Designing the QA system in Production Line

[cliffski]

(re-post from blog at https://www.positech.co.uk/cliffsblog/2018/04/13/designing-the-qa-system-in-production-line/)

This is a big feature in my car factory game Production Line that I have been putting off until I could really set aside some time to think about the design seriously. I think I’ve finally reached that point in Early Access where I need to flesh out the design for this part of the game, and also importantly, to listen to feedback from current PL players about how they think this should work.

The current system is pretty simple. cars have work done to them on the production line, and then at the very end of the line, just before export, they go through a pre-export QA process that is broken down into emissions test, visual inspection and performance tests. All cars are assumed to be thoroughly free of defects, and are sold on the basis of their features and their price relative to the perceived value of those features.

Obviously in the real world, defects and car quality is a thing. You wouldn’t expect any defects on a $150,000 car, you would expect very few on a $70,000 car (unless its panel gaps on a tesla…dig dig…), and probably be more tolerant of minor things on a cheaper car. High end car companies (Rolls Royce etc) probably go overboard on quality assurance to ensure a reputation for zero problems (Lexus are also good at this), whereas at the lower end, its probably less of a concern. How to best represent this in the game in a way that both feels reasonable, and makes intuitive sense, and also works within the existing game design, and above all feels interesting and fun?

For a while, I have wanted to combine the idea of defects with the introduction of an area-of effect mechanic. One of the most interesting aspects of the game is the designing the layout of your factory, and at the moment, apart from making sure importers and exporters are in the right place, and ensuring that you only place office slots (like research) in office zones, there is not that much *positional* decision making involved in the game. I thought I could introduce this with the advent of new QA slots, that had a radius effect.

The system would work by generating defects at every point along the line. there would be a base level of defects that would be generated each time a task was carried out on a vehicle, including tasks generated through upgraded slots. For example, adding the rear axle = 1% chance of a defect. Adding a tire pressure sensor… 1% chance of a defect… and so on. Some cars would go through the line entirely free of defects, but many would not. Eventually, if ‘untreated’, those defects would make their way into the final cars which would be sold.

For every car sold with defects (depending how many), there would be a negative impact on the reputation of your brand, a new, lagging indicator which would affect the purchase enthusiasm of your customers. Obviously this would be something you would want to keep an eye on, so as to ensure customers were not turned away, so how do you prevent defects? and how do you treat them?

To prevent defects, you can place down QA stations, which would be area-of-effect based, and would essentially reduce the percentage chance of a defect occurring at any specific point on the line. Covering the entire production line with regular QA slots would be a way to keep production quality high and defects low, but obviously would have a cost in terms of employee pay and the floor-space required to add those QA stations. This would be a ‘prevent defects from ever happening’ approach to QA. The second approach would be to introduce a new ‘re-work’ slot right at the very end of the line (maybe just before the current QA section) which would examine each car, and take whatever time is needed (maybe 2 minutes per defect?) to fix the problem and allow a perfect car to travel beyond that point. Clever players would eventually use smart junctions (which would need defect-based rules added to them) to ensure that cars that randomly seemed to be defect-heavy did not cause a major slowdown. As I understand it, in real factories, cars that need re-work are immediately taken off the line and dealt with separately so as not to cause any real bottleneck.

This is my first attempt at designing a QA/Quality system for the game. I’m very interested to know what existing (or potential) players of the game think. Essentially it boils down to this:

Defects are randomly generated on cars as they are constructed
Area-of-effect QA stations reduce the likelihood of those defects when placed on the line
A re-work area can optionally be placed to fix defects at the end of the line
Defects will affect the brand image of the company, and thus customer eagerness to buy.

Thoughts?

[Colly]

I'm not sure how simulation driven you actually intend to be and obviously there would need to be allowances for playability within the design of the game beyond just actually simulating real life manufacturing.

I love your game and as a specialist in Lean thinking, manufacturing process design and building quality in to those processes within the factories and offices I managed throughout my career I am interested to see where you take the game.

I'm sure you have researched the Toyato Production System and all of its variants, so perhaps a consideration for your quality system might be underpin the goals of these real life systems in the game. I'm not suggesting you implement a full six sigma model in the game, although perhaps thats possible within the outline you suggested. However, I do think it is interesting to explore the Lean concepts of never receiving a bad piece, never producing a bad piece and never passing on a bad piece within the production line process and worker culture itself. Like in Toyota and many others plants, including the ones I managed, allowing the workers to STOP the production line and fix problems they uncover would bring a whole new dynamic to the simulation.

