[{"data":1,"prerenderedAt":356},["ShallowReactive",2],{"blog-content-batch-quality":3,"blog-prev-content-batch-quality":123,"blog-next-content-batch-quality":242},{"id":4,"title":5,"author":6,"body":7,"canonical":107,"category":108,"cover":107,"date":109,"description":110,"draft":111,"extension":112,"keywords":107,"meta":113,"navigation":114,"noindex":111,"ogImage":107,"path":115,"seo":116,"seoDescription":107,"seoTitle":107,"stem":117,"tags":118,"updated":107,"__hash__":122},"blog\u002Fblog\u002Fcontent-batch-quality.md","批量内容的质量底线：模板化到什么程度算过头","元析智能",{"type":8,"value":9,"toc":96},"minimark",[10,14,18,21,24,27,30,33,36,39,42,45,48,51,54,57,60,63,66,69,72,75,78,81,84,87,90,93],[11,12,13],"p",{},"批量生产内容不是原罪，模板化过头才是。我们给批量内容定了四条底线：每篇有独家信息、结构可以有模板、案例不能重复、数据必须核实。",[15,16,17],"h2",{"id":17},"批量生产的两种做法",[11,19,20],{},"第一种，模板化生产：固定结构、固定句式，像填表一样写。效率高，但容易同质化。",[11,22,23],{},"第二种，模块化生产：固定的是\"模块清单\"（比如每篇必须有：问题、方案、数据、判断），但每个模块的内容独立写。",[11,25,26],{},"我们选第二种。效率比纯手工高 40%，质量比填表式高得多。",[15,28,29],{"id":29},"什么可以模板化",[11,31,32],{},"结构：每篇都用\"问题、方案、数据、判断\"四段，读者读起来顺畅。",[11,34,35],{},"格式：标题层级、列表、表格用法，统一。",[11,37,38],{},"流程：选题、审稿、发布的标准动作，统一。",[11,40,41],{},"这些模板化不会伤害内容，反而让读者更容易读。",[15,43,44],{"id":44},"什么绝对不能",[11,46,47],{},"第一，案例不能模板化。每篇必须是不同的客户、不同的场景。写\"某客户\"然后编一个通用案例，一眼就能看出来。",[11,49,50],{},"第二，数据不能模板化。每个数据都要有来源，不能\"大约\"\"业内人士估计\"。",[11,52,53],{},"第三，判断不能模板化。每篇的结论应该基于这篇的事实，不能所有文章都得出同一个结论。",[11,55,56],{},"第四，开头不能模板化。开头是读者决定读不读的地方，模板化开头会直接劝退。",[15,58,59],{"id":59},"一个越界的案例",[11,61,62],{},"2025 年 8 月，一个客户让我们批量写 200 篇\"行业解决方案\"页面，要求是同一套模板，换行业名。",[11,64,65],{},"我们评估后拒绝了。原因：200 个行业里，真正有搜索需求、有客户案例的不到 30 个。其余 170 个就是模板套壳，写出来也不会有排名，还会拉低整站质量。",[11,67,68],{},"后来只做了 30 个，每个都有真实案例和数据，效果很好。",[15,70,71],{"id":71},"我们的底线清单",[11,73,74],{},"每篇发布前问四个问题：这篇有没有别人写不出来的内容？案例是不是这个客户真实的？数据能不能找到来源？结论是不是这篇推导出来的？",[11,76,77],{},"四个都是\"是\"，发；有一个\"否\"，回去改。",[15,79,80],{"id":80},"一个提醒",[11,82,83],{},"批量内容的效率优势，建立在\"有足够素材\"的基础上。素材不够，批量就是批量制造垃圾。",[11,85,86],{},"我们一般先攒素材库：客户案例、实测数据、常见问题。素材够 100 篇的量，才开批量。",[15,88,89],{"id":89},"一句判断",[11,91,92],{},"批量不是错，错的是\"为了数量放弃判断\"。模板可以统一格式，不能统一思想。",[11,94,95],{},"200 篇平庸内容，比 30 篇有判断的内容，伤害大得多。",{"title":97,"searchDepth":98,"depth":98,"links":99},"",2,[100,101,102,103,104,105,106],{"id":17,"depth":98,"text":17},{"id":29,"depth":98,"text":29},{"id":44,"depth":98,"text":44},{"id":59,"depth":98,"text":59},{"id":71,"depth":98,"text":71},{"id":80,"depth":98,"text":80},{"id":89,"depth":98,"text":89},null,"内容运营","2026-10-27","批量生产内容不是原罪，模板化过头才是。我们给批量内容定了四条底线：每篇有独家信息、结构可以有模板、案例不能重复、数据必须核实。