Kanban Pull Improvement Case Analysis: A Manufacturing Company Solves Inventory and Delivery Challenges with "One Card"

1. Case Background: A Company Trapped by "Push Inertia"

A certain electronic component manufacturing company, primarily engaged in the assembly of power modules and control boards, has an annual revenue of about 800 million yuan. The company's production organization method is a typical "push" system: the planning department, based on sales forecasts and MRP calculations, issues a material input plan to all workshops once a week, and each process focuses on production according to the plan, pushing the completed products to the next process, and finally consolidating them into the finished goods warehouse.

On the surface, this system has been running for over a decade, with each link supported by a system. However, the management is well aware that problems are accumulating: the raw materials warehouse is expanding, work-in-progress (WIP) is piling up in the aisles, and the finished goods warehouse is filled with both slow-moving and out-of-stock items. By the time the project started, the company's inventory turnover days had risen to 68, the material completeness rate was hovering around 70%, and the on-time delivery rate for finished goods was only 82%. Even more frustrating was that the sales department frequently inserted orders and expedited materials, forcing the planning department to constantly re-plan, while the production floor was enveloped in a "firefighting" culture—wherever there was a shortage, it was addressed, and whoever shouted the loudest got priority.

What truly motivated the management to make a change was a customer audit. The customer, seeing a large number of semi-finished products stored for over three months in the warehouse, directly questioned, "If you can't manage your own inventory, how can we trust you to manage our orders?" This question stung everyone present. The general manager immediately decided to form a cross-departmental improvement team, starting with the most problematic final assembly line, to reduce inventory and improve delivery within a year.

The first step for the improvement team was not to design a solution but to quantify the problems. They analyzed three months of production and inventory data and discovered an abnormal phenomenon: the final assembly line spent two hours each day "searching for materials" before starting work. The line-side warehouse contained thousands of materials, but the few needed were often missing. Despite the large amount of stagnant material in the warehouse, the planning department continued to purchase according to forecasts. The data also showed that the average dwell time for WIP on the final assembly line was four days, while the actual processing time was less than two hours—materials spent over 95% of their time "waiting."

This discovery made the improvement team realize that the problem was not in the processing speed of a particular process but in the entire system being bound by "push inertia." Everyone knew the principle of "producing on demand," but years of operational inertia had made "overstocking, early production, and pushing out" the default behavior. To break this inertia, what was needed was not another round of motivational meetings but a mechanism that could enforce behavioral change. The improvement team turned their attention to the simplest and most effective tool in lean manufacturing—kanban.

2. Problem Diagnosis: Using Data to Understand "Why Inventory Can't Be Reduced"

Before making any changes, the improvement team conducted a three-week system diagnosis. They did not rush to draw conclusions but instead broke down the inventory issue into four dimensions for analysis.

Inventory Structure Analysis. The team categorized the inventory into "raw materials, WIP, and finished goods" and then classified the items using the ABC method. The results were surprising: A-class materials, which accounted for less than 15% of the total number of items, made up over 70% of the inventory value; nearly 40% of the WIP had not moved for over two weeks. These "zombie materials" indicated a severe mismatch between the material input rhythm and the consumption rhythm—planning was based on forecasts, but the actual structure of customer orders had changed.

Material Completeness Tracking. They tracked the material completeness of the final assembly line for two consecutive weeks, finding that the lowest completeness rate was 58%. After investigating the shortage list, they discovered that one-third of the "missing" materials were actually in the line-side warehouse but were buried under other materials; one-third had long procurement cycles and unstable consumption; the remaining were genuine forecast errors. This finding overturned the intuitive assumption that "shortages are due to poor procurement"—a significant portion of the "shortages" were caused by chaotic management.

Information Flow Analysis. The team drew a rough flowchart from order to material input, identifying that planning information had to pass through five stages: sales, planning, procurement, warehouse, and workshop. Each stage had time lags and discrepancies. The inventory data used by planners came from weekly stock counts, while actual consumption varied daily; material handlers often "grabbed materials nearby" for convenience, leading to increasing discrepancies between records and reality. Information distortion forced planners to increase safety stock, creating a vicious cycle of "the less accurate, the more stock, the more stock, the less accurate."

On-Site Observation. The improvement team spent 30 minutes each morning at the final assembly line, observing typical scenarios: material handlers replenishing dozens of boxes at once, placing them on the floor when there was no space; operators bypassing the process to call the warehouse directly when short of materials; the warehouse prioritizing lines with the loudest voices, with delivery order depending on "who shouts the loudest."

