WEEK 1 ASSIGNMENTSDay 1 ~ Day 5Video & Advertising Platform
Week 1 Individual Homework Missions
Apply the AI-Native SW design ideas you learned directly to a real business scenario (Video & Advertising Platform),
complete requirements definition, user flows, data integrity, API contracts and UI/UX checks, and record your decision evidence in ai_review.md.
10 assignments
2 per day / 10 problems total
50 pts / 100 pts
50 pts per problem (100 pts per day)
90โ120 min
Suggested time per day
ZIP submission
Submit StudentID_Name_DayNN.zip
๐ Day 1โ5 homework submission and grading rules
File structure: inside the StudentID_Name_DayNN.zip file, create HW-DNN-01 and HW-DNN-02 folders and submit the required docs and ai_review.md
AI review record (ai_review.md required): must include the questions you used, AI's key answers and your reasons for accepting, changing or rejecting (simply copying AI answers loses points)
Grading: 2 problems per day (50 pts each, 100 total); your Week 1 final grade is the average of the 5 days' scores.
DAY 01 ASSIGNMENTS
Day 1. Requirements and Domain Responsibility
HW-D01-01
Day 1 โข HW-D01-0150 pts
HW-D01-01. Decide Between Conflicting Service Needs
Resolve conflicting needs of 4 roles โ Viewer (watch without account), Creator (keep private), Advertiser (show ads everywhere), Admin (show only approved) โ and write a requirements doc (requirements.md).
Tech SkillTECH-BA, TECH-DDD
AI SkillAI-CTX, AI-REV, AI-DOC
Lab areaHYBRID
Suggested time45โ60 min
RequirementsConflict ResolutionFR ID
Output:requirements.md, ai_review.md
HW-D01-02
Day 1 โข HW-D01-0250 pts
HW-D01-02. Who Owns View Counts and Ad Costs?
Clearly separate the events of playback request, actual viewing, ad impression and cost, and work out ownership across 5+ Domains and responsibility for preventing double charges.
Tech SkillTECH-DDD, TECH-ARCH
AI SkillAI-CTX, AI-REV, AI-DOC
Lab areaHYBRID
Suggested time45โ60 min
Domain BoundaryData OwnershipEvent Deduplication
Output:domain_map.md, ai_review.md
DAY 02 ASSIGNMENTS
Day 2. Flows and Design Changes
HW-D02-01
Day 2 โข HW-D02-0150 pts
HW-D02-01. Explain the Playback Experience Even If the Ad Server Stops
Design a Mermaid Sequence Diagram and a user wait policy so viewing never freezes forever, even with ad response delays, no candidates or the user leaving before playback.
Tech SkillTECH-FLOW, TECH-ARCH
AI SkillAI-CTX, AI-REV, AI-DOC
Lab areaHYBRID
Suggested time45โ60 min
Sequence DiagramException FlowTimeout Policy
Output:playback_flow.md, ai_review.md
HW-D02-02
Day 2 โข HW-D02-0250 pts
HW-D02-02. Find the Docs Broken by a Policy Change
When policy v2 changes to "at most 3 impressions per user per day," find 4+ mismatches between existing screen/sequence/data docs, and write an impact analysis and Decision Log.
Tech SkillTECH-FLOW, TECH-BA, TECH-QA
AI SkillAI-CTX, AI-REV, AI-DOC
Lab areaHYBRID
Suggested time45โ60 min
Consistency AuditImpact SyncDecision Log
Output:change_review.md, ai_review.md
DAY 03 ASSIGNMENTS
Day 3. Data Models and Integrity
HW-D03-01
Day 3 โข HW-D03-0150 pts
HW-D03-01. Design a Data Model That Separates Views, Impressions and Charges
Design a 3NF ERD with 6+ Entities like Campaign, Ad, Impression and Charge, and prove normalization integrity with past charge history and 8 sample records.
