Operational Improvement for Back-End Factories
Challenges specific to assembly and test facilities, and ways to mitigate them.
By Jennifer Robinson
In this article, we turn our long-time focus on factory improvement to the operational challenges of semiconductor assembly and test facilities. Some of these challenges, such as managing variability, are common to all factories. Others, such as long, sequence-dependent setups, are more specific to back-end manufacturing.
In contrast with wafer fab cycle times, assembly and test cycle times are shorter, measured in a few days to a week rather than weeks or months. However, this reduced process complexity usually comes with a dramatic increase in factory scale.
In this article, we will review the assembly and test process flow, discuss back-end-specific operational challenges, and suggest ways to mitigate them.
The post-fab process
There is variation in the post-fab process, particularly between what happens at more manual legacy assembly and test factories and newer, advanced packaging facilities. Here we broadly describe a typical legacy production flow.
Wafer Sort/Probe: This is when each die is electrically tested (probed) to sort good from bad while still part of the wafer. A wafer map indicates the good vs. bad die before the wafer is shipped from the fab to an assembly and test facility.
Die Prep: This stage includes backgrind to make the wafer thinner, putting the wafer on tape, and then saw, laser, or plasma dicing to cut the wafer into die. Each die will be on the tape in its original mapped position and ready for die bond.
Die Attach / Die Bond: At die bond, each good die is picked off and attached to a leadframe, substrate or carrier with epoxy or solder. The frames are grouped temporarily into magazines. Sometimes, with epoxy bonds, there is a bake step to cure the epoxy before wire bond. This cure step takes place in a batch oven.
Wire Bond: At wire bond, each die is electrically connected to the frame via gold, copper, or aluminum wire. A device may require anywhere from one to thousands of wires. Wire bonders can make many bonds per second. A flip-chip process is sometimes used instead of wire bond. The die is flipped over and connected via solder bumps.
Molding / Encapsulation: To protect the package, the assembly is then encapsulated in epoxy mold. Many packages can be molded together as a strip or matrix array. The mold is cured in an oven afterward.
Marking, Ball Attach and Singulation: Each part is laser-marked with identification. Some packages get solder balls attached. The strip or matrix is separated into individual units.
Final test: The individual units are electrically tested, often at multiple temperatures in a handler that presses them against test sockets. Some devices also go through burn-in, a long high-temperature stress test done in batch ovens.
Pack and ship: Good units are binned by speed or grade, placed into trays or tape-and-reel, moisture-protected (dry pack for moisture-sensitive parts), and shipped.
Inspection and reliability sampling is done throughout the process and may lead to scrap or rework

Things about this process that are more straightforward to manage vs. the front-end fab
Three things about the above process are helpful in terms of achieving reasonable cycle times from the back-end.
- The linear process flow does not include any reentrant loops (except for possible rework). This reduces variability and helps reduce scheduling complexity.
- The relatively short total cycle time (typically not more than a week, and often much shorter) and small number of steps also reduce scheduling complexity.
- There are typically no single path tools in assembly and test, which means that a single down tool will not bring the line to a halt.
There are, however, several challenges to manage in assembly and test factories.
Operational challenges in back-end manufacturing
Setups, scale (high WIP levels and many like tools), batching, lack of WIP linearity, manual processing (for legacy assembly and test sites), and other sources of variability all lend complexity to the back-end process.
Setups: One of the biggest operational challenges in back-end manufacturing is tool changeovers. Setups can be many hours to as much as a day long and usually require manual intervention by operators. Longer, more complex setups may require dedicated maintenance technicians. Operations like die bond and wire bond can have ten or more different physical components that need to be swapped in or out by a technician to run a subsequent lot. Changeovers of testers often require a set of durables to be available before starting the changeover. The management of these physical items adds to overall complexity. The cost profile and capacity to perform these different types of changeovers can also vary significantly, making the problem of setup optimization a notable contributor to factory profitability. Some facilities are labor constrained in terms of staff who can do setups. High levels of product mix and frequent mix changes also compound the setup challenge.
Scale/Complexity: Although assembly and test process flows are not long, the factories can be challenging to manage due to sheer scale. WIP levels may be high, and the number of tools of a given tool type can, in some cases, exceed 1000. While this abundance of like tools is better for cycle time than having a series of one-of-a-kind tools, qualification rules can become quite complex. A single lot might have hundreds of qualified tools at a given step, leading to an enormous number of candidate lot to tool assignments for any decision. Setup minimization, tool downtimes, and varying lot priorities also need to be considered.
Batching: There are several steps in the assembly and test process flow in which units are batched together for processing: batch ovens can be found in epoxy cure, encapsulation, and, notably, final test (burn-in). The arrival process downstream of this batching shows elevated levels of variability. This is especially a problem where the arrival variability from epoxy cure leads to starvation of wire bond tools. This batch processing is on top of other form changes in assembly and test, as when die are grouped into leadframes and magazines, combined into assemblies and arrays, and later separated into individual units. Binning at different stages can also lead to smaller lot sizes. All of this adds variability and complexity to the process.
