Test Case Design Principles — From Real Luxury Scenarios to Synthetic PUNTO Cases
Artefact · Applied AI Engineer Written Challenge · Step 2 — Test Case Design
Test Case Design Principles
From real luxury scenarios to synthetic PUNTO cases — what the research actually teaches, and what each lesson becomes as a test.
Companion to the luxury reference page (04). Brand: PUNTO, fictional leather house. Price ladder fixed: Tier 1 RMB 4,200–9,800, Tier 2 RMB 22,800–34,800, Tier 3 RMB 48,500. Date: 3 November 2026.
The derivation rule for this whole report
Real sources decide what risk to test . They never supply a test case. No real client name, private chat, real purchase record, internal CRM record, real brand product or real brand copy is used in any case below. Every client, history, product and conversation in the sketches is invented, and every PUNTO product and price is taken only from the fixed PUNTO table.
Derivation chain, used in every section: Real source → Observed principle → PUNTO rule → Test risk → Brief scenario → Hard fail if the tool breaks it.
1. Luxury client journey — the map the six brief scenarios sit on
The luxury journey is usually described in five stages: awareness, consideration, purchase, post-purchase engagement, and loyalty / advocacy. Luxury clients are described as meticulous researchers who gather information before buying, and as emotional decision-makers who seek exclusivity, pleasure and self-expression, not price or function alone. The brief’s six scenarios are not the journey itself — they are the moments a Client Advisor (CA) writes in, at different journey points.
Journey stage What the client is doing (research) Brief scenario that writes at this stage What a good WeChat message must do here
Awareness Discovers through search, social, friends, boutique window Event invitation (first touch, low pressure) Give one reason the event fits this client, not a mass blast
Consideration Researches, compares models, checks price, asks an advisor Product recommendation; Price objection Answer the exact use and capacity question; compare inside PUNTO, never invent a competitor fact
Purchase Chooses, often after a boutique visit or advisor hold — (no brief scenario is the checkout itself) —
Post-purchase Uses, cares, notices wear, needs reassurance Post-purchase care Care that fits the actual material and use; boutique assessment for anything beyond wiping dry
Loyalty / advocacy Returns, gifts, celebrates, brings a friend Birthday / festive greetings; Event invitation (returning client) Remember the person, not just the purchase; make the next step easy to decline
Silence / churn (journey breaks) Stops visiting or buying, for a reason the brand may not know Reactivation Restore contact first, sell second — or not at all in that message
Reference links — Client journey
[https://blog.osum.com/luxury-consumer-behavior/] Luxury consumer behaviour — meticulous research, emotional decision-making, journey stages awareness, consideration, purchase, post-purchase engagement
[https://www.shopify.com/hk-en/blog/customer-engagement-model] Customer engagement model — awareness, consideration, conversion, retention, advocacy; high-touch model for luxury, one-to-one before purchase and after purchase
2. CRM lifecycle — acquisition, engagement, retention, reactivation
Luxury CRM job descriptions describe the same lifecycle in the same words: from acquisition and onboarding to retention, reactivation and advocacy, segmented by value, behaviour and preferences, with KPIs for engagement, repeat rate, churn and lifetime value. Clienteling is the one-to-one layer on top of CRM: bespoke communications, collection previews, private appointments and milestone moments, delivered by a named advisor.
CRM stage What CRM does in real luxury retail PUNTO test variable (vary this, not the copy) Risk if the tool gets it wrong
Acquisition First contact, often after a boutique visit, referral or event; welcome, not a hard sell New client, no purchase yet, one stated interest Assumes a taste or budget the client never stated
Engagement One-to-one outreach with a genuine reason: a new colour in an owned line, a private preview, a care check Client with one owned PUNTO piece and one stated use Recommends a product that does not fit the stated use or capacity
Retention Milestones, birthdays, festivals, care and repair; keep the client, do not discount the relationship Loyal client, 2+ purchases, birthday or festival in the next 2 weeks Turns a greeting into a promotion; invents a gift the client did not ask for
Reactivation Win-back after silence; first message has no offer, asks why, or offers a non-selling reason to return Silent 6 months, silent 18 months, previously declined marketing, previously had a service problem Direct selling to a long-silent or marketing-averse client; guilt or pressure
Advocacy Client brings a friend, gifts, attends and returns; invitation and recognition, not a referral bribe Client who gifted before, or who attended an event before Asks for a referral as a transaction; promises a reward that does not exist
Reactivation principle, in detail: win-back research stresses timing by exit reason, not by calendar alone — a message timed to a life event or a natural re-engagement window performs better than a calendar blast, and the first diagnostic message should contain no offer. A good win-back sequence is check-in, then useful tip about a product already owned, then an open door — and if there is no reply, stop. Discount ladders are explicitly warned against: they train clients to wait for a code and devalue the brand. For luxury, the non-discount equivalent is a care check, a repair assessment, or a private preview — never a price cut.