Tech SkillTECH-DATA, TECH-DDD
AI SkillAI-CTX, AI-REV, AI-DOC
Lab areaHYBRID
Suggested time45โ60 min
3NF ERDData DictionarySample Data
Output:data_model.md, ai_review.md
HW-D03-02
Day 3 โข HW-D03-0250 pts
HW-D03-02. Resolve the Conflict Between an Ad Deletion Request and Past Charges
Resolve the conflict between an advertiser's request to delete creatives and the operator's need to keep past charges with a Soft Delete / de-identification model, and write before/after data examples and check cases.
Tech SkillTECH-DATA, TECH-QA
AI SkillAI-CTX, AI-REV, AI-DOC
Lab areaHYBRID
Suggested time45โ60 min
Soft DeleteIntegrity GuardValidation Cases
Output:data_change_review.md, ai_review.md
DAY 04 ASSIGNMENTS
Day 4. API Contracts and Review
HW-D04-01
Day 4 โข HW-D04-0150 pts
HW-D04-01. Submit an API Contract Another Team Can Build From
Write OpenAPI 3.0 specs for 4 endpoints โ video lookup, playback, campaign creation and ad request โ and define success/failure Payload examples and ownership conditions.
HW-D04-02. Restore Compatibility of a Broken Contract
Identify 4+ consumer impacts from a new API response change (a Breaking Change like returning failures as HTTP 200), and set a backward-compatibility recovery strategy and test cases.
Tech SkillTECH-API, TECH-QA
AI SkillAI-CTX, AI-REV, AI-DOC
Lab areaHYBRID
Suggested time45โ60 min
Breaking ChangeBackward CompatibilityContract Test
Output:api_review.md, ai_review.md
DAY 05 ASSIGNMENTS
Day 5. UI/UX and User Checks
HW-D05-01
Day 5 โข HW-D05-0150 pts
HW-D05-01. Design the Screens for a First-Time Ad Creator
Design 3 mobile campaign creation screens for a beginner local trader, the 6 UI states (Loading/Error, etc.), recovery actions and an API mapping table.
Critically review the screens from assignment 1, judge 5 pieces of mock user feedback like "mistaking it for waiting for approval" and "hard to find retry," and produce before/after screens and recheck criteria.
You're the first planner at "VidAce", a Nigeria-Korea joint video ad platform. Two weeks before launch, the CEO drops the requirement lists from 4 stakeholder teams on your desk.
RoleKey needConflict point
๐งโ๐ป Viewer (no account)Watch right away without logging inConflicts with Creator's need for privacy
๐ฌ CreatorPrivate by default right after uploadConflicts with Advertiser's need for instant ad display
๐ผ AdvertiserInsert ads in every public videoConflicts with Admin's limit on posting before approval
๐ก๏ธ AdminAllow only reviewed videos to be publicConflicts with Viewer's need to watch right away
๐ง Questions to think about
How could you define the boundary between videos a Viewer can and can't watch without an account?
If a Creator sets a video private but an Advertiser wants ads on it, whose need should come first? Why?
When should Admin review happen? Right after upload? When publishing is requested? Which timing makes business sense?
What is a "checkable requirement"? What's the difference between "must load fast" and "the first frame must show within 2 seconds"?
Is this platform's main revenue ads or creator subscriptions? How does that affect requirement priorities?
๐ก Example prompt (for reference โ don't use it as-is)
Below is an example to help you understand how to use AI. In the real assignment, you must analyze the scenario yourself first and write a more specific prompt that fits the situation.
Example prompt โ HW-D01-01
[Role] You are a Business Analyst.
[Context] The 4 roles (Viewer, Creator, Advertiser, Admin) of the video ad platform VidAce
each have different needs, and some directly conflict with each other.
[Request]
1) Summarize each of the 4 roles' goals in 1โ2 sentences
2) Identify 2 clearly conflicting pairs of needs, and explain the policy adopted and the reason for giving up the other for each
3) Write 8 checkable requirements from FR-001 in requirements.md format
(each: ID, role, requirement, acceptance condition, exception condition)
[Output format] Use Markdown tables and section headers
๐ฏ The real assignment โ do it yourself
1
List requirements by role For each of the 4 roles, separate explicit needs and implicit needs and derive at least 12 items.
2
Write 8 checkable requirements with FR IDs Write each requirement in the form FR-001: [role] must be able to [action] when [condition], with 2+ testable acceptance criteria (AC).
