First, the conclusion: an LMS warehouse rollout can start without engineered standards. Run a four-week observed-rate baseline first, then decide whether engineered standards are worth building. Waiting until the standards are complete before switching the system on is the most common delay I see in small warehouses. Standards take six months to measure, the system slips six months, and when peak season arrives the supervisor is still scheduling by gut feel.
Here is why the wait is a mistake. Engineered standards are precision work: breaking jobs into motions, timing them, adding allowances, and reviewing the results. A small warehouse without a dedicated analyst and without that budget cannot carry this honestly. The measured numbers come out, nobody on the floor believes them, and execution stalls. I have seen this exact failure three times. The script never changes: consultants measure for two weeks, hand over a standards manual, the shift lead flips through two pages and leaves it in a cabinet, and the standards in the system describe a different warehouse from the one on the floor. The standards are not useless. The warehouse simply is not ready for them yet.
An observed-rate baseline takes the opposite route. It does not argue about how much a worker should theoretically do. It records how much the worker actually did, for four straight weeks, and takes a middle value everyone can accept. That number is imperfect, and it gives you two things engineered standards cannot give you at the start: it comes from your own floor, so a shift lead cannot dismiss it as theory; and it works for scheduling, headcount math, and exception spotting from week one. Get the system moving first, then make the standards precise. That order is what makes small-warehouse projects survive.
Last year I ran this play in an e-commerce warehouse in eastern China. The picking team ran to more than twenty people, peak season survived on overtime, and the supervisor scheduled entirely from memory. Some days the floor was short-handed, other days workers stood idle. In week one of the system we changed nothing. We only connected task data and attendance data and watched actual output per person and per stage, every day. After four weeks, a middle value for picking emerged, and the swings in the checking stage became visible. For the first time, the supervisor could point at data for Tuesday afternoon headcount instead of guessing two extra hands were required. Overtime fell back in that same month, not because workers got faster, but because scheduling stopped being a guess.
A second warehouse, an auto-parts site in northern China, showed the reverse problem. Its shift lead was strong, and experience-based scheduling had never caused a major failure, so the owner saw no reason for a system. After the four-week pilot, the data exposed something experience could not see: the morning wave and the afternoon wave, same people doing the same work, produced visibly different output. Digging in, we found afternoon replenishment was not keeping up, and pickers spent much of their time waiting for stock. Experience can build a roster. It cannot show cross-stage waiting. A system is not there to watch workers. It is there to make that waiting visible.
LMS warehouse baseline: four weeks of observed rates
Run the four-week pilot on the checklist below. No week is optional.
- Week one: connect data only, no evaluation. Connect task completions, attendance, and stage timestamps. Tell the crew plainly that this week's data will not feed performance reviews, only the baseline. If that message is unclear, workers will start cherry-picking jobs, and the data will be worthless on arrival.
- Week two: produce a middle value per stage. Picking, checking, packing, and putaway each get their own number. One blended average across stages guides nobody, because a fast stage will hide a slow one.
- Week three: schedule against the baseline. The supervisor builds rosters from the baseline numbers, compares them with what actually happened, and writes down every large gap. Roster error is the most valuable feedback a baseline gives, better than any report.
- Week four: set a trial target. The baseline is not the target. The target sits one step above the baseline, set jointly by the supervisor and the crew, then posted in the open. Review exceptions daily, asking only three questions: is the data accurate, where is the stage stuck, and what changes tomorrow.
This LMS warehouse pilot does only three things in phase one: record counts, compute the baseline, and help the supervisor roster. No performance linkage, no automatic dispatching, no complex incentive rules. Fewer functions mean the pilot is more likely to survive. A pilot that survives earns the right to discuss phase two.
| Item | Buying on day one | What to use instead |
|---|---|---|
| Engineered standards module | Expensive, never fully measured, rejected on the floor | Four-week observed baseline plus supervisor review |
| Automatic dispatching | Dispatch logic depends on standards; wrong standards mean wrong dispatch | Supervisor rosters by hand from the baseline; the system only suggests headcount |
| Performance incentive module | Link pay to an unstable baseline and workers start gaming the numbers | Publish the data with no pay link until the baseline holds steady |
| Exception dashboard | Worth buying; the most valuable pilot feature | — |
Pilot mechanics: accurate data beats a full feature list
A pilot lives or dies on data accuracy, not on feature count. Timestamps must come from the system's task records, never from workers typing their own hours. Hand-typed hours shrink on busy days and stretch on quiet ones. I once worked with a warehouse whose handwritten timesheets looked like a textbook, and the moment we connected system timestamps, three stages of data changed shape completely. If the source is wrong, every analysis downstream is wasted effort.
One more detail: do not change the process during the pilot. A changed process destroys the baseline's reference point. Run the old process for four weeks, measure its real level, and only then discuss improvement. Many projects die because they try to launch a system and redesign the process at the same time. Two variables moving together means nobody can say which one caused a problem.
When an LMS warehouse graduates to engineered standards
When two of the three signals below appear, engineered standards are worth discussing. Signal one: the baseline holds steady for two straight months, swings narrow, and the supervisor already rosters from data by habit. Signal two: crews start asking why one stage always runs slower than baseline, which means they treat data as a management tool rather than a monitor. Signal three: a gap the baseline cannot explain shows up, such as two teams on the same stage and the same product running apart for weeks, and that gap needs motion-level breakdown to explain.
Graduation itself needs care. Do not switch the whole warehouse at once. Pick one stage, such as picking, and run the engineered standard in parallel with the observed baseline for two weeks. Switch only where the two agree, and investigate where they disagree. For how to build standards in depth, see the engineered standards approach. Parallel running is the easiest step to skip and the least skippable step in the whole project.
Failure points: what not to buy on day one
Walk through this list at the pilot kickoff meeting, item by item.
- Borrowing another warehouse's standards outright. Product mix, location layout, and wave rhythm all differ, so another site's numbers are wrong numbers at your site.
- Linking pay to the baseline the moment it appears. The baseline is a starting point, not a passing grade. Link pay too early, and workers will optimize the number instead of the work.
- Letting only the IT team push the pilot while the operations lead watches. The system is the supervisor's management tool. If the supervisor does not use it, the pilot is decoration.
- Judging the pilot before four weeks are done. Week one data is mostly noise, week two starts to take shape, and an early verdict means deciding from noise.
One counter-intuitive point belongs up front: an observed baseline measures existing waste along with existing output, so a slack stage gets a low baseline. That is not a flaw in the baseline. It is its honesty. The baseline says where you are, the target says where you are going, and keeping the two separate is what gives management a grip. Treating the baseline as a target to pass is the same as certifying waste.
Small warehouses should not wait for everything to be ready. Run four weeks and argue from your own numbers. Once the numbers exist, rostering, overtime, and stage bottlenecks all have a starting point for discussion. Without those four weeks, every argument with a shift lead is about feeling. With them, the argument is about data.
By Lin Xiao, WMS Implementation Consultant