The diagnosis concluded with three key points: first, push-based material input caused a disconnect between inventory and actual consumption, leading to a severely imbalanced inventory structure; second, information distortion amplified fluctuations, forcing planners to increase inventory to hedge against uncertainty; third, the production floor lacked a "consumption-triggered replenishment" mechanism, relying on people, voices, and relationships for replenishment. Based on these conclusions, the team decided to pilot the final assembly line, using kanban to pull and rebuild the "consumption-replenishment" order, starting with the line-side inventory, which was the easiest to see results.

3. Improvement Plan: Pilot First, Calculate Kanban, Then Establish Rules

The improvement team did not roll out the plan company-wide from the start but chose the final assembly line, where problems were most concentrated and improvement willingness was strongest, as the pilot. The pilot scope was clearly defined as the final assembly line and its direct upstream electronic material warehouse and semi-finished product storage area, involving about 400 types of materials. The plan design followed a "three-step approach."

Step One: Define the Pull Boundary. The team categorized the materials for the final assembly line into three types: high-frequency stable consumption items were included in the kanban pull; long-cycle imported items and custom items maintained planned procurement but with a safety stock red line; low-frequency sporadic items continued to be requisitioned on demand. The principle was simple—kanban was only used for "repetitive, stable, and predictable" flows. Setting the boundary too wide would make the rules unmanageable due to complex material characteristics; setting it too narrow would reduce the improvement effect.

Step Two: Accurately Calculate Kanban Numbers. This was the most critical and error-prone step in the pilot. The formula for calculating the number of kanban cards is straightforward: the number of kanban cards equals the daily consumption rate multiplied by the replenishment lead time and the safety factor, divided by the number of items per box. However, accurately calculating each parameter required on-site data: the daily consumption rate should be the average of the actual consumption over the past three months, not the planned amount; the replenishment lead time should be measured from the moment an "empty kanban card is issued" to when the material is delivered to the line, including each step of information transmission, picking, delivery, and acceptance; the safety factor should be determined based on the consumption fluctuation, with higher values for more volatile materials and lower values for stable ones.

Initially, the team's calculated kanban numbers were generally too high—safety factors were habitually set at 30%. They tested the calculated numbers on the pilot line for a week, observing whether the supermarket inventory remained consistently low. They found that a safety factor of 10% to 15% was sufficient for most materials, with only a few seasonal materials requiring higher buffers. After two rounds of "calculation—testing—correction," the kanban numbers for the 400 materials were finalized, and the theoretical upper limit of line-side inventory decreased by about 40% compared to before the improvement.

Step Three: Establish the Supermarket and Rules. A fixed line-side supermarket area was designated next to the final assembly line, with each storage location defined by material type, capacity, and quantity. Labels were used to indicate the material number, the number of items per box, and the number of kanban cards. Three ironclad rules were established: no replenishment without a kanban card, material handlers only recognize kanban cards and not phone calls; empty kanban cards must be returned on the same day, with no accumulation allowed; quality abnormal materials must be immediately removed from the supermarket channel and only returned after being processed by the quality department. These rules were incorporated into the job responsibilities of material handlers and operators, with team leaders checking compliance weekly.

After the plan was designed, the improvement team spent a week training all relevant personnel: material handlers learned to replenish according to the kanban cycle, operators learned to place kanban cards when materials were consumed, and planners learned not to interfere with material input in the pilot line. The pilot officially switched on the following Monday.

4. Implementation Challenges: Real Difficulties and Solutions

No improvement is smooth sailing, and this project was no exception. In the first two months after the switch, the improvement team faced three significant challenges, each forcing them to refine their plan.

First Challenge: Kanban Numbers "Calculated Correctly" but "Ran Out of Control." In the second week after the switch, the supermarket inventory began to lose control, with some materials being replenished more than consumed. The team rechecked the data and found no errors, but on-site observation revealed the truth: material handlers were accustomed to "bringing a few extra boxes," and they added extra quantities when replenishing according to kanban; some operators, fearing material shortages, held onto kanban cards and released them all at once near the end of their shift, causing a "pulsed" surge in replenishment signals.

The solution was two-pronged: installing physical barriers at the supermarket storage locations to prevent over-replenishment, and incorporating the rule "kanban cards must be placed and returned on the same day" into the daily inspection of the team. After two weeks, the supermarket inventory returned to a controlled state.

Second Challenge: Discrepancies in Data Metrics. After the pilot showed results, the finance department raised a question at the monthly business analysis meeting: why had the inventory turnover days not significantly decreased? The improvement team discovered that the discrepancy was due to different data metrics—finance calculated total inventory based on book value, while the team calculated line-side inventory based on physical quantities. The improvements in the pilot line were offset by inventory increases in other areas. This dispute almost derailed the project, as management began to doubt whether the improvement was just a "numbers game."

The solution was to standardize the metrics. The improvement team and finance jointly defined three consistent indicators: line-side inventory value, material completeness rate, and on-time delivery rate, and agreed to use these as the sole evaluation criteria for the project. They also included the pilot line's inventory in an independent monthly stock count, using physical data to speak. After standardizing the metrics, the improvement results became clear, and the dispute naturally subsided.