3
Record conflict resolution decisions For 2 conflicting pairs, write โ the policy adopted, โก the alternative given up, โข the business reason for giving it up, โฃ affected roles and how to make up for it.
4
Sort out-of-scope features and open questions Write 3 features excluded from the v1.0 release and 3 open questions, explaining each reason in one sentence.
5
Write ai_review.md Record at least 1 each of AI suggestions you โ accepted, โก changed (with reason) and โข rejected (with reason).
๐ Grading (50 pts):Checkable requirements 20 ptsLogical conflict resolution reasons 15 ptsClear separation of scope and assumptions 10 ptsEvidence of AI judgment and self-critique 5 pts
DAY 1 ยท HW-D01-02 ยท 50 pts
๐ฅ Who Owns View Counts and Ad Costs?
๐ Business scenario
VidAce's data team released a shocking report: ad impression events are being recorded 2.3 times more than real impressions, and advertisers are being overcharged at month end. Every team says, "It's not our fault."
๐ด A Viewer pressed play but closed it after 3 seconds โ was it counted as a view?
๐ด The ad was called twice but the Viewer saw it once โ what's the billing basis?
๐ด A network error sent the playback signal to the server twice โ who's responsible for preventing double records?
๐ด The user watched 25 seconds of a 30-second ad, then closed the app โ counted as completed? Partial?
๐ง Questions to think about
Events vs facts: are "clicking play" and "actually starting to watch" the same event? By what criteria can you separate them?
What ownership means: which data should be stored in which team's DB? What problems happen if two teams each store the same data?
Preventing double processing: when the same event comes in twice, what's the difference between "use the first" and "use the last"?
Billing responsibility: who should bill the Advertiser for ad costs? Should recording billing events and billing be handled in the same Domain?
๐ก Example prompt (for reference โ don't use it as-is)
For Domain analysis, sort the event list and scenarios yourself first, then ask AI to review.
Example prompt โ HW-D01-02
[Role] You are a Domain-Driven Design expert.
[Context] On a video ad platform, playback request, actual viewing, ad impression and billing
are different events, and there's currently a double-recording problem.
[Domains to analyze] Video, Playback, Ad, Impression, Charge, User
[Request]
1) Define each Domain's Primary Responsibility and source data ownership in one line
2) Compare 2 alternative ways of splitting responsibility for "playback request โ start watching โ ad impression โ billing"
3) Suggest a policy for handling duplicate events (e.g. Idempotency Key)
๐ฏ The real assignment โ do it yourself
1
Event type table Define at least 8 event types and mark each as "counted? / billed? / double-processing risk?" with Yes/No and a reason.
2
Domain responsibility map (domain_map.md) Define at least 5 Domains and state for each โ main responsibility, โก data it owns, โข events/data to share, โฃ data it must never write directly.
3
Design a double-processing prevention policy Compare the pros and cons of 2+ technical strategies (Idempotency Key, Redis distributed lock, etc.) to prevent double ad impression records, and describe your chosen strategy and why.
4
Decide how to handle 4 boundary scenarios For each of 4 scenarios (leaving after 3 sec, ad called twice, network double send, closing after 25 sec), decide the handling policy and how data is recorded, with reasons.
๐ Grading (50 pts):Consistent Domain responsibility and ownership 20 ptsPrecise event separation 15 ptsFeasible double-processing prevention 10 ptsEvidence of AI judgment and self-critique 5 pts
DAY 2 ยท HW-D02-01 ยท 50 pts
๐ฅ Guarantee Playback Even If the Ad Server Stops
๐ Business scenario
At 3 p.m., a surge of VidAce users pushes ad server response time past 8 seconds. Thousands of Viewers pressed play, but are waiting with a frozen screen and no message.
โก Ad server timeout: no response in 3 seconds โ just play the video? Keep waiting?
โก No ad candidates: inventory for that time/region is used up โ how do you tell the Viewer?
โก Leaving right before playback: the ad request finished but the Viewer closed it โ bill the ad?
โก Partial success: the ad server responded but the creative URL failed to load โ play right away?
