Anchoring on Last Year's Number

Why this matters

Most small-shop budgets, revenue targets and raise rounds are built the same way: take last year's figure and add a bit. It is fast, it feels conservative, and it is right more often than not. It is also how a shop sets a growth target its crew cannot physically deliver, keeps funding a lead source that stopped working two years ago, and gives its best tech the same raise as its weakest. The skill is not abandoning last year's number. It is knowing which lines it is still good for and rebuilding the ones where it is not.

What anchoring is

Anchoring is the tendency for a first number you see to pull your estimate toward it, even when you know the number is irrelevant. Amos Tversky and Daniel Kahneman showed it in 1974 with a rigged wheel of fortune: people who watched the wheel stop on 10 then guessed that about 25 percent of United Nations member countries were African, while people who watched it stop on 65 guessed about 45 percent. A random number moved a factual estimate by 20 points.

Later work by Nicholas Epley and Thomas Gilovich found that when people start from an anchor and adjust, they reliably stop adjusting too soon. They move away from the anchor until the answer seems plausible, and plausibility arrives early.

Last year's number is a far better anchor than a wheel of fortune. It is real data about your shop. That is precisely what makes it dangerous: it carries all the conditions that produced it, including the ones that no longer hold, and it feels so reasonable that nobody asks what those conditions were. Pricing has its own anchoring story about which option the customer sees first; this card is about the anchors owners set for themselves. See related: Anchoring: Present the High Option First.

The one question that decides it

Every line in a budget or target came from something: a count of trucks, a number of techs, a lead source's volume, a supplier's price list. Call those its drivers. The question is:

Has any driver of this number changed since last year by more than about a tenth?

If no driver moved much, last year plus a known adjustment (inflation, a price change you have been told about) is a perfectly good estimate, and rebuilding it from scratch is wasted effort. If a driver moved by more than a tenth, last year's number is describing a shop you no longer are, and adjusting from it will stop short. Rebuild that line from its drivers.

The tenth is a starting threshold, not a law. Tighten it for your five largest lines, where a small miss is a large amount, and loosen it for small lines where the time spent rebuilding outweighs any error.

The method for rebuilding from drivers rather than history has a name: zero-based budgeting, developed by Peter Pyhrr at Texas Instruments around 1970. Big companies apply it to every line every year, which is expensive. A small shop does not need to. It needs the question above, and zero-based thinking only where the answer is yes.

Case one: fuel, where the anchor holds

A five-truck shop sets next year's fuel budget. Last year's fuel was about 4 percent of total spend. The owner's instinct: last year plus 4 percent.

Run the question on the drivers:

• Trucks: five last year, five this year. No change.
• Miles per truck: the service area is the same, the job count is expected to be within a few percent, and routing has not changed. Last year's miles came in within 3 percent of the year before.
• Fuel price: the main unknown. The owner cannot forecast it better than anyone else, and an anchor of last year plus a few percent is as good a guess as a rebuild would produce.

No driver moved by a tenth. Last year plus 4 percent stands. Spending an afternoon rebuilding this line from miles, fuel economy and a price forecast would produce roughly the same number with false precision attached. Note what the question saved: the time to rebuild the lines that need it.

Case two: the revenue target, where the anchor breaks

Same shop, same morning. Last year: 2,150 completed jobs. The owner writes the target as last year plus 10 percent, 2,365 jobs, because 10 percent is what they grew the year before and it feels like the right ambition.

Run the question. The main driver of completed jobs is tech capacity, and it changed:

• Last year: five experienced techs, averaging 430 completed jobs each. 5 x 430 = 2,150, which matches the total.
• This year: one experienced tech left in December and was replaced by an apprentice. Four experienced techs plus one apprentice.

That is a driver moving by far more than a tenth: experienced capacity fell by a fifth. So rebuild from the drivers:

• Four experienced techs at last year's 430 each: 4 x 430 = 1,720 jobs.
• The apprentice: the shop's own record of its last two first-year hires shows them completing a little over half of what an experienced tech completes in year one. Use 55 percent as the shop's own figure, labeled as an estimate: 0.55 x 430, about 237 jobs.
• Capacity total: 1,720 + 237 = about 1,957 jobs, call it 1,955.