Not batching exactly, but there are also cases where processing of Material A at a given step is dependent on having Material B available at that step. An example is multichip module die bond, where multiple die bond operations take place on subsequent tools to build a leadframe. Die from Material A are bonded to the leadframe in one tool at the first die bond step, then from Material B in a different tool at the second step. (Sometimes there are additional steps.) This material dependency means that to continue processing at the second step, Material B must be available on time. This adds to schedule complexity.
Lack of WIP Linearity: Although process flows in assembly and test are not reentrant, different throughput rates at different production stages, together with grouping and ungrouping of units, result in a non-linear WIP profile. Decisions of how to group material at one stage influence the required setups at the next. When not well managed, this lack of WIP linearity can lead to starved constraint tools. Throughput imbalances between handlers and testers can exacerbate this issue.
This situation is compounded by high product mix and by non-linear arrivals from wafer fab / sort into the back-end facility. It can be challenging for assembly and test managers to respond to unpredictable arrival patterns from the fab.
Constraint Management: Because assembly and test process flows are shorter, the focus of back-end KPIs tends to be on utilization of constraint tools, typically wire bond tools and testers. Die bond can operate several times faster than wire bond, requiring decisions to be made at die bond to avoid starved tools or unnecessary setups at wire bond. Testers can similarly starve due to upstream delays from downtime or setups, leading to capacity loss. Tester downtime, throughput limits, or chamber temperature limitations can also slow product flow.
Manual Processing / Labor Inefficiency for Legacy Factories: Unlike newer advanced packaging facilities, legacy assembly and test factories can be people-constrained, especially in performing setups. Man-to-machine ratio is often used by supervisors as a productivity index, with a goal (for cost reduction) of having fewer operators per tool. This can run counter to operational improvement, as expensive tools end up with forced idle time while waiting for operators. Productivity losses can also occur during shift change and breaks, when all operators leave at the same time. Short processing times make it difficult to keep tools running during the break. Compliance can also be a challenge when implementing modern scheduling solutions in previously operator-managed facilities, especially when a new schedule increases setups.
Back-end tools also have a much higher number of “assist” events than front-end tools. This is especially true in wire bond and die attach. The tools may experience have short pauses requiring operator intervention. While operators can normally resolve these issues quickly:
- Frequent assists can cause actual tool throughput rates to diverge significantly from theoretical throughput rates, leading to missed shipment targets
- These assists can drive down man-to-machine ratio
- Due to assists and other issues, tools often run below their target speeds. Where this speed loss goes unrecognized, overall performance can suffer (e.g. many pattern recognition repeats may be required before success). In general, legacy facilities may operate tools that are not optimally maintained or adjusted
Other Sources of Variability: Of course, other sources of variability compound the back-end manufacturing challenge. These include:
- Yield losses: Failures may require rework, scrap, or re-test cycles, which inflate overall cycle time and can complicate scheduling.
- Test time variability: Some products take longer to test than planned. Some test times can be unpredictable.
- Downtime: With the high number of redundant tools in a back-end factory, a single down tool doesn’t take down the entire line. However, equipment downtime still poses a challenge. Downtime is a capacity loss on high-utilization tools and can drive the need for extra setups.
- Bowl feeders: Small, non-wafer parts (pins, clips, lids) are sometimes delivered to assembly or test stations via bowl feeders. If the bowl feeder is not working well, it may feed parts more slowly than the capacity of the machine, causing throughput losses. When the bowl feeder is running close to empty, the last parts may require many iterations in the bowl before they finally make it into the machine in the correct orientation. The machine will have forced idle time in the meantime.
Meeting Expected Shipment Dates / OTD: All the variability described above means that assembly and test facilities, like wafer fabs, can experience challenges with meeting expected shipment dates. Although the process flows are shorter than those in the fab, the sheer volume of WIP to be managed at one time can make meeting the due date for a particular unit challenging, particularly in the presence of setups and equipment downtime. Starts planning and setup planning are both critical in meeting shipment dates.
Lack of Data Visibility: Legacy assembly and test factories may lack the data visibility needed to identify improvement opportunities. Even where factories implement modern schedulers, good data is necessary to see gaps in compliance. Minimizing future setups may require analyzing historical setup data. Similarly, knowing where forced idle time on bottleneck tools has occurred in the past can help to avoid starvation in the future. All of this requires strong data visibility.
Recommendations and Solutions
Here, we turn to recommendations for mitigating the operational challenges outlined above. We include brief mentions of where INFICON’s solutions can help, though of course other products may be used.
