← All games

03 / PROJECT STORY · DESIGNER DIARY

Made With Care

Run a sushi shop across five shifts. Hire your crew and plan their work, then program robots to take over.

Open the restaurant

01Inspiration

Made With Care is a five-chapter sushi restaurant management game. At first, you assign staff to slice fish, cook rice, prepare dishes, and serve customers. By the end, you can set priorities for robots and let the kitchen and conveyor belt run automatically.

The idea came from my experience developing customer service systems. I saw the work move from local teams in the United States to overseas contractors, then gradually to chatbots. The tools kept improving while the people who had provided the service left in successive waves.

I chose a sushi restaurant because the staff's work can be shown directly. Players recruit people, divide up tasks, and learn the preparation and service routine before deciding whether to replace those employees with machines. The name “Made With Care” contrasts with the increasingly automated service.

Day one · staff service
Day one · staff service
Day five · robots and conveyor belt
Day five · robots and conveyor belt
Day one requires assigning staff to prepare and serve food. By day five, robot rules and a conveyor belt can handle service.

02How a day works

Each day begins with a newspaper. Its stories suggest recent food preferences, such as tuna coming into season, while the supply page estimates customer numbers. Players use this information to choose a menu, hire staff, and buy ingredients and equipment. They then have 30 seconds to prepare before opening. During service, they assign staff to cook and deliver orders placed by seated customers.

The news relates to newly unlocked recipes but does not list each customer's order in advance. Players still have to estimate how much to buy. I also rewrote the newspaper itself. Early articles directly said that new dishes would increase revenue or new employees could help, which did not sound like news. I changed these to neighborhood reports about local lunches and gatherings that suggest what people might want to eat.

Newspaper
Newspaper
Menu
Menu
The newspaper suggests food preferences. The menu shows ingredients, selling prices, and earnings after ingredient costs.

03Adjusting the second level

The second level was originally much harder than the first. Buying equipment made processing faster, but players could not operate the controls any faster. They still could not keep up.

Whenever a worker finishes a job, the player needs to assign another. Finding the employee, checking the orders, and choosing an action all take time. Faster processing can also mean having to give new instructions more frequently. Equipment output alone does not account for that work.

I added the rate of task assignment to the simulation. All employees shared one input interval, allowing at most one instruction every 2, 3, 4, or 5 seconds. Selecting an employee and assigning a task both counted as part of that instruction. In level two of the early three-level version, 16 customers arrived over 99 seconds. Even with all the equipment, assigning a task every 4 seconds served only 10 customers and left 318 in cash, below the target of 400. One instruction every 2 seconds served everyone.

For this adjustment, I kept the 16 customers, staff, purchases, and target, but extended the span between the first and last arrival from 99 to 160 seconds. I standardized patience at 95 seconds across ordinary levels, removing the second level's shorter 40-second allowance. Preparation still felt rushed in play, so I then increased the preparation period across all levels from 20 to 30 seconds.

Early three-level version · level two pacing test
Early three-level version · level two pacing test
The test keeps the scheduling strategy fixed while limiting the task assignment rate shared by all employees. Arrivals, patience, and preparation time changed together; the comparison is between complete configurations.

After the changes, one instruction every 4 seconds served all 16 customers and left 502 in cash. A 5-second interval also passed. In the same test, arrival spans from 150 to 190 seconds all allowed every customer to be served. I chose 160 seconds for this revision to keep the day short while allowing full service.