๐ง Questions to think about
Choosing the timeout: what if the ad server wait is 1 second? 10 seconds? Who should decide the best value, and how?
Fallback strategy: how much business is lost when playing without an ad? And how many users leave when the video isn't shown? Which loss is bigger?
User communication: in the same situation, how do "Preparing your ad" and "Your video will start shortly" feel different?
Billing boundary: should an ad that was requested but not shown be billed? Who should make this decision?
๐ก Example prompt (for reference โ don't use it as-is)
If this is your first Mermaid Sequence Diagram, learn the basic format from the example below. The real assignment must include all 4 exception cases.
Example prompt โ HW-D02-01
[Role] You are a distributed systems architect.
[Context] Video playback needs an ad server call, but the ad server can fail for many reasons,
like timeout/no inventory/creative loading failure.
[Request]
1) Write the normal sequence as a Mermaid sequenceDiagram
2) Add a sequence for the "ad server 3-second timeout" exception, including the Fallback policy and user message
3) Organize in a table whether each exception is billed, and why
๐ฏ The real assignment โ do it yourself
1
Normal flow sequence diagram Write the normal case of the Viewer โ VideoService โ AdService โ CDN โ Viewer flow in Mermaid. Include 5+ Actors, 10+ messages and timing notes.
2
Sequences and policies for 4 exception cases For each of the 4 exceptions, make a separate Mermaid diagram and a table of โ detection condition, โก Fallback behavior, โข exact message, โฃ whether it's billed.
3
Timeout Policy decision doc Decide the ad server wait limit and explain your reasoning logically in terms of the trade-off between Viewer drop-off and ad revenue. Include a table comparing 2 alternatives.
4
Decide whether to use a Circuit Breaker Give the pros and cons of opening a Circuit Breaker after 5 straight ad server failures, and conclude whether to use it on this platform.
๐ Grading (50 pts):Diagram accuracy and completeness 15 ptsHandling of 4 exception cases 20 ptsLogical reasons for the Timeout policy 10 ptsEvidence of AI judgment and self-critique 5 pts
DAY 2 ยท HW-D02-02 ยท 50 pts
๐ฅ Find and Fix the Docs Broken by a Policy Change
๐ Business scenario
3 months after VidAce launched, user drop-off surged because of ad fatigue. The ad team decided on an urgent policy change.
โ Old policy v1: "same user, same ad โ unlimited impressions"
โ New policy v2: "same user, same ad โ at most 3 impressions per day"
The policy changed, but these 5 docs are still based on v1: the ad playback screen spec, ad request sequence diagram, Impression DB table design, ad billing policy doc and QA test cases.
๐ง Questions to think about
Predict the ripple effect: how many changes across the whole system will one rule change, "max 3 per day," require? List them first.
Data model change: which column is likely missing from the Impression table? If you add it, what happens to existing data?
Real-time check: how can you quickly look up "how many times did this user see this ad today"? Is a DB query enough?
Policy boundary cases: what if the 3rd ad was seen right before midnight? Is "a day" based on UTC or the user's local time?
๐ก Example prompt (for reference โ don't use it as-is)
People must think about doc mismatches first. Analyze each of the 5 docs, then hand the review to AI.
Example prompt โ HW-D02-02
[Role] You are a QA Lead and Technical Writer.
[Context] The ad impression policy changed from "unlimited" to "max 3 per day."
The existing 5 docs were written based on the v1 policy.
[Request]
1) Identify 2 items in each doc that need fixing because of the policy change (10+ total)
2) For each mismatch, organize: original | correct content after the change | impact, in a table
3) Write 5 Decision Log entries for the policy change
๐ฏ The real assignment โ do it yourself
1
Find mismatches per doc (at least 4) For each mismatch, describe โ doc name and location, โก current v1 content, โข correct v2 content, โฃ the specific problem if not fixed.
2
DB Schema change design Based on the Impression table, include columns to add/change, an index strategy, an aggregation query example (pseudo-SQL) and a migration strategy for existing data.
3
Updated ad request sequence diagram Write a new Mermaid sequence adding a "check whether over 3 per day" step, including the branches for under/over 3 and the Viewer message when over.
