, as far as a shop can confirm it. The jump coincides with the pay change, it is concentrated where both answers are defensible, and nothing else about his territory, his skills or his honesty moved at the same time. This is not proof in a laboratory sense; it is the explanation that fits every piece of the evidence, where the first three each fit only part of it.

Note what it does not say. It does not say Marcus decided to oversell. It says that on the calls where a good tech could go either way, a plan that paid heavily for one answer and nothing for the other tilted which way he went. The substitution test from the incentives card applies: put another strong tech on the same pilot and the honest expectation is the same drift.

The fix

The owner changed the plan, not the person, and told Marcus why, plainly and without accusation:

• Pay both honest answers. An approved major repair now carries an advisor payment too, so the two options no longer pay wildly differently.
• Second look in the judgment band. A replacement recommendation on a system under 15 years old gets a quick review of the readings by the senior tech before it goes to the customer. Above 15 years, no review.
• Both options on paper, every time. Every repair-or-replace customer gets the repair and the replacement written up side by side, with what each would cost them and the expected life of each. The customer chooses with both in front of them.

How to structure sales pay so it does not reward overselling is its own subject. See related: Commission Structures That Don't Encourage Overselling. Any change to how techs are paid also touches overtime rules for non-exempt employees, and that is covered, with its state layer, in the incentives card.

How they confirmed it worked

The office tracked Marcus's rate by age band for the next two quarters: replacement recommended on 26 of 92 repair-or-replace calls, about 28%, against the roughly 25% his age mix predicts. In the 10 to 14 year band it fell to 4 of 31, about 13%, still above the crew's 7%, but each of those four recommendations went through the senior tech's review under the new rule, and he agreed with all four.

His total equipment sales fell by less than the recommendation rate did, because customers who were shown both options side by side accepted more of the replacements they were offered. And the shop has had no second-opinion complaint about his calls since.

The order mattered

The owner ran the four checks from cheapest and least accusatory to most. Skill and honesty both took a dozen record reviews. Mix took one report. The pay plan took reading two documents and one more report. Had he started with the honesty conversation, he would have damaged his best tech's trust in the shop to discover something a pay plan explained, and the age-band table would still have been needed to see where the tilt was.

There was also a reason the pay plan came last in the owner's own head, and it is worth naming. He had designed the pilot himself, a few months earlier, as a reward for a tech he rated highly. Owners rarely suspect their own designs, and a plan built as a thank-you does not feel like something that could bend anyone's judgment. That is exactly why reading the plan as written, rather than as intended, belongs on the list whenever one person's pattern changes.

The same lens cuts both ways. The second-opinion company, which recommended the cheaper coil, works on its own pay plan too, and a shop that pays heavily for repairs and little for replacements tilts the other way in the same judgment band. Neither recommendation was wrong. Each was a defensible answer that its tech's pay made a little easier to give.

References

• See related: Incentives Explain Behavior Better Than Character
• See related: Commission Structures That Don't Encourage Overselling
• See related: Correlation, Causation and Your Job Data
• See related: Base Rates: The Number You Should Start From