一个批量 200 篇翻车的案例和我们的检查表都在这篇，四条底线要守住。",false,"md",{},true,"\u002Fblog\u002Fcontent-batch-quality",{"title":5,"description":110},"blog\u002Fcontent-batch-quality",[119,120,121],"批量内容","内容质量","内容生产","AF6J5N8U_1ccSpPLqaP8s91isAKofOogCF2KJvUvygU",{"id":124,"title":125,"author":6,"body":126,"canonical":107,"category":230,"cover":107,"date":231,"description":232,"draft":111,"extension":112,"keywords":107,"meta":233,"navigation":114,"noindex":111,"ogImage":107,"path":234,"seo":235,"seoDescription":107,"seoTitle":107,"stem":236,"tags":237,"updated":107,"__hash__":241},"blog\u002Fblog\u002Fedge-cache-static-generation.md","静态化加边缘缓存，把首屏压进一秒内",{"type":8,"value":127,"toc":221},[128,131,134,137,140,143,146,149,152,155,158,161,164,167,170,173,176,179,182,185,188,191,194,197,200,203,206,209,212,215,218],[11,129,130],{},"2026 年 10 月我们给一个文档站做了静态化和边缘缓存，首字节从 900 毫秒降到 120 毫秒，首屏进入一秒内。改动不涉及业务逻辑，只是把渲染的结果提前算好，放到离用户近的地方。",[15,132,133],{"id":133},"静态化解决什么",[11,135,136],{},"动态渲染每次请求都要查数据、拼模板，再快也有几百毫秒。",[11,138,139],{},"静态化就是提前生成 HTML，请求来了直接返回文件。服务器压力也降下来了，一个静态文件能扛住大量并发。",[11,141,142],{},"对搜索引擎也有好处：返回稳定、内容完整、不依赖 JS。",[15,144,145],{"id":145},"哪些页面能静态化",[11,147,148],{},"营销页、产品页、文档页、博客文章、案例页，这些内容变化不频繁，全部可以。",[11,150,151],{},"不能静态化的是个性化内容：登录后的后台、购物车、实时数据看板。",[11,153,154],{},"中间状态是列表页和搜索页。可以静态化基础部分，动态部分用客户端填充，但要小心 CLS。",[15,156,157],{"id":157},"边缘缓存怎么配",[11,159,160],{},"把静态文件推到 CDN，缓存时间设长，用文件名哈希做版本。",[11,162,163],{},"HTML 的缓存时间设短，五分钟到一小时，配合主动刷新。",[11,165,166],{},"缓存键只保留路径，别把查询参数和 Cookie 算进去，否则命中率会很低。",[15,168,169],{"id":169},"内容更新怎么触发",[11,171,172],{},"三种方式。",[11,174,175],{},"第一，增量构建。内容变了只重新生成受影响的页面，几分钟完成。",[11,177,178],{},"第二，webhook 触发。CMS 发布时调用构建接口，全量或增量重建。",[11,180,181],{},"第三，短 TTL 兜底。就算触发失败，缓存到期也会自动回源，最多晚一小时。",[11,183,184],{},"三种一起用最稳，我们给客户的配置就是这样。",[15,186,187],{"id":187},"预热很重要",[11,189,190],{},"静态化之后第一批访问的用户，如果缓存还是冷的，体验和以前一样。",[11,192,193],{},"我们的做法是发布后主动请求一遍核心页面，把缓存预热。二十个页面，两分钟跑完。",[11,195,196],{},"预热脚本放在发布流程里，自动执行，别靠人记得。",[15,198,199],{"id":199},"一个常见的误解",[11,201,202],{},"静态化不等于不用服务器。构建、预热、缓存刷新都需要服务，只是请求量大时不再实时渲染。",[11,204,205],{},"还有一层容易忽略：静态文件如果没开压缩，传输体积会大两三倍。构建产物要先压好再上传，别指望 CDN 自己处理所有格式。",[15,207,208],{"id":208},"一个态度",[11,210,211],{},"静态化是 TTFB 优化里最彻底的一招，能到百毫秒级。但它的前提是内容能提前算好。",[11,213,214],{},"对于需要实时数据的站，静态化只能覆盖一部分页面，别硬上。",[11,216,217],{},"还有一个隐形成本：构建时间。几千页的站，全量构建可能要十几分钟，所以增量构建是必须的。",[11,219,220],{},"我们给客户的验收标准是：核心页面首字节低于 300 毫秒，首屏 LCP 低于 1.5 秒。做到这两个数，性能这条线基本不用再操心。",{"title":97,"searchDepth":98,"depth":98,"links":222},[223,224,225,226,227,228,229],{"id":133,"depth":98,"text":133},{"id":145,"depth":98,"text":145},{"id":157,"depth":98,"text":157},{"id":169,"depth":98,"text":169},{"id":187,"depth":98,"text":187},{"id":199,"depth":98,"text":199},{"id":208,"depth":98,"text":208},"技术SEO","2026-10-26","一个文档站改成静态生成加边缘缓存，首字节从 900 毫秒降到 120 毫秒。