Reference links — CRM lifecycle and win-back
[https://career.io/job/pomellato-crm-manager-americas-new-york-kering-group-0121a440732833b8e0078c603f5d619e] Luxury CRM lifecycle — acquisition and onboarding to retention, reactivation and advocacy; segmentation by value, behaviour and preferences; previews, private appointments, milestone moments
[https://www.rajivgopinath.com/blogs/marketing-hub/consumer-experience-and-engagement/cx-and-retention-strategies/retention-and-loyalty/win-back-campaigns-that-dont-feel-desperate] Win-back timing by exit reason and life event, not calendar; regained clients and lifetime value; timing as the difference between welcome and intrusive
[https://www.campaignrequest.com/pulse/winback-flows-that-work?hs_amp=true] Win-back flows — define lapse per category, diagnostic message with no offer first, prefer non-discount value, graceful sunset after no reply
3. Product information structure — what a message may quote
Official product pages, across luxury and leather houses, use the same field structure. That structure becomes the input schema for every test case, and the groundedness rule for the tool: a message may quote only a field that is present in the input for that case.
Field (from public product pages) Example in PUNTO terms Tool rule
Material Grained calf, smooth calf, vegetable-tanned calf Quote the material in the input; never upgrade it to exotic
Dimensions / capacity What fits: phone, wallet, keys, notebook, 13-inch laptop Capacity is a fit claim — if it does not fit, do not recommend it
Compartments / closure Card slots, zip coin pocket, zip or flap closure Use only compartments listed for that product
Carry Hand, shoulder, crossbody, detachable strap Match carry to the stated use (commuting, travel, evening)
Care Wipe dry with a soft cloth; keep from prolonged water, heat, light Safe care only; anything else is a boutique assessment, not a home remedy
Price tier PUNTO Tier 1 / 2 / 3, RMB, fixed table Quote the fixed price or band only; never a discount, never a “special price”
Service Boutique care assessment, stitch repair, edge renewal No timeline, fee or guaranteed outcome in a message
If a field is missing from a test input, the correct behaviour is to omit it, not to fill it. That omission is itself a test: a tool that invents a size, a capacity or a repair time fails groundedness, even if the message reads well.
Reference links — Product information structure
[http://steelhorseleather.com/products/the-bard-weekender-handmade-leather-duffle-bag?_kx=WpLsIARBtU2XfR8fHyTU4qoniBkpVT5E_dnVkxvMnzBQZkpB_knWWT9vxaHnigLk.RzexA2] Leather product specification pattern — material, dimensions, capacity, compartments, carry, care instructions
[https://himileathers.com/products/ladies-leather-laptop-bag] Product specification table pattern — material, laptop capacity, compartments, closure, handle, dimensions, weight, hardware, care
4. Public customer-service scenarios — the service patterns to test
After-sales service (post-purchase care)
The classic luxury after-sales pattern, described in service literature, is: greet the client as warmly as any other client, ask open questions and listen, let the client express the emotional part of the frustration, take ownership (“here is what I can do for you”), agree what will be done and what it will cost, update the client’s information for recontact, and follow up when the service is ready. For PUNTO, that pattern is adapted, not copied: the message must not promise a timeline or a fee, and must not diagnose damage from a photo or a description. The PUNTO version is: acknowledge, ask one clarifying question about use, offer a boutique assessment, and leave the next step with the client.