4
Document the Decision Log and assumptions Record 3 clearly decided items and 3 undecided items needing more discussion separately. For undecided items, write the options and the impact of each.
๐ Grading (50 pts):Evidence-based mismatch detection 20 ptsSpecific DB schema changes 15 ptsClear separation of decided/undecided 10 ptsEvidence of AI judgment and self-critique 5 pts
DAY 3 ยท HW-D03-01 ยท 50 pts
๐ฅ Design a Data Model That Separates Views, Impressions and Charges
๐ Business scenario
A serious problem came up in VidAce's month-end settlement: changing an ad campaign's budget midway also changed the previous month's billed amount. Advertiser complaints are pouring in.
โ Problem: Charge.amount looks up Campaign.budget_per_day live, so past charges change automatically when the budget changes
๐ง Questions to think about
Why normalize: why is 3NF needed? What specific anomalies happen in this scenario if 3NF is broken?
History vs normalization: to keep past billed amounts, you may ironically need to store some data twice. How will you resolve this trade-off?
Video and Impression relationship: if one video can show many ads, and one ad can show in many videos, what kind of relationship is it?
Aggregated vs raw data: which is more useful โ a table storing total daily impressions, or one recording each impression?
๐ก Example prompt (for reference โ don't use it as-is)
For data model design, always draw the ERD yourself first, then ask AI to check it.
Example prompt โ HW-D03-01
[Role] You are a data architect.
[Context] There's a design flaw where Campaign budget changes affect past Charge history.
[Entities] Campaign, CampaignBudgetHistory, Ad, Video, Impression, Charge, User, Advertiser
[Request]
1) Define each Entity's PK, FK, main attributes and relationships
2) Explain a design that keeps past Charge history unchanged when the budget changes
3) Prove "billed amounts stay consistent before and after a campaign budget change" with 8 sample records
๐ฏ The real assignment โ do it yourself
1
3NF ERD design (at least 6 Entities) Write the ERD as a Mermaid erDiagram or text table. Include PK, FK, required attributes and data types for each Entity. Use a separate Entity for Campaign budget history.
2
Data dictionary For each of the 6 Entities, describe in one sentence each โ the Entity's purpose, โก each column and its constraints, โข related business rules.
3
Prove integrity โ sample data scenario Build the scenario "Campaign A starts in January with a daily budget of โฆ100,000 โ changes to โฆ80,000 on January 15 โ January invoice issued" with 8 sample records.
4
Design a duplicate impression query Write logic in pseudo-SQL that judges a duplicate when the same user requests the same ad twice in the same session, and explain why you chose the time window.
๐ Grading (50 pts):Correct relationships and constraints 20 ptsComplete history-keeping design 15 ptsConsistent sample data 10 ptsEvidence of AI judgment and self-critique 5 pts
DAY 3 ยท HW-D03-02 ยท 50 pts
๐ฅ Resolve the Conflict Between an Ad Deletion Request and Keeping Past Charges
๐ Business scenario
An urgent request came from the legal team. Advertiser X demands their ad (confirmed as an illegal medicine ad) be deleted immediately along with all past impression records. But the finance team pushes back: the ad's past billing records are subject to tax audit and must be kept for 5 years.
โ Delete Ad record โ Impression.ad_id FK violation cascades to delete Impressions
โ Delete Impression โ Charge.impression_id FK violation cascades to delete Charges
โ Past billing records lost โ fail tax audit, legal liability
โ Advertiser's personal data deletion request (GDPR/NDPA) conflicts with the duty to keep billing records
๐ง Questions to think about
Limits of Soft Delete: even with is_deleted=true, the data stays in the DB. If the advertiser exercises the right to erasure, is a simple Soft Delete enough?
Scope of de-identification: what info in the ad creative can you erase so it counts as "personal data deleted" while billing records stay?
FK constraints and archives: to stop cascading deletes, which fits better โ ON DELETE SET NULL or ON DELETE RESTRICT?
Priority of legal requirements: when the right to erasure conflicts with local tax retention duties, which law takes priority?
๐ก Example prompt (for reference โ don't use it as-is)
The key is balancing technical solutions with legal requirements. Analyze the 4 conflict scenarios yourself.
