Once you know which process to fix and what's in the way, the temptation is to roll the fix out everywhere. Don't. Run it small first.

Running it small is the R in WORTH. It's how you prove a change is worth doing before you spend the effort making it permanent.

What "small" means

Small means one change, on real work, for a short, defined stretch. Not a pilot program with a steering committee. A handful of real cases, run the new way, by the people who actually do the work.

A good rule of thumb is around twenty real cases. Few enough to finish quickly. Enough that you'll hit the awkward ones, not just the easy ones.

Use Tuesday's messy cases

There's a difference between "did we build it right?" and "did we build the right thing?" A demo on clean data answers the first. Only a person running Tuesday's messy cases answers the second.

So pick real work, including the half-filled forms, the customer who describes the problem in a way nobody planned for, and the job that changes midway. If the new way only works on the easy ones, you've learned something important before it cost you anything.

Exhibit 1

Speeding up the open road just gets you to the merge sooner

Work piles up where the road narrows. Widen the merge and everything behind it moves.

The open road: repave it or add a lane, and traffic just reaches the merge sooner.

The merge: where work waits.

Widen the merge and everything behind it moves.

Source: Earn the Change, Troy Green

Exhibit 2

Only the owner's review passes the twice-as-fast test

If this step ran twice as fast, would the result arrive sooner? Ask it of every step.

Customer calls
Site visit

Within two days

Twice as fast?

The quote still waits

Not the obstacle

Owner review

Waits four to five days

Twice as fast?

Quotes go out days sooner

The constraint

Quote goes out

Obstacle statement

Quotes wait four to five days for the owner's review because every standard job is priced from scratch.

Source: Earn the Change, Troy Green

Three numbers, no more

Before you start, write down three numbers:

  • Prove it. The one number that should move if the change is right. Where does it start, what's the target, and by when?
  • Protect. The one thing you can't afford to break while you change something else. Quality, customer satisfaction, error rate, time spent checking.
  • Show it stuck. The number you'll check later, after the change becomes routine, to make sure it didn't quietly slide back.

Three numbers keeps everyone honest. More than three and nobody watches any of them.

The protect number is the one people forget

Most changes that fail don't fail on their main goal. They succeed at it while quietly breaking something else. Quotes go out faster, but more of them have errors. Calls get answered sooner, but customers have to call back. The protect number catches that before it becomes the new normal.

Decide what happens next

At the end of the small run, you have one of three answers:

  1. It worked. The prove-it number moved and the protect number held. Move to trade-offs and habit.
  2. It didn't work. That's not failure. You just avoided rolling out something that wouldn't have helped. Go back to the obstacle and try a different fix.
  3. It's unclear. Run it a little longer, or a little differently, before deciding.

Each of those is a good outcome, because each one is based on evidence instead of opinion.

Start with one week

Pick your first process. Pick the one change you want to try. Write down your three numbers. Then run it on the next twenty real cases and see what happens.