Price objection
Price-objection research separates the stated objection from the root cause: budget, fear of a costly mistake, comparison with elsewhere, or a value question. The recommended sequence is acknowledge and empathise, probe for the root cause, then reinforce value — after-sales service, durability, repair and daily use — rather than discounting. For PUNTO, the value answer is: work hours and stitching are not claimed (they are not in the input), but material, capacity, daily use, patina and lifetime boutique care assessment are. The step-down answer is a lower PUNTO tier, never a lower price for the same piece.
Gifting and events
Exclusivity in current luxury research is moving from simple scarcity to meaningful access: useful knowledge, trusted guidance, or a privilege that fits the relationship. A thoughtful invitation or useful information works better than a noisy event. For PUNTO, an event invitation must therefore state what the client will actually see or do, give a clear time and place, and make declining easy — and must not invent scarcity (“only a few places left”) or a guest the client did not ask to bring.
Reference links — Service scenarios
[https://www.oreilly.com/library/view/selling-luxury-connect/9780470457993/ch88.html] Ideal after-sales service scenario — warm greeting, open questions, ownership, agreed intervention, recontact information, follow-up when ready
[https://fastercapital.com/content/Price-Objection--Turning-Price-Concerns-into-Sales-Opportunities.html] Price objection — budget, fear of mistake, social proof, after-sales value; shift from cost to benefits, not to discount
[https://blog.living-luxe.shop/exploring-the-thriving-luxury-accessories-market/] Exclusivity as meaningful access — early access, preview, trusted guidance; aftercare as cleaning, repair intake and service reminders
5. Chinese WeChat communication — what the channel itself demands
WeChat is not a broadcast channel in China. It is the operating system of daily life — messaging, payment, e-commerce, service and loyalty in one app, with over 1.3 billion monthly users — and the luxury use of it is private-domain: a named advisor, a WeCom contact, a Mini Program, and a one-to-one relationship the brand owns. That structure creates both the opportunity and the risk the test cases must cover.
WeChat principle (research) PUNTO rule Test it in
Private domain, named advisor, 1:1 — not a mass blast Every message is from one named CA to one named client, with one reason All 6 scenarios
Attention is limited; always-on culture creates fatigue One message, one ask, no follow-up pressure in the same message Reactivation, event invitation
Useful, timely, context-aware — not frequent or intrusive No message without a trigger: birthday, festival, care need, new colour in an owned line, event that fits All 6 scenarios
Private-message etiquette: greeting + specific reason + value + closing; avoid late night and weekend peak hours Message structure and timing are part of the input; a late-night send is a test variable, not a default Birthday, reactivation
Research-first: buyers search an exact model on Xiaohongshu, compare owner notes, check price, then talk to an advisor Recommendation must survive comparison: state material, capacity and price tier, so the client does not need to ask twice Recommendation, price objection
Shift from status display to self-expression, cultural consumption and value for daily use No status language; scenario and use are the reason, not wealth or rank All 6 scenarios
Younger, self-purchase growth; gifting still matters, but personal reward and milestone gifting are distinct from business gifting Separate test types: self-purchase, gift for a named person with a known taste, gift with unknown taste Birthday, recommendation
Reference links — WeChat and Chinese luxury consumers
[https://fashionchinaagency.com/wechat-crm-why-fashion-luxury-brands-in-china-have-to-get-one/] WeChat CRM and private domain — named advisor, WeCom, Mini Program, owned direct relationship; 1.382 billion users, opened 9–11 times a day
[https://fashionchinaagency.com/2017-luxury-brands-china-still-underestimate-wechat/] WeChat for luxury in 2026 — two-way relationship infrastructure, personalised communication, VIP group management, human not automated
[https://www.gogetop.com/Blog/wechat-always-online-culture.html] WeChat always-on culture — digital fatigue, limited attention, brands should be useful, timely and context-aware, not frequent or intrusive
[https://www.theredscroll.com/industries/luxury/] China luxury 2025 — from aspiration to connoisseurship, buyers pay for what they know, mainland spending 65% in 2025, younger buyers more cautious
6. The six brief scenarios — principle, risk and synthetic sketch
Each scenario below has one core risk, the input variables that vary it, and a short synthetic sketch. Sketches are examples of case design , not finished test cases — the full 50-case set will expand each sketch into several cases with different clients, products and histories.
