A target nobody believes doesn’t function as a target. It’s wallpaper. Operators clock in, glance at it, and mentally file it under “the number that’s always wrong,” and once that happens, the target stops doing the one job it had, telling people whether the shift is going well. Most bad targets aren’t bad because anyone was careless. They’re bad because they were set as one plant-wide guess instead of built from what each specific job actually requires.
Where a bad target usually comes from
A shift target is often set once, years ago, by someone estimating “what this machine should be able to do,” rounded to a clean number, and never revisited. That number gets treated as gospel regardless of which product is actually running today, even though different products on the same machine can have wildly different real cycle times. A single flat target painted on a whiteboard flatters some jobs and punishes others, and the operators running the punished job every day learn fast that the number has nothing to do with them.
The fix is skipping the guess entirely. A target built from the routing, the actual documented cycle time for this specific product on this specific machine, changes with the job instead of pretending every job is the same one.
Building a target from the routing
A routing already breaks a product down into its operations, each one pinned to a specific machine, with a cycle time and a setup time attached. That’s the target, already sitting in the system that governs how the job gets planned and costed, not a separate number someone has to remember to maintain in a second place. When a job for that product runs on that machine, the target on the floor should be whatever the routing says for this operation on this machine, not a general house number that happens to be close for some jobs and wildly off for others.
This matters even more on a machine that swaps between products within a single shift. A window that spans a changeover shouldn’t grade the whole stretch against one blended average, because that smears a fast product’s speed onto a slow one and produces a performance number that’s accurate for neither. Grading each sub-interval against its own product’s target, most specific first, this machine’s own routing step, then the work center’s general step, then the product standard, keeps a changeover from distorting the number on either side of it without anyone catching it.
The mold time trap
Injection molding shops hit a specific version of this problem often enough that it’s worth naming directly. A mold has a rated cycle time from the toolmaker or the mold’s own spec sheet, and it’s tempting to treat that number as the target the way a nameplate rate gets treated as a target elsewhere. The trap is that a mold’s rated cycle time is usually measured under close to ideal conditions, and a single mold can run at meaningfully different real speeds depending on the press it’s mounted in, the material lot, the cooling setup, and how well-maintained the mold itself currently is.
A rated mold time used as a blanket target across every press that mold ever runs on produces the same problem as a nameplate rate used as a performance benchmark: a number that’s frequently wrong in one direction or the other, trusted anyway because it came from somewhere official-sounding. The fix is the same discipline as anywhere else, use the mold’s actual demonstrated cycle time on this specific press as the working target, and treat the toolmaker’s rated number as a reference point for what might be achievable eventually, not as the number graded against today.
Why belief matters more than accuracy on its own
A target can be numerically defensible and still fail if nobody on the floor believes it. Belief is what actually changes behavior, a target people trust gets people adjusting pace, flagging a problem early, asking for help before a shift ends behind, because the number feels like a fair description of what’s expected. A target people have already written off as wrong gets ignored entirely, even on days it happens to be accurate, because trust, once lost on a number, doesn’t come back just because that particular instance was correct.
That’s the real cost of a bad target, and it’s bigger than the immediate confusion it causes. Once operators decide the target painted on the board doesn’t reflect reality, they stop using it as a signal at all, and a plant loses the single fastest piece of feedback a shift has for knowing, in real time, whether it’s on track. Getting the target right isn’t a fairness exercise for its own sake. It’s restoring a working instrument that a bad number had slowly turned into noise.
Revising a target without starting a fight
Changing a target that’s been in place a while is politically touchy, because it feels like an accusation, either the old target was wrong and someone should have caught it sooner, or the new target is a stealth speed-up dressed up as a data correction. Both readings turn a data conversation into a trust conversation, and that’s usually where a good revision gets stuck.
The way through is grounding the change in something nobody in the room disputes: what the machine has actually demonstrated it can do, not a percentage bump imposed from above. A revised target that’s set to match a documented cycle time already observed on the routing, or a machine’s own recent sustained pace, is a fact, not a demand. It’s much harder to argue with “the machine already ran this fast for two hours last Tuesday, that’s the number” than with “corporate thinks you should be faster.” Bring the receipt, not the opinion, and the conversation shrinks from a negotiation into a fact check.
Revisit targets on a set schedule, quarterly is reasonable for most shops, instead of only when someone notices a number looks wrong. A scheduled review feels like maintenance. A surprise review feels like scrutiny, and people respond to those very differently even when the underlying change is identical.
Quick recap
- A flat plant-wide target fails the moment two products on the same machine have real cycle times that differ by a wide margin.
- Build the target from the routing, the actual documented cycle time for this product on this machine, not a house number nobody’s revisited in years.
- Grade a changeover window by each sub-interval’s own product target, not one blended average that smears a fast job’s speed onto a slow one.
- A mold’s rated cycle time is a spec sheet claim, not a target. Use the mold’s actual demonstrated speed on the specific press it’s running on instead.
- Revise a target by pointing at what the machine already demonstrably did, not a percentage handed down from above.
- Review targets on a set schedule. A scheduled check feels like upkeep. A surprise change feels like scrutiny.