全球一致不是总部的审美偏好那么简单。一致的产品结构让问题可以被比较，让质量标准不会在每个市场各写一版。品牌如果在每个市场说一套互相矛盾的话，用户和合作方都会失去参照。成本、测试和发布如果完全按市场拆开，规模带不来效率，只会带来更多版本。可维护性是这里最容易被低估的一点：今天为了迁就一个市场写进主干的例外，过一段时间会变成没人敢动的部分。

没有一个稳定内核，每个新市场都像第一次做产品。那不是本地适配，是反复重来。一致所保护的，是产品还能被理解、被测试、被继续做下去。

Global consistency is not just a headquarters preference for how things should look. A shared product structure lets problems be compared, so the quality bar is not rewritten in every market. If the brand says contradictory things in each market, users and partners lose their reference. If cost, testing and release are split completely by market, scale does not buy efficiency. It buys more versions. Maintainability is the part most easily underestimated: an exception written into the core to accommodate one market becomes, later, a part nobody dares to touch.

Without a stable core, every new market is a first product. That is not local adaptation. It is starting over. What consistency protects is the chance to keep understanding, testing and continuing the product.

另一边的理由同样具体。规则不同，一个在原地能走通的流程，换一个市场可能就不能用。用户完成同一任务的方式不同。数据的有无和更新节奏不同。渠道和服务能力不同。语言和文化会改变人如何理解提示、风险和责任。当地的竞争也会改写“够用”的标准。

如果这些差异被一句“全球统一”盖过去，统一往往只是对总部方便。产品在计划里仍然是一个版本，在用户手里却已经不可用，或者要靠当地团队用看不见的人工把它撑住。适配不是为了证明市场特殊，而是为了产品在那个市场里仍然成立。

The case for adaptation is just as concrete. Rules differ, so a flow that works in one place may be unusable in another. People complete the same task differently. Data may or may not exist, and it may update at a different pace. Channels and service capacity differ. Language and culture change how people read a prompt, a risk or a responsibility. Local competition rewrites what counts as good enough.

If those differences are covered by the phrase “keep it global,” the unity is often convenient mainly for headquarters. The plan still shows one version. In the user’s hands the product no longer works, or a local team is holding it up with manual effort nobody sees. Adaptation is not a claim that the market is special. It is what keeps the product real in that market.

所以问题通常不是选择标准化，还是选择本地化。需要逐层问的是：什么必须统一，否则产品不再是同一个产品；什么可以配置，让差异被容纳而不必另起一套；什么必须本地化，否则在那个市场不能用；什么只能由当地判断，因为证据和后果都在当地；什么应该回到总部重新学习，因为当地看到的不是例外，而是原来的全球假设本身有问题。

在汽车软件里，这组问题会变得很具体，但并不因此需要拿出某个客户或某款车才能讨论。导航、地图和车载功能要在不同区域里保持可识别的任务和基本质量，同时又要面对道路、数据、法规和使用习惯的差别。哪些放进统一的产品结构，哪些做成区域可以调整的部分，哪些必须按当地条件重做，是持续发生的产品判断。长期做这类产品的人，会反复碰到同一类边界。公开这些边界如何被判断，并不需要公开谁的项目、哪一款产品或哪一份不能说的材料。

The question is usually not whether to choose standardization or localization. It is, layer by layer: what must stay unified or the product is no longer the same product; what can be configured so difference is absorbed without starting over; what must be localized or the product cannot be used in that market; what only the local team can judge, because both the evidence and the consequences are local; and what headquarters needs to learn again, because the local observation is not an exception but a flaw in the global assumption.

In automotive software these questions become concrete without requiring a named customer or a named vehicle. Navigation, maps and in-vehicle functions have to keep a recognizable task and a basic quality bar across regions, while roads, data, regulation and habits of use differ. What sits in the shared product structure, what a region can adjust, and what has to be redone for local conditions, is an ordinary product judgment. People who work on this kind of product meet the same class of boundary again and again. Explaining how the boundary is judged does