What the comparison shows. The anchored target of 2,365 is 410 jobs above what the crew can complete, about 21 percent over capacity (410 against 1,955). Even holding flat at last year's 2,150 would need 195 jobs, about 10 percent, that the crew does not have.

Why the anchor fails here and not on fuel. Adjusting from 2,150 would never have produced 1,955. Every adjustment an owner would naturally make starts from "we did 2,150" and asks how much more is possible. The rebuild starts from the people doing the work and finds the number goes down.

What the owner did with it. He did not set 1,955 as the goal and stop. The rebuild told him the target of 2,365 was a hiring decision, not a sales decision: closing the gap needs roughly one more experienced tech (another 430 jobs would put capacity near 2,385), and if that hire lands by April instead of January, the shop gets most of a year from it at best. He set a target of about 2,100 jobs, roughly flat, plus a hire, and moved the sales push to maintenance agreements for the shoulder months, where the crew has slack. Setting 2,365 and pushing marketing to hit it would have produced a backlog, overtime, and the kind of rushed work that turns into callbacks.

The anchor hidden inside a percentage

Raises are the most common anchored decision in a shop, and the anchor is often not a number but a percentage: "we gave 3 percent last year, so 3 percent again."

An across-the-board percentage anchors every person to their own last-year wage, which preserves every past mistake. A tech who was hired low because the shop was desperate two years ago stays low forever, three percent at a time. A tech whose skills grew sharply, say who now runs commissioning or handles the hardest diagnostic calls, gets the same raise as one whose skills stood still.

The driver question works here too. For each person, the drivers of what they should be paid are what the role pays in your market and what they can now do. If either moved by more than a tenth since their pay was last set, rebuild their number from those, not from their old wage. Where neither moved, a uniform percentage is fair and quick. The result is usually a round where most people get the standard increase and two or three get a rebuild, which is a much smaller conversation than it sounds. See related: Incentives Explain Behavior Better Than Character.

What changes the answer

Two conditions flip the method:

• The record itself was unusual. If last year contained a one-time event (a hailstorm that tripled roofing calls, a flood that did the same for restoration), then even with no driver change, last year is not the baseline. Strip the event out and rebuild from the year before or from the drivers. A great year looks like a new normal and usually is not. See related: Regression to the Mean and the Tech Who Had a Great Month.
• Nobody knows the drivers. For a new lead source or a new service, there is no "last year" worth anchoring to, and also no drivers you can count yet. Then the starting point is how similar things went for similar shops, not a rebuild from your own figures. See related: Base Rates: The Number You Should Start From.

How to check your budget for anchors

A quick pass over a draft budget or target sheet takes under an hour:

1. For each of the ten largest lines, write its drivers in one line next to it. If you cannot name the drivers, you cannot tell whether last year still describes it, and that line is a rebuild candidate by default.
2. Mark any driver that moved by more than a tenth. Headcount, truck count, a lead source's volume, a supplier's price list, a change in service mix.
3. Rebuild only the marked lines from their drivers, and write the rebuilt number next to the anchored one. Do not delete the anchored one; the gap between them is the useful part.
4. Read the gaps. A rebuilt number within a few percent of the anchored one means the anchor was fine. A gap of more than a tenth means a decision is hiding there, usually a hire, a cut, or a price change, and that decision deserves its own conversation rather than a budget line.

The point is not to distrust last year. It is to stop letting it answer questions it was never asked.

References

• Tversky, A. and Kahneman, D., "Judgment under Uncertainty: Heuristics and Biases," Science, 1974 (the wheel-of-fortune anchoring study)
• Epley, N. and Gilovich, T., work on insufficient adjustment from self-generated anchors
• Pyhrr, P., "Zero-Base Budgeting," Harvard Business Review, 1970
• See related: Anchoring: Present the High Option First; Regression to the Mean and the Tech Who Had a Great Month; Base Rates: The Number You Should Start From