Pacing test record saved during development
SetupCommand intervalServedCashTargetResult
Before2 s16/16502400Pass
Before3 s14/16442400Pass
Before4 s10/16318400Fail
Before5 s8/16254400Fail
After2 s16/16502400Pass
After3 s16/16502400Pass
After4 s16/16502400Pass
After5 s16/16502400Pass
One instruction includes selecting an employee and assigning a task. All employees share the input interval. The simulation excludes misclicks and hesitation. In the original output, “current” refers to the configuration after this adjustment.
Original test output
┌─────────┬──────────────────────────────┬───────────────────────┬─────────┬──────┬──────┬───────┬──────────────┬──────────┬───────────────────────┐
│ (index) │ scenario                     │ secondsPerInstruction │ served  │ cash │ goal │ won   │ instructions │ duration │ instructionsPerMinute │
├─────────┼──────────────────────────────┼───────────────────────┼─────────┼──────┼──────┼───────┼──────────────┼──────────┼───────────────────────┤
│ 0       │ 'Day 1 current'              │ 2                     │ '12/12' │ 360  │ 260  │ true  │ 32           │ 194.8    │ 9.9                   │
│ 1       │ 'Day 1 current'              │ 3                     │ '12/12' │ 360  │ 260  │ true  │ 32           │ 198.5    │ 9.7                   │
│ 2       │ 'Day 1 current'              │ 4                     │ '12/12' │ 360  │ 260  │ true  │ 32           │ 199.7    │ 9.6                   │
│ 3       │ 'Day 1 current'              │ 5                     │ '11/12' │ 328  │ 260  │ true  │ 32           │ 208.7    │ 9.2                   │
│ 4       │ 'Day 2 before, machines'     │ 2                     │ '16/16' │ 502  │ 400  │ true  │ 43           │ 145.7    │ 17.7                  │
│ 5       │ 'Day 2 before, machines'     │ 3                     │ '14/16' │ 442  │ 400  │ true  │ 43           │ 151.7    │ 17                    │
│ 6       │ 'Day 2 before, machines'     │ 4                     │ '10/16' │ 318  │ 400  │ false │ 38           │ 158.9    │ 14.3                  │
│ 7       │ 'Day 2 before, machines'     │ 5                     │ '8/16'  │ 254  │ 400  │ false │ 29           │ 153.8    │ 11.3                  │
│ 8       │ 'Day 2 before, no machines'  │ 2                     │ '8/16'  │ 330  │ 400  │ false │ 28           │ 155.1    │ 10.8                  │
│ 9       │ 'Day 2 before, no machines'  │ 3                     │ '6/16'  │ 268  │ 400  │ false │ 28           │ 158      │ 10.6                  │
│ 10      │ 'Day 2 before, no machines'  │ 4                     │ '5/16'  │ 242  │ 400  │ false │ 26           │ 159      │ 9.8                   │
│ 11      │ 'Day 2 before, no machines'  │ 5                     │ '5/16'  │ 242  │ 400  │ false │ 21           │ 159      │ 7.9                   │
│ 12      │ 'Day 2 current, machines'    │ 2                     │ '16/16' │ 502  │ 400  │ true  │ 43           │ 209.7    │ 12.3                  │
│ 13      │ 'Day 2 current, machines'    │ 3                     │ '16/16' │ 502  │ 400  │ true  │ 43           │ 207      │ 12.5                  │
│ 14      │ 'Day 2 current, machines'    │ 4                     │ '16/16' │ 502  │ 400  │ true  │ 43           │ 206.8    │ 12.5                  │
│ 15      │ 'Day 2 current, machines'    │ 5                     │ '16/16' │ 502  │ 400  │ true  │ 43           │ 249.7    │ 10.3                  │
│ 16      │ 'Day 2 current, no machines' │ 2                     │ '16/16' │ 582  │ 400  │ true  │ 43           │ 271.4    │ 9.5                   │
│ 17      │ 'Day 2 current, no machines' │ 3                     │ '15/16' │ 550  │ 400  │ true  │ 43           │ 274.9    │ 9.4                   │
│ 18      │ 'Day 2 current, no machines' │ 4                     │ '15/16' │ 550  │ 400  │ true  │ 42           │ 274.2    │ 9.2                   │
│ 19      │ 'Day 2 current, no machines' │ 5                     │ '12/16' │ 458  │ 400  │ true  │ 37           │ 285      │ 7.8                   │
└─────────┴──────────────────────────────┴───────────────────────┴─────────┴──────┴──────┴───────┴──────────────┴──────────┴───────────────────────┘
Download record

Slower arrivals also changed how necessary equipment was. In the same simulation, buying no equipment and assigning a task every 5 seconds served only 12 of 16 customers, but still left 458, above the 400 target. I kept this option, allowing players to pass through task planning and savings on equipment.

04Batch preparation

Fish slicing originally produced exactly 2 slices per task. I changed it to a selectable quantity, with a maximum of 4 after upgrading the knife. Rice changed from a fixed batch of 4 portions to a maximum of 2. Larger fish batches reduce repeated instructions; smaller rice batches make it easier to control preparation quantities and storage use.

With Haru and the upgraded knife, 2 slices take about 6.15 seconds and 4 take about 12.31 seconds, counting processing only, not walking. Two batches of 2 and one batch of 4 take the same total processing time. The difference is that the player does not have to assign another task halfway through. Tests checked batch limits, duration, material use, and storage occupancy.

Batch limits and processing time
Batch limits and processing time
With the same employee and knife, combining batches saves one instruction while processing time still scales with the quantity. Rice changed from a fixed batch of four to a maximum of two to make preparation quantities easier to control.

Batching had another problem: if the player requested 4 slices but only had enough ingredients for 2, the employee did nothing. I changed it to process as much as possible. If the fridge has only 2 spaces left, the worker makes 2 slices. Players no longer need to calculate the remaining space and then edit the instruction themselves.