Scenario 1 — Birthday and festive greetings
Principle: CRM retention and milestone moments. Risk: greeting becomes promotion; or the tool invents a personal detail (a child, a job, a taste) the input never gave. Vary: birthday vs Chinese New Year vs Qixi / 520 vs Mid-Autumn; loyal client vs new client; self-purchase history vs gift history; client in Shanghai, Shenzhen, Guangzhou, Hangzhou or Nanjing.
Sketch (synthetic): Client “Ms Chen”, Shanghai, bought a PUNTO Crossbody last spring for commuting, birthday in 10 days. Fail if: the message offers a discount, invents a gift for a partner the input never mentioned, or pressures a reply by the birthday.
Scenario 2 — Product recommendation
Principle: Consideration stage, research-first buyer. Risk: recommending a bag that does not fit the stated use, or inventing a preference (“you love red”) to justify it. Vary: commuting (laptop fit) vs travel vs evening; stated colour or material vs unstated; budget tier stated vs unstated; client already owns a PUNTO piece in a line vs owns nothing.
Sketch (synthetic): Client “Mr Zhou”, Shenzhen, commutes by metro, carries a 13-inch laptop and a notebook, no colour preference stated. Fail if: the tool recommends the Evening Pouch (phone only) for a laptop commute, or invents a colour preference.
Scenario 3 — Post-purchase care
Principle: After-sales service pattern, adapted for PUNTO. Risk: home remedy, diagnosis from a description, or a promised timeline / fee. Vary: water mark vs edge wear vs stitch loosening vs hardware; recent purchase vs 2-year-old piece; client anxious vs client calm; client far from a boutique vs near one.
Sketch (synthetic): Client “Ms Lin”, Hangzhou, bought a PUNTO Daily Tote 8 months ago, caught in rain, water mark on grained calf. Fail if: the message tells her to apply a cream, polish or home treatment, or promises “ready in a week”.
Scenario 4 — Reactivation
Principle: CRM reactivation, win-back without discount. Risk: direct selling to a silent or marketing-averse client, guilt, or treating all silence as the same. Vary: silent 6 months vs 18 months; previously declined marketing vs simply busy; previously had a service problem vs no problem; owned piece that needs care vs owned piece with no care need.
Sketch (synthetic, from the earlier principle): Client “Ms Lin” (a different synthetic client from Scenario 3, same first name is a coincidence in the sketch set — full cases will use distinct surnames), 18 months without a purchase, last conversation said she is not considering adding anything soon. Fail if: the message opens with a new launch or a price, instead of a care check or a soft, declinable reason to reconnect.
Scenario 5 — Event invitation
Principle: Meaningful access, not scarcity. Risk: invented scarcity, invented guest list, or an invitation that does not fit the client’s city, schedule or taste. Vary: boutique atelier preview vs private preview of a new colour; client in the boutique city vs travelling; client attended before vs never attended; client may bring a friend vs invitation is single-client.
Sketch (synthetic): Client “Ms Wang”, Guangzhou, attended a PUNTO care workshop last year, owns a Top-Handle. Fail if: the message says “only a few places left” when the input gives no capacity, or invites her to a Shanghai event when the input says she is in Guangzhou with no travel stated.
Scenario 6 — Price objection
Principle: Price objection as a root-cause question, answered with value and a tier step-down, never a discount. Risk: discount, competitor disparagement, or a value claim the input cannot support. Vary: “too expensive” vs “cheaper elsewhere” vs “can you discount?”; client compares with a named real brand (tool must not disparage it, and must not invent its price) vs compares with no brand; client budget is Tier 2 vs Tier 1.
Sketch (synthetic): Client “Mr Liu”, Nanjing, likes the PUNTO Top-Handle at RMB 48,500, says it is higher than he expected and asks about a discount. Fail if: the message offers any discount, says another brand is worse, or claims a stitching time or craftsman count not in the input.
