Why Are We Still Relying on Excel After Replacing Three Scheduling Systems? — Three Structural Prerequisites for Digital Planning

After replacing three scheduling systems, the planner’s desk is still running Excel. This is not because the software was chosen incorrectly.

Failed projects in digital scheduling share a common characteristic: there is a "plan" in the system, but the workshop runs the Excel version. The system becomes an archiving tool, while Excel becomes the decision-making tool.

The reasons are often diagnosed as "lack of user cooperation," "the system is not user-friendly," and "inaccurate data." All these reasons are valid, but they are not the root cause. The root cause is that the three structural prerequisites for digital scheduling have not been met—if these prerequisites are not met, the fourth system will yield the same result.

Prerequisite One: Accuracy of Process Routes and BOM

The first task of a scheduling system is to "calculate how long a process will take and what materials are needed." This depends on two primary data sets:

• Process Flow: sequence of operations, equipment groups, standard labor hours
• BOM: materials consumed and quantities required for each operation

Criterion: the proportion of process routes that do not match the actual process should be ≤ 2%.

How to measure this? Sample 50 work orders in progress, compare the "process routes in the system" with the "actual reported sequence of operations," and calculate the inconsistency rate. If it exceeds 5%, the output of any scheduling system is unreliable—it calculates a different world.

The three most common distortions are:

The cost of fixing this step is severely underestimated. A medium-sized factory typically needs 2～4 months to calibrate process routes and labor hours to a usable level, involving collaboration between process, production, and IE departments. Many companies are unwilling to spend these months, opting to implement the system directly and then spend two years manually adjusting it within the system.

Prerequisite Two: Reliable Capacity Model

The core constraint in scheduling is capacity. The capacity model must answer four questions:

1. How long can the equipment be used: available labor hours = calendar time − planned downtime − historical failure rate
2. How long does a changeover take: the changeover matrix for different products and specifications, not an average value
3. What is the first-time yield: yield determines the amount of material to be input and the machine hours required
4. Where are the personnel constraints: which equipment requires certified personnel, which operations are limited by team size

Criterion: the deviation between the system-generated schedule and actual execution should be ≤ 10%.

How is this deviation calculated? Compare the "system-scheduled completion time" with the "actual completion time" and calculate the median deviation by operation. If the deviation exceeds 20%, it indicates that the capacity model has become detached from reality—planners will naturally use Excel because only Excel can accommodate the information they know but the system does not.

A specific counterexample: a company uniformly fills in 30 minutes for changeover time, but the actual time ranges from 30 minutes to 4 hours (depending on the specification differences between the two products). As a result, the system's schedule is always overly optimistic for high-changeover products, and the planner's first task every morning is to reschedule—naturally, this is done in Excel.

Prerequisite Three: Comprehensive Constraints and Rules Before Algorithms

This is the easiest prerequisite to skip and the most fatal.

Scheduling is not just about "finding the optimal solution"; it is a series of judgments:

• Three orders are competing for the same equipment, which one goes first?
• An urgent order is inserted, how does the original plan adjust?
• Materials are not complete, should we wait for the materials or start with semi-finished products?
• Customer priority, promised delivery date, and profit contribution, which one takes precedence?

These judgments are currently in the planner’s mind and in Excel. If the rules are not written down, the system can only schedule according to default logic, and the results will not align with the planner's judgment, leading to manual adjustments—always in Excel.

Criterion: for any scheduling decision, the planner can explain "why it is scheduled this way." If they cannot explain clearly or say "based on experience," the rule has not been formalized and should not be entered into the system.

The recommended sequence for solidifying rules is:

Most companies fail by skipping ① and going straight to ③. No matter how advanced the algorithm, if the rules fed into it are empty, the results will still rely on human intervention.

Why Excel Persists: It Carries the Decision-Making Authority

When these three prerequisites are considered together, the conclusion is clear:

Excel is not used because it is convenient; it is used because it accommodates "all the rules that have not yet been written down."

Every cell the planner modifies in Excel is a rule that has not been entered into the system. Therefore, the essence of digital scheduling is not "implementing a scheduling software," but extracting the decision-making logic from the planner’s mind, writing it into rules, and handing it over to the system. This is a management engineering task, not an IT engineering task.

This also explains why IT-led scheduling projects have a low success rate—IT can install the software, but it cannot extract the rules. The rules must be written by operational and production planning personnel themselves.

Four Key Indicators to Watch

To determine if digital scheduling has truly been implemented, watch these four indicators:

If, six months after the system goes live, the "manual adjustment rate of plans" is still above 50%, it indicates that at least one of the three prerequisites is missing. At this point, the system should not be replaced; instead, the missing prerequisites should be addressed.

In a Nutshell

The correct sequence for digital scheduling is: first, calibrate the primary data, then build a reliable capacity model, next, write the decision-making rules into understandable tables, and finally, discuss algorithms and systems. Projects that reverse this sequence will end up with a "system with a plan and an Excel sheet in the workshop"—and typically, it takes until the second system replacement for this to be acknowledged.

This article is part of the "Digitalization and AI Zone · Operational Digitalization" series. Other articles in this direction include "The Seven Breakpoints from Order to Delivery" and "Why No One Looks at the Business Dashboard," which can be found in the zone list.

Knowledge code: 16.2.2 Author: QTank Version: v20261026