Batch processing boundary checks
Tests cover quantity limits, linear processing time, and partial completion when ingredients or storage space run short.
Original test output
# Subtest: batch limits describe final outputs rather than ingredient package multipliers
ok 14 - batch limits describe final outputs rather than ingredient package multipliers
  ---
  duration_ms: 0.651917
  type: 'test'
  ...
# Subtest: four fish slices consume exactly two fish and stop after one selected batch
ok 15 - four fish slices consume exactly two fish and stop after one selected batch
  ---
  duration_ms: 9.150542
  type: 'test'
  ...
# Subtest: work time scales by finished portions, then skill and coffee apply
ok 18 - work time scales by finished portions, then skill and coffee apply
  ---
  duration_ms: 2.838584
  type: 'test'
  ...
# Subtest: a batch uses available storage immediately and stops after the clamped quantity
ok 21 - a batch uses available storage immediately and stops after the clamped quantity
  ---
  duration_ms: 17.704875
  type: 'test'
  ...
# Subtest: partial raw stock makes two or three fish slices and one rice portion without topping up later
ok 22 - partial raw stock makes two or three fish slices and one rice portion without topping up later
  ---
  duration_ms: 19.6105
  type: 'test'
  ...
Download record

Customers now arrive in groups, with time to replenish supplies in between. Delivering a dish pays immediately, removing the extra progress bar that previously ran after the food had reached the customer.

Day one · serving a customer · 5 seconds · silent
Payment arrives as soon as the dish is delivered.

Patience and preparation time stay consistent across ordinary levels. Difficulty changes through arrivals, orders, and equipment. To keep each level around five minutes, I had once added a forced closing time, but customers could still be waiting for food when it triggered. I replaced that with adjustments to customer numbers and arrival timing, allowing the final order to finish before closing.

05Recipes and earnings

Even with fewer repetitive inputs, play still largely consisted of checking who was most impatient and deciding what to do first. I added recipes so players could choose different ways to run the restaurant before opening.

Players select the day's menu from the dishes they have learned. Onigiri needs only rice and nori; chirashi don needs two kinds of fish, rice, and egg, making purchasing and preparation more involved. Recipes unlock directly after passing a level. Players decide whether to add them to the menu.

Dishes that require more types of ingredients and more preparation should earn more per serving. For example, onigiri sells for 20 with an ingredient cost of 2, leaving 18. Chirashi don sells for 84, costs 17 in ingredients, and leaves 67. That difference gives players a reason to prepare more ingredients. It also lets them pass up some ordinary orders to prioritize dishes with a higher return.

Onigiri and chirashi don cost comparison
Onigiri and chirashi don cost comparison
After ingredients, onigiri earns 18 and chirashi don earns 67. Wages, equipment, and waste are separate costs.

Those figures subtract only ingredients. Wages, equipment, and waste are separate costs. I also relaxed the original target calibration, which allowed three missed customers but not four, to accommodate different menus and business strategies. Savings from previous days or more high-return orders can compensate for additional missed customers.

I revised the accounting too. The test version had left too much starting cash, and money remaining after buying ingredients was not fully counted toward the target. I changed it to provide funding only once, at the start. Each day's target is then checked against the previous balance plus sales revenue, minus ingredients, equipment, and wages. I reduced purchasing warnings as well: buying fewer ingredients may be an intentional decision to serve only certain dishes.

06Equipment and robots

The original three chapters represented the early, middle, and late stages. Expanding to five allowed more gradual equipment upgrades: better knives first, then rice cookers, ovens, steam cabinets, and commercial machines.

In one version, upgraded equipment sliced fish and cooked by itself, with names and shapes that did not resemble actual kitchen equipment. I replaced these with more familiar tools that increase batch size or speed but still require an employee to operate them. The player decides what to replenish and which dish to make first. Later, robots can make those decisions by following rules.

Robot priorities run from top to bottom. For example: serve a finished dish if one is available, cook when there is an order, and replenish ingredients when needed. Players can change the conditions, quantities, and order. The conveyor belt only transports completed dishes. By day five, the player's main tasks are adjusting rules and checking how the system runs.

Day five · robot rules
Day five · robot rules
Set conditions, quantities, and priorities for serving, cooking, and replenishing ingredients.

The game originally started with three employees. I changed it so players must hire first. Candidates have different specialties, abilities, and wages, requiring players to divide the work according to their budget and menu. Replacing them with robots later also means opening the staff page and dismissing the original employees manually. I kept that step so replacing staff remains the player's own decision.

Day five · automated service · 6 seconds · silent
Robots follow their rules while the conveyor belt carries finished dishes.

07Art, music, and interface

Day one uses warm wood, paper, and leather details. Later kitchens gradually introduce cooler colors and more metal. Early music centers on relaxed piano and saxophone; later tracks become more mechanical as the equipment and workforce change.

Raw ingredients and prepared food sit together in the top bar, so players do not need to keep opening the inventory during service. Earlier versions had a dense background that distracted attention from the restaurant. I removed the background houses and enlarged the restaurant to make the staff, food, and customer actions easier to see.

Next projectFraud Score