这篇写哪些页面能静态化、缓存怎么失效、内容更新怎么触发。",{},"\u002Fblog\u002Fedge-cache-static-generation",{"title":125,"description":232},"blog\u002Fedge-cache-static-generation",[238,239,240],"静态化","边缘缓存","TTFB","sy-zNNIR7TINDZIiu3RNeDIdS1F9nCyefVfmqTOHvMU",{"id":243,"title":244,"author":6,"body":245,"canonical":107,"category":108,"cover":107,"date":345,"description":346,"draft":111,"extension":112,"keywords":107,"meta":347,"navigation":114,"noindex":111,"ogImage":107,"path":348,"seo":349,"seoDescription":107,"seoTitle":107,"stem":350,"tags":351,"updated":107,"__hash__":355},"blog\u002Fblog\u002Fcontent-kpi-layers.md","内容 KPI 分四层，只看最后一层会做错决定",{"type":8,"value":246,"toc":334},[247,250,254,257,260,264,267,270,274,277,280,284,287,290,293,296,299,302,305,308,310,313,316,319,322,325,328,331],[11,248,249],{},"内容 KPI 只看注册数，会把该做的事全砍掉。我们把考核分四层：产出层、可见层、互动层、商业层。2026 年 1 月一个客户只考核注册，半年把文档内容全砍了，结果支持成本涨了 30%。",[15,251,253],{"id":252},"第一层产出","第一层：产出",[11,255,256],{},"发了多少篇、覆盖多少个主题、更新了多少篇旧文。这一层最容易被当成全部，它其实只是投入的计量。",[11,258,259],{},"它的用处只有一个：确认团队真的在干活。",[15,261,263],{"id":262},"第二层可见","第二层：可见",[11,265,266],{},"收录数、有曝光的页面数、关键词覆盖数、排名分布。这一层反映内容有没有被搜到，滞后大约两到三个月。",[11,268,269],{},"新项目的头三个月只能看这一层。",[15,271,273],{"id":272},"第三层互动","第三层：互动",[11,275,276],{},"点击率、阅读完成率、页面停留、内容页到功能页的点击。这一层反映内容有没有兑现点击。",[11,278,279],{},"互动层不好看，通常问题在标题或页面结构，改起来最快。",[15,281,283],{"id":282},"第四层商业","第四层：商业",[11,285,286],{},"注册、试用、咨询、成交。这一层滞后三到六个月，而且受销售、产品、定价影响，归因复杂。",[11,288,289],{},"只看这一层，等于用一个滞后的、多因素的数字来评价一个团队。",[15,291,292],{"id":292},"权重怎么定",[11,294,295],{},"我们建议新项目 2:3:2:3，成熟项目 1:3:3:3。产出层权重下调，是因为成熟期不需要靠篇数证明什么。",[11,297,298],{},"具体数字可以调，关键是四层都要有，不能只留一层。",[15,300,301],{"id":301},"那个被砍掉文档的案例",[11,303,304],{},"2026 年 1 月，客户只看注册，发现文档页带来 0 个注册，把文档团队砍了两个人。",[11,306,307],{},"三个月后客服工单涨了 40%，销售在演示时找不到可发的链接。7 月又把文档恢复了，重新招人。",[15,309,80],{"id":80},[11,311,312],{},"KPI 分层不是为了让考核好看，是为了让决策看得到过程。只考核结果，过程就会塌。",[15,314,315],{"id":315},"怎么跟销售对齐",[11,317,318],{},"商业层的数要和销售一起看。内容带来的线索，销售跟进后成交与否，把结果回填到内容页，这个数据比任何报表都有说服力。",[11,320,321],{},"回填频率一个月一次就够，太频繁没人做。",[15,323,324],{"id":324},"一个小团队的做法",[11,326,327],{},"没有数据团队的时候，四层里先抓两个：有曝光的页面数、内容页到试用的转化。",[11,329,330],{},"这两个数手工能算，一个月花半天，足够判断方向对不对。",[11,332,333],{},"内容这件事，过程塌了，结果一般是半年后才显出来。",{"title":97,"searchDepth":98,"depth":98,"links":335},[336,337,338,339,340,341,342,343,344],{"id":252,"depth":98,"text":253},{"id":262,"depth":98,"text":263},{"id":272,"depth":98,"text":273},{"id":282,"depth":98,"text":283},{"id":292,"depth":98,"text":292},{"id":301,"depth":98,"text":301},{"id":80,"depth":98,"text":80},{"id":315,"depth":98,"text":315},{"id":324,"depth":98,"text":324},"2026-10-28","只看注册数，会把该做的事全砍掉。我们把内容考核分四层：产出、可见、互动、商业。前两层看新项目，后两层看成熟项目，权重我们定成 2:3:2:3。",{},"\u002Fblog\u002Fcontent-kpi-layers",{"title":244,"description":346},"blog\u002Fcontent-kpi-layers",[352,353,354],"内容评估","KPI","数据指标","We9oActN48bjEDehS2pY6U2ONDMB4mQNpjrZbQE9Uog",1794088100315]