Third Challenge: Quality Abnormal Materials Returning to the Line. During an incoming quality control (IQC) inspection, a batch of electronic materials was found to have batch appearance defects, and the quality department, following the rules, removed the materials from the supermarket. However, the production line's consumption quickly approached the safety line, and under pressure, the material handler bypassed the process and directly took materials from the abnormal area to the line. The quality department immediately stopped the operation, and the two departments argued in the workshop.

This incident made the improvement team realize that the weak link in kanban pull was not material flow but abnormal handling. They subsequently established an "abnormal kanban" mechanism—quality abnormal materials were marked with yellow abnormal kanban cards, and quality engineers were required to provide a disposition within 24 hours. Material handlers were also authorized: any material removed from the supermarket was considered "unavailable" and could not be taken without going through the abnormal channel. Clear rules reduced conflicts, and no instances of abnormal materials being used without disposition occurred in the following three months.

5. Effect Verification: Speaking with Numbers

Six months after the pilot began, the improvement team conducted a comprehensive effectiveness assessment: the line-side inventory value of the final assembly line decreased by 52%, the inventory turnover days dropped from 68 to 31; the material completeness rate improved from less than 70% to 94%; the on-time delivery rate increased from 82% to 96%. The line-side warehouse freed up over 300 square meters of space. More importantly, the behavior patterns on the production floor changed: material handlers no longer ran around the workshop with carts but followed fixed routes for cyclic distribution; operators no longer called the warehouse directly when short of materials but placed kanban cards in the return box and continued their work; the warehouse's record-physical discrepancy rate decreased from over 8% to within 2%. As the final assembly workshop supervisor put it, "Previously, people were chasing materials; now, materials follow the signals."

After the pilot's success, the improvement team gradually extended the same method to the insertion line, packaging line, and raw material warehouse, adapting it to the characteristics of each area: standard kanban cycles for processes with large batches and few changeovers; increased safety stock coefficients for processes with high volatility; and external kanban introduced at the supplier end to directly transmit replenishment signals to key suppliers. By the end of the project, the factory's inventory turnover days stabilized at 35, and annual cash flow improved by over 30 million yuan.

6. Insights from the Case: Key Success Factors of Pull Improvement

The most valuable aspect of this case is not the kanban calculation formula but the simple rules it reveals—any company looking to replicate this method can draw inspiration from them.

First, the Starting Point of Improvement is Diagnosis, Not Tools. This company initially wanted to implement an electronic kanban system, but the diagnosis revealed that the root cause of the problems was push-based material input and information distortion. No matter how advanced the tool, it cannot change the incorrect operational logic. First, understand the problem, then decide on the tool, and do not reverse the order.

Second, the Pilot is the Source of Confidence and the Testing Ground for Rules. Start with one line to establish and test the rules, and use visible results to gain support from various departments. The material range, indicator metrics, and abnormal handling rules must be refined during the pilot phase and then replicated when mature.

Third, Rules Must Be "Physicalized," Not Just Rely on Self-Discipline. Physical barriers, kanban return boxes, and abnormal kanban cards, though seemingly cumbersome, are essential to ensure that rules are not violated. Turning rules into physical constraints and daily inspections helps good habits become stable.

Fourth, Standardized Metrics Ensure Visible Improvement. This project almost failed due to discrepancies in data metrics. From the first day of improvement, align the indicator definitions with finance and operations to make the results visible in management reports. Otherwise, even the best improvements cannot withstand data scrutiny.

Fifth, Pulling is Not the Goal, Exposing Problems Is. After the kanban reduced inventory, the real issues that were previously hidden by inventory surfaced: supplier quality fluctuations, unstable equipment, and inconsistent work methods. Following the inventory reduction, the company initiated specialized improvements for bottleneck processes and quality guidance for suppliers—kanban systems allowed them to see the full scope of the problems for the first time. This is the profound intention of kanban: it is not to hide inventory but to expose problems, forcing the organization to address them one by one.

For companies hesitating between push and pull systems, this case provides a simple answer: start with one line, establish rules with kanban, verify the effects with data, and drive improvements through abnormalities. When "consumption-triggered replenishment" becomes an instinct on the production floor, inventory, delivery, and quality issues will be resolved on a more solid foundation.

The essence of kanban pull is not to reduce inventory but to expose problems—using one card to rebuild the "consumption-replenishment" order and teach the organization to flow on demand.

Knowledge code: 7.1.3

Version: v20260802

Author: Quality Think Tank Quality Think Tank is dedicated to providing systematic professional knowledge, methodologies, and practical tools to quality management practitioners, helping companies continuously improve their quality capabilities.