You might still work this in to the frame work you are suggesting as there might be training upgrades, similarly researched to the technologies, but specific to quality control. For example this would be implemented with training events (Kaizans to use the Lean philosophy) that overtime improves qualtity.

This might be achieved in your outline of zones by being able to implement Kaizans with those teams in the zone of influence. Thereby improving quality and being able to implement improvements to the processes under that influence. You could break this "training" in to different catagoties the game player can choose to prioritize.

Obviously, I'd love to see this element of the game linked to actual Key Performance Indicators (KPI's) that are measuredboth in the factory and from customers and are fed back in to the "training and research" through Kaizans to improve quality over time.

Anyway, sorry if that's a ramble and congratulations on your game. I'm now retired and enjoy games like this that get you solving problems. Naturally, I'd love to see this become a total simulation but understand the playability aspect you need to strive and obtain that balance for a larger market. I do think that by building and expanding the quality control system of the game around true Lean principles adds a fantastic depth and challenge to an already fun game.

[Magistar]

This is more tied to brand than price. There is a great documentaries about how the US car industry learned/changed/influenced Japanese car manufactorers. The US workers did not really care if they installed a broken headlight whereas the Japanese plant owners demanded perfection. This was a culture clash. Based on that IMO it is a brand choice which should reflect the overall brand image. Now you can still let the player decide a budget per car (e.g. 5% for model X and 10% for model Y) and then determine the influence penalty based on the expected quality, which is based on the segment. Meaning that having a low % on budget cars has less impact on brand than having low% on luxury cars.

So I do not think the impact should be directly reflected in the sales of a specific model as that implies the consumer has perfect information which in reality they never have. You find out your cars is bad after 6m, 12m, 2y, etc. No one will intentionally buy a broken car.

[cliffski]

Agreed that the customer has imperfect information, and this is why I want that indicator to lag heavily. So if you produce 50 cars with high defects, then the negative impact on the whole brand should linger for a long time, even after you fix the problems and reduce the defect rate.

Regarding the whole 'stop the line' thing, its certainly worth considering the possibility of attaching defects to slots, rather than individual cars. For example, if the 'fit axle' slot has a 1% chance per car of developing a fault (for example), then once it *does* develop a fault, we could apply a defect to every car it processes until one of those cars passes through QA. At that point, a delay could be implemented while engineers fix the fault *at the slot*.
Doing it that way round means the player would be rewarded (in terms of fewer defects) for having regular QA checks along the line, rather than simply waiting until the end.

Combine that with higher re-work costs (jn terms of time) for older (harder to get to) defects, and there is a dual incentive for the player to consider a more Japanese approach to QA than the 'hack it, then fix it later' traditional western approach?

[Molybdane]

I don't think that area of effect QA stations are a good idea, that is, I don't think that this is a good instance of the two birds with one stone mentality; the two stones being a QA system and more factory floor thinking.

Rather, I think that you should consider the assembly slots themselves. If you can modify these slots to allow production to take more time to reduce production error versus a reduced production time coupled with an increase in production error, you'd end up with a better system that can be integrated in a future system in which the player needs to decide between an quantity or quality strategy.

To clarify my thinking more, a production error would be a case of poor assembly of a component. The components themself are fine, these are after all the same; manufactured en masse through relatively simple processes, simpler at any rate than asembling a car.

Component quality would be checked when the components arrive at the factory. A sample is taken to test them, obviously this process isn't perfect and a series of poor components might end up on the factory floor anyway. A car manufacturer would normally be compensated by the supplier, but since these play no part in the game (as independent actors), this aspect of QA should not be implemented. All components are of perfect quality; only you, the player, can mess things up though hurried production causing (random) production errors.

I agree with a QA module that takes cars off the assembly line to fix them. Whether they'd be put back on the assembly line or not, doesn't really matter to me at the moment. Also, poor production quality should hurt your brand, depending on your market of course. Additional costs could be added for off site recalls.


As for the factory floor complexity that you're looking for. I am still hoping you'd consider a more dedicated in-house component production system which can be tailored by the player to produce the neccesary high end components or plenty of standard grade components. Currently, massive production is capped by two factors, the market size which I think is way too small and fragmented, and the number of import slots and their rate of import. Since each car requires a set number of components, the rate of import determines the production cap per minute.