Example prompt โ HW-D03-02
[Role] You are a data governance expert.
[Context] When an ad deletion is requested, foreign key cascading deletes destroy past billing records.
The right to erasure (GDPR/NDPA) conflicts with tax retention duties.
[Request]
1) Design a model combining Soft Delete + personal data de-identification, with record examples before/after deletion
2) State 2 problems this model solves and 2 it doesn't
3) Evaluate a separate data archive table strategy as an alternative
[Note] Mark legal judgments as "needs review by a legal expert"
๐ฏ The real assignment โ do it yourself
1
Detail 3 cascading delete problems Explain step by step, from the SQL FK constraint view, how an Ad Hard Delete cascades Impression โ Charge โ tax report. Include the error type and business impact at each step.
2
Design a Soft Delete + de-identification model Design the columns to add to the Ad table and how deletion requests are handled. Write a Before/After table of records before and after (de-identified), and check that billing can still be calculated after de-identification.
3
Write 4 check cases โ Charges remain after normal deletion, โก advertiser info can't be looked up after de-identification, โข billing totals stay accurate, โฃ reactivation (undoing deletion) isn't possible. Include expected input, steps and expected result for each.
4
Resolve the right-to-erasure vs tax retention conflict Suggest 2 technically possible solutions and clearly separate what each solves and doesn't.
๐ Grading (50 pts):Meets deletion request and retention duty together 20 ptsTechnical completeness of the de-identification model 15 ptsSpecific check cases 10 ptsEvidence of AI judgment and self-critique 5 pts
DAY 4 ยท HW-D04-01 ยท 50 pts
๐ฅ Write an API Contract Another Team Can Build From Right Away
๐ Business scenario
VidAce's back-end team (Lagos) and front-end team (Seoul) must start development at the same time. The back end won't be done for 2 weeks, but the front end must start tomorrow using a Mock server. A clear API Contract is needed right now.
GET /videos โ video list (filters: category, approval status, page)
POST /videos/{id}/play โ record playback start (auth required, triggers ad request)
POST /campaigns โ create an ad campaign (Advertiser only)
GET /ads/request โ request ad candidates (targeting by Video ID, User ID)
๐ง Questions to think about
What a contract means: how is an API contract different from a simple doc? What must it include to be a "contract"?
Auth scope: for POST /videos/{id}/play, does a "video viewable without an account" need auth? How is playback recorded?
Error code design: when there are no ad candidates, which is better โ HTTP 200 with an empty array or HTTP 404? How does it affect client code?
Idempotency: how should POST /videos/{id}/play be handled when called twice for the same request?
๐ก Example prompt (for reference โ don't use it as-is)
If OpenAPI 3.0 YAML is new to you, learn the structure from the example below. In the real assignment, all 4 endpoints must be designed to fit the exact business rules.
Example prompt โ HW-D04-01
[Role] You are an API design expert.
[Context] Front end and back end need an API contract to develop in parallel.
[APIs] GET /videos, POST /videos/{id}/play, POST /campaigns, GET /ads/request
[Include for each API]
1) OpenAPI 3.0 YAML spec (path, method, parameters, requestBody, responses)
2) JSON Payload examples for success (200/201) and failure (400/401/403/404/409/500)
3) Auth method and access conditions per role
4) Whether and how idempotency is handled
๐ฏ The real assignment โ do it yourself
1
Write OpenAPI 3.0 YAML (4 endpoints) Fully write path, method, summary, parameters/requestBody (with schema) and responses (at least 3 status codes) for each endpoint. Make it match the ERD you designed on Day 3.
2
Error code system reflecting business rules Design an error code system for the response Body (e.g. VIDACE_ERR_001: ad inventory used up). Define at least 8 error codes and the action the client should take for each.
3
Ownership and access control spec For each API, clearly define โ which roles can call it, โก how to check access when only your own resources are allowed, โข the response when accessing someone else's resource.
4
Write a Mock server guide Based on the OpenAPI spec above, write Mock response examples (success/failure) in JSON, pick one Mock server tool and briefly explain how to set it up.
