Analysis Of Shopee’s Taiwan Store Group Product Selection Strategy And Practical Methods For Creating Hit Items

1. overview: treat store product selection and hot product creation as a "technology-driven" process

1) the traffic and conversion of store groups in shopee taiwan are affected by server response, cdn cache and domain name resolution;
2) the product selection and trial order stages require a stable environment to ensure reliable a/b test data;
3) traffic peaks when popular products are launched will amplify infrastructure shortcomings (bandwidth, concurrency, i/o);
4) the product selection process should be combined with monitoring, automatic expansion and contraction, and backup plans;
5) technical indicators (ttfb, 95th percentile delay, cache hit rate) directly affect advertising roi and conversion rate.

2. server and vps selection strategy (goal: cost controllable and availability)

1) node location selection: prioritize local taiwan or nearby asia-pacific nodes (taiwan/hong kong/singapore) to reduce latency;
2) cpu/ram configuration: 4vcpu/8gb is available for small-scale trial period, and 8vcpu/16gb is recommended for expansion period;
3) storage and iops: use nvme or high iops ssd when product details contain many pictures or videos;
4) bandwidth and billing: estimate based on peak bandwidth, set burst guarantee and monitor traffic;
5) backup and sla: choose a vendor that supports snapshots and a 99.95% sla, ensuring a recovery time objective (rto) ≤ 15 minutes.

3. domain name, dns and cdn deployment details (improve page loading and anti-brushing)

1) domain name strategy: the main store uses an independent primary domain name, and small store groups can be isolated by secondary domain names or paths to reduce management costs;
2) dns: use anycast dns and low ttl (60s in the trial phase, 300s in the stable phase) with grayscale release;
3) cdn: static resources and images must go through cdn. it is recommended that the cdn hit rate target be ≥ 80%;
4) https and certificates: automated certificate renewal (acme) avoids conversion losses due to certificate expiration;
5) caching strategy: product page cache ttl can be set to short cache (60–300s) + edge cache layer cache image 1d–7d.

4. ddos defense and high-availability architecture (to ensure stable sales period)

1) use a cloud protection/access layer with cleaning capabilities. the basic defense capability is recommended to be ≥ 100 gbps, and the critical period should have an elasticity of more than 200 gbps;
2) waf and rate limitation: add whitelist and frequency limit to login, order, and api interface;
3) multi-az/multiple computer rooms: database master-slave and application deployment across availability zones, switching time ≤ 30s;
4) health check and automatic failover: the load balancer health check lasts for 10 seconds, and exceptions are triggered and removed three times in a row;
5) monitor alarms: set bandwidth, request exception, error code (5xx) threshold alarms and connect with the operation and maintenance duty.

5. practical case: technical architecture and specific server configuration examples of a 10-store cluster

1) background: 10 small and medium-sized stores, average daily uv of 30k, peak concurrent requests on promotion days is about 1,200 rps;
2) target: page 95p latency < 300ms, cache hit rate ≥ 85%, origin bandwidth ≤ 1 gbps during peak period;
3) architecture: two-layer load balancing + 4 web services + db master-slave + redis cache + cdn + cloud protection;
4) operation and maintenance: automatic expansion and contraction strategy: horizontally expand a web when cpu > 65% or rps > 250;
5) results: through cdn and redis, the origin bandwidth was reduced from the original 3.2 gbps to an average of 420 mbps (hit rate 86%).

6. technical optimization steps and monitoring kpis for creating popular products (implementation operation list)

1) image and resource optimization: webp + lazy loading, page first screen resources ≤ 150 kb, ttfb ≤ 200ms;
2) dynamic caching rules: use edge cache + cache sharding strategy for high-traffic products to reduce back-to-origin;
3) back-end optimization: database index, query cache, and asynchronous order queue reduce blocking;
4) monitoring and rollback: set up ab test + indicators (ctr, conversion rate, refund rate) linkage for automatic rollback;
5) cost control: reduce traffic costs by more than 30% through cdn hit rate improvement and appropriate instance specifications.

7. summary and recommendations (deployment priority list)

1) prioritize ensuring the stability and low latency of domain names and dns;
2) use cdn as the first layer of protection and improve user experience;
3) establish automatic expansion and contraction and health check to deal with short-term traffic bursts;
4) configure cloud protection and waf, and conduct traffic drills for promotional days;
5) incorporate technical indicators into the product selection and hot product evaluation system, and use "environmental availability" as the threshold for product selection.