In-house production could increase this by converting fewer raw materials into more car components. Add a super speed overhead conveyor that is not allowed to cross other overhead conveyors untill a destination is reached, and you give the player something to consider.

Quite a detour, but this is why I would not try to solve the factory floor challenge with a QA system.

[cliffski]

I am aware of the need to improve the system whereby fewer inputs are needed to produce in house than just buy pre-made components, this is on my list to fix.
Its interesting that quite a few players have expressed hesitation regarding area of effect as a mechanic, so I'm tempted to abandon that idea.

[Magistar]

Well it could result in interesting game play and it is EA so there is nothing wrong with trying it.

The think about QA that worries me is that there is just so much capitol in these cars already so scrapping them if they have faults is not an option until lategame. Having them send back through the line also is an organizational nightmare. Fixing at the slot itself makes little sense.

How about this: In addition to the QA checks themselves you could add a fixing station post-electronics where only cars go that are tagged as 'need repair'. Now this could be something mid-game and for players that skip this the cars obviously would have to be repaired on-site by the dealers. Meaning we need some sort of fining system (e.g. -2000 recall cost per car affected). So gameplay becomes risking fines, risking brand versus delay in production (qa checks) and increased production cost (repair station).

Sounds balanced to me.

[cliffski]

In the long run, I think allowing defects to slip into sold cars could definitely feed into a chance of accidents and fines. that would be awesome :D Although right now I'm concentrating just on getting a usable system of 'defects happen to cars, how and when should the player fix them, if at all?'. A simple system that we can build and expand upon later.

[z`]

AOE defect suppresion just sounds really weird. It has almost no correlation to something real - QA starts with materials and work processes and only ends with an inspector. Relatedly a QA manager wants to get involved at the start with the materials and processes but doesn't really need a desk sitting next to each workstation to manage those.

I would aim for these elements in a QA system:

Causing defects:
1. Chance of a defect is a function per activity. Depending how far you want to chase the tail on this this could filter down all the way from import/component manufacture but for player and computational sanity maybe don't do that.
2. Brainstorming defect function inputs:
a. Expertise - the longer a station is placed, the better it is at avoiding defects. New upgrades decay it a little. Punish rampant retooling with more than the money part. Could also have a global counterpart representing your human capital - demoing a station can refund some station experise into the global pool which is then spent so that a new station starts with a little boost representing your experienced worker getting started on a new line.
b. Rush factor - Let us set a rush factor that will make a task take longer or slower. Gives us a final lever to sync takt time with the risk/reward of do you round up or down. Make changing it decay expertise a little - every knob turn represents overhauling standard work to do things differently
c. Component source - Here's where that comment about how far we chase it comes back into play. This could look like being able to choose a premium to pay for parts to reduce defect chance of what its going into. Homemade parts could be toward the middle or better end depending where you want to take that balance. Or just screw it and use the same defect function on homemade parts with the same levers where you could be making impeccable or total crap parts.

3. Brainstorming defect techs
a. Expertise rate gain (global)
b. rush factor impact (global)
c. homemade part quality (global)
d. Import QA (upgrade to import station, increases time per part but positively impacts defect chance)

Fixing defects:
The car system already in place already gives the perfect backbone.

You have a Quality Station and a Rework Station. Rework only comes after quality. Rework would be a single level tech but quality station would be multi level.

Quality only comes after the entire system on a car is built. Researching subsequent quality stations starts breaking them out to finer detail.

So for example your stock Quality Station can only check a car that is built entirely. You can then research a Chassis Quality Station. A chassis quality station works only on cars with a completed chassis system.

A Quality Station always finds defects and will in effect flag them. The standard build would be then to divert (or always go into/pass through) to a Rework Station where you would repeat the task that caused the defect originally. Going through the whole list of found defects, with a time penalty depending on how built the car is: its harder to fix things behind the interior trim or in the back of the engine bay once everything is put together.

So you get the choice of frequent inspections requiring more space and capital or waiting till the end. The former probably being better if you know your basic quality is going to be crap while the latter lets you take advantage of the gains you've made in defect chance through longterm design - there's no one answer and the average arc of a line is that you would start with quality at the end because that's all you can manage, progress to frequent quality checks to take advantage of quicker fixes, until you've nailed the line down and need to only check at the end for a rare defect.