7. Principle matrix — 20 principles that become test requirements
# Principle (from real scenarios) PUNTO rule Tested in
P1 One named advisor, one named client, one reason No mass-blast language; no “dear clients” All
P2 Research-first buyer compares before buying State material, capacity and price tier so the message survives comparison Rec., Price
P3 Personalisation uses only input facts Never invent a taste, job, family member or past purchase All
P4 Emotional decision, self-expression over status No status, wealth or rank language All
P5 Attention is limited; WeChat fatigue is real One ask, no pressure, easy decline Reactivation, Event
P6 Greeting + specific reason + value + closing WeChat structure, warm and concise, not an essay All
P7 Milestone moments are retention, not promotion Birthday / festival message may mention a product only if the input gives a reason Birthday
P8 Capacity is a fit claim Do not recommend a bag that cannot hold the stated items Recommendation
P9 After-sales: greet warmly, listen, own the problem Acknowledge, ask one clarifying question, offer boutique assessment Care
P10 No diagnosis or home remedy from a description Safe care only: wipe dry, keep from water / heat / light; else boutique Care
P11 Long silence needs a soft reason, not a sale Care check or preview first; no launch-led reactivation Reactivation
P12 Silence has different reasons — branch on reason Declined marketing, busy, service problem, moved city are different cases Reactivation
P13 Win-back first message has no offer No discount, no “special price”, no urgency in reactivation Reactivation
P14 Access, not scarcity, is the event value State what the client will see / do; never invent capacity or scarcity Event
P15 Price objection: acknowledge, probe, then value Ask what feels high — budget, comparison or value — before answering Price
P16 Value is daily use, material, patina and lifetime care Never a discount, never a competitor put-down, never an invented craft claim Price
P17 Tier step-down is the no-discount answer Top-Handle → core bag → Tier 1 piece, in PUNTO’s fixed ladder Price
P18 Self-purchase and gifting are different intents Gift cases must state recipient taste or state that it is unknown Birthday, Rec.
P19 Festival and city matter in China Use the festival and city in the input; never move a client to another city Birthday, Event
P20 Silence after no reply is a boundary Tool output is one message set, not a chase sequence; no “second reminder” pressure Reactivation
8. How the 50 cases are distributed from these principles
Distribution, to be confirmed before generation — the same distribution proposed earlier, now justified by the principles above:
Scenario Cases Must cover
1 Birthday / festive 9 Birthday, Chinese New Year, Qixi / 520, Mid-Autumn; loyal, new, gift-buyer, self-buyer
2 Product recommendation 9 Commuting, travel, evening; laptop fit, no fit, stated taste, unstated taste, owned-line colour
3 Post-purchase care 8 Water, edge, stitch, hardware; recent, 2-year-old; anxious, calm; near boutique, far from boutique
4 Reactivation 8 Silent 6 months, silent 18 months, declined marketing, service problem, moved city
5 Event invitation 8 Atelier preview, private colour preview; attended before, never attended; same city, travelling
6 Price objection 8 Too expensive, cheaper elsewhere, asks for discount; Tier 2 budget, Tier 1 budget; named competitor, no competitor
Total 50 Every principle P1–P20 appears in at least 2 cases; every PUNTO product appears at least once; every price tier appears in at least 5 cases
Client cities for the synthetic set: Shanghai, Shenzhen, Guangzhou, Hangzhou, Nanjing — the priority cities from the China search plan. No case uses a real person, a real address, a real phone number, or a real purchase record.
9. What is deliberately NOT in the test set
No real client names, private WeChat chats, real purchase records, or internal CRM data.
No real brand products, prices or copy as PUNTO material — real brands appear only in this report, as reference.
No invented stock, discount, repair timeline, repair fee, event capacity, or craftsman count.
No case that requires the tool to know something the input does not give — if the input is silent, the correct output is silent on that point.
No case that can only be passed by flattery, status language, scarcity or pressure — those are fails, not styles.
Step 2 report: test case design principles derived from real luxury client journey, CRM lifecycle, product information structure, public service scenarios, the brief’s six scenarios, and Chinese WeChat research. Cases themselves are synthetic and fictional. Next step, when confirmed, is the full 50-case input matrix built from Section 8, using only PUNTO’s fixed products and prices.