Variability Reduction: Focusing on variability reduction related to equipment downtime, test times, and scrap and rework can drive down cycle times. Lower cycle times track with lower WIP (from Little’s Law, see Issue 21.06). Lower WIP helps reduce the scale challenge in back-end fabs (fewer lots to manage at the same time).
INFICON’s contribution: Our four-hour web-based cycle time management course includes extensive discussion about the impact of variability on cycle time and ways to reduce that impact. Past issues of this newsletter also delve into variability reduction on a regular basis.
Scheduling: The linear flow and small number of steps make it computationally feasible today to compute a schedule for the entire back-end factory. A scheduler can help with minimizing setups, keeping bottleneck tools running, balancing throughput across the different stages to enable more linear WIP flow, and meeting delivery commitments.
INFICON’s contribution: Our next-generation back-end scheduling engine, Scheduler XTR, is now live at a factory in Asia. XTR combines mathematical programming and optimization with the ability to generate thousands of fast candidate schedules. Together these map out the Pareto-optimal frontier, so planners can see exactly how trade-offs like on-time delivery and throughput play against each other and pick the schedule that best fits the business objectives. A sample Gantt chart for a wire bond area is shown below.

Reporting, Alerting and Simulation: An operations digital twin with a reporting front end can improve data visibility to help identify problems. Automatic alerting about actionable situations can also help with improvement. Reporting should also include integration with fab shipment data, to better optimize starts into the line. A key benefit of a reporting system with alerting is to identify tools that are idle with WIP waiting, as these represent lost capacity.
Another key use of reporting is to show the people running the factory what operators accomplished on the floor compared to what they could have accomplished had they followed the plan. This is where simulation can be helpful. Planners can show a simulation of what the factory could have done (by following a recommended schedule) and compare that with what was achieved, highlighting gaps. Such information about the consequences of behavior is important for manual fabs. For newer, more automated fabs, simulation can help show the impact of the KPI optimization weights that planners selected in comparison to other KPI weights that might have been implemented.
INFICON’s contribution: Our FabTime reporting system, which will be integrated into our upcoming FabOptix product, can identify real-time problems, such as tools idle with WIP waiting, as well as longer-term KPIs. In the example shown below, the left four tool groups spent most of their idle time while WIP was waiting, representing significant lost opportunity.
Our back-end scheduler effectively creates a high resolution digital twin-based simulation of the next 3-4 days, allowing customers to see the consequences at a KPI level of business practices. Together with our reporting solution, Scheduler XTR helps companies compare what could have been done with what happened. Seeing this delta is a necessary first step towards improvement.

Conclusions
Although assembly and test factories have relatively short, linear process flows, and few one-of-a-kind tools, they have many complexities that create operational improvement challenges. These include:
- Long, sequence dependent setups that need to be minimized, and for which available operator staffing can be limited
- Batching and unbatching of units at different production steps
- Form-factor changes at various stages of the process flow
- Non-linear arrivals from wafer fab
- Large numbers of tools, making qualification management and dispatching decisions challenging
- Downtime, test time variation, and other sources of variability
- Lack of data visibility, particularly for operator compliance with schedules
- And lots more…
Solutions to these challenges include variability reduction efforts (including education), advanced scheduling solutions, operations digital twins and reporting dashboards, and simulation tools for scenario assessment. Some of these solutions are available from INFICON, others, like variability reduction, require education and effort by factory management.
We would like this article to be the start of a discussion about how industry experts, industrial engineers, software developers, and factory managers can work together to drive improvements in back-end manufacturing performance.
Closing Questions for Subscribers
What back-end challenges does this article miss? If you work in a back-end factory, what solutions would be most useful for you?
Acknowledgements
I learned a great deal about scheduling for assembly and test factories from a recent SEMICON West presentation about back-end scheduling by Holland Smith and Igor Kuvychko. I also learned from discussions with John Behnke (GM of Smart Manufacturing), and from the references below. John Behnke, Lothar Mergili (Smart Manufacturing Advisor), Eric Wiedemann (Marketing Program Manager) and Frank Chance (Sr. Software Architect and FabTime co-founder) were all kind enough to review a draft version of this article and provide valuable feedback. I also used Claude Haiku 5.5 to learn some of the background information about assembly and test process flows. Any errors are mine, and I would welcome feedback to improve future versions of this article.
Further Resources
See more from this issue:
All past FabTime newsletters are available in PDF format from the FabTime Newsletter Archive. Please look for the link in the most recent email issue of the newsletter or reach out to Jennifer on LinkedIn. You can download individual issues or download a zip file containing all past issues. Some articles have been re-published on the INFICON website. Those are linked above where mentioned. See also:
- From dicing to packing: Examining the packaging process | Semiconductor Digest. This is a good overview but dates to 2001.
- Factory Scheduler for Backend Manufacturing, INFICON white paper, 2026.
For a more in-depth discussion of methods to improve semiconductor factory cycle time and operation performance, consider hosting a session of our four-hour web-based cycle time management course.