๐ Grading (50 pts):Complete and accurate API contract 20 ptsConsistency between ERD and API schema 15 ptsError handling and ownership policy 10 ptsEvidence of AI judgment and self-critique 5 pts
DAY 4 ยท HW-D04-02 ยท 50 pts
๐ฅ Fix the Front End Broken by Back-End Breaking Changes
๐ Business scenario
The back-end team deployed these 5 changes without notice. The front-end app stopped working at once and user reports are coming in.
๐ฅ Change 5: method changed from GET /ads/request โ POST /ads/request
๐ง Questions to think about
Defining Breaking Changes: which of the 5 changes are true Breaking Changes, and which are acceptable?
Immediate response vs transition period: should you roll back right away, or build an Adapter Layer to support both versions?
HTTP 200 vs 404: why did changing no ads from 404 to 200 + empty array break the front end? Is it actually the "more correct" design?
Why contract tests: what tests or process should have existed beforehand to prevent this?
๐ก Example prompt (for reference โ don't use it as-is)
Think about Breaking Change impact from the Consumer's view. Analyze the impact of each of the 5 changes specifically.
Example prompt โ HW-D04-02
[Role] You are an API integration engineer.
[Context] The back-end team deployed 5 Breaking Changes without notice.
[Request]
1) For each of the 5 changes, analyze where front-end code breaks, the symptoms and an urgent fix
2) Compare the pros and cons of an immediate rollback vs an Adapter Layer
3) Explain a Consumer-Driven Contract Testing process to prevent this
4) Write 5 contract test cases (Given-When-Then format)
๐ฏ The real assignment โ do it yourself
1
Impact table for 5 Breaking Changes For each change, organize โ the front-end code/logic that breaks, โก user experience symptoms, โข urgent fix, โฃ scope of a permanent fix, in a table.
2
Choose and build a backward-compatibility recovery strategy Choose one of โ immediate rollback, โก API Versioning, โข Adapter/Gateway Layer, and write the reason and a list of real implementation changes.
3
Write 5 contract test cases Write 5 contract tests in Given-When-Then format. Each test must be able to detect a specific Breaking Change.
4
Suggest an API change management process Write an "API Breaking Change pre-review checklist" of at least 8 items, explaining in one line why each must be checked.
๐ Grading (50 pts):Specific Consumer impact analysis 20 ptsFeasible recovery strategy 15 ptsObservable contract tests 10 ptsEvidence of AI judgment and self-critique 5 pts
DAY 5 ยท HW-D05-01 ยท 50 pts
๐ฅ Design the Screens for a Local Trader Making Their First Ad
๐ Business scenario
VidAce decided to bring small local traders in Lagos on as new advertisers. The target user is "a local trader running a digital ad for the first time."
๐ค Persona: Amina (45, runs a food shop in Lagos)
Uses a smartphone but doesn't use apps much. Ad budget โฆ50,000 a month. Doesn't know ad terms ("What's CTR?"). Uses it where the network is unstable.
๐ Doesn't know the word "campaign" โ how do you guide her?
๐ Network drops while entering the budget โ is the input lost?
๐ File too big while uploading the ad creative โ how do you tell her and what help do you give?
๐ 3-day review wait โ is there a reason to come back to the app during this time?
๐ง Questions to think about
Word choice: what words can you use instead of industry terms like "campaign," "inventory" and "CTR"?
Kinds of UI states: besides Loading, Empty, Error, Disabled and Success, does this platform need its own UI states? What about "under review"?
Quality of error messages: how do "An error occurred" and "The file is too big. Only 10MB or less. [See how to compress a file]" feel different?
Handling unstable networks: what UX patterns are needed so input isn't lost? (Auto-save? Offline mode?)
๐ก Example prompt (for reference โ don't use it as-is)
For UI design, first draw the user's journey yourself. Sketch every step from Amina opening the app to making an ad on paper, then look at the example below.
Example prompt โ HW-D05-01
[Role] You are a UX designer and front-end architect.
[Context] Design mobile campaign creation screens for Amina (45),
a local trader with no digital ad experience.
[Request]
1) Describe in text the wireframe structure of each of 4 screens: "Start campaign" โ "Upload creative" โ "Set budget" โ "Submitted"
2) Write 6 UI states that can happen on each screen and a user message for each
3) Write a mapping table of the API endpoints and Request/Response fields linked to each UI component
[Note] Use language a local trader understands, without technical terms
๐ฏ The real assignment โ do it yourself
1
User Journey Map List the steps of Amina making her first ad. For each step, organize in a table โ what she does, โก how she feels (hope/worry/confusion, etc.), โข possible problems, โฃ what good design should give.
2
Wireframe spec for 3 screens Describe the UI components each screen needs as a hierarchy. Write all text in Amina's language, without technical terms.
3
Full spec of 6 UI states For each screen's main states (Loading, Error, Disabled, Empty, Success, Pending-under review), specify โ the exact message, โก the next action the user can take (recovery button), โข how long the state lasts or how to leave it.
4
API-UI mapping table Write a table linking the APIs designed on Day 4 to these screens: which API each UI event calls, the Request fields used, and the UI response to each Response status.
๐ Grading (50 pts):Complete journey design from the user's view 20 ptsSpecific UI states and recovery actions 15 ptsAccurate API links 10 ptsEvidence of AI judgment and self-critique 5 pts
DAY 5 ยท HW-D05-02 ยท 50 pts
๐ฅ Critique Your Own Screen Design and Make It Better
๐ Business scenario
VidAce's UX researcher ran a usability test with 5 local traders on the campaign creation screens designed in HW-D05-01. The results were shocking.
๐ฃ๏ธ Participant A: "I don't know what 'under review' means. Was my ad rejected?"
๐ฃ๏ธ Participant B: "I got a phone call while entering the budget, and when I opened the app again I had to start over."
๐ฃ๏ธ Participant C: "The upload failed, but there's no explanation of what file I should upload."
๐ฃ๏ธ Participant D: "I pressed Submit and nothing happened. When I pressed it again it said 'This campaign was already submitted.'"
๐ฃ๏ธ Participant E: "I want to see how much of my ad money was spent, but I don't know how to check after making it."
๐ง Questions to think about
Direction of critique: which design principle does each piece of feedback break? Which of Nielsen's 10 heuristics does it relate to?
Tech vs UX problem: is Participant B's problem (lost progress) a UX problem or a technical one? What must change to fix it?
The real problem in feedback: Participant A said they didn't know the word "under review," but is the real problem the word, or the way the state is explained?
Checking improvements: how can you confirm the improved screen is really better? What measures should you use?
๐ก Example prompt (for reference โ don't use it as-is)
Critique must be specific. First read the 5 pieces of feedback and find each root cause yourself. Then talk with AI.
Example prompt โ HW-D05-02
[Role] You are a UX Auditor.
[Context] 5 pieces of feedback were collected in a usability test of the campaign creation screens.
[Feedback to analyze] (paste the 5 feedback items)
[Request]
1) Classify the root cause of each feedback (UX principle broken or technical implementation problem)
2) Give an improvement for each in Before/After format
3) Write 3 usability test criteria to check the improvements
๐ฏ The real assignment โ do it yourself
1
Root cause analysis of 5 pieces of feedback For each, write โ the screen/component where the problem happened, โก the UX principle broken (with evidence like Nielsen's heuristics), โข severity (1โ5, by business impact), โฃ whether the cause is technical or UX.
2
Before/After screen spec For each of the 5 pieces of feedback, write the improved screen design in Before/After format. Explain how component structure, user flow and information hierarchy changed, based on UX principles.
3
3 usability recheck criteria Write 3 measurable criteria to confirm the improved screens are really better. Each must include a measurement method (observation, survey, completion rate, etc.) and a success threshold (number).
4
Design decision record Record separately the feedback you โ accepted, โก partly accepted (reason for not fully accepting), โข didn't accept (reason). For ones not accepted, also write the recheck condition.
๐ Grading (50 pts):Evidence-based problem judgment and cause analysis 20 ptsSpecific Before/After improvements 15 ptsMeasurable recheck criteria 10 ptsEvidence of AI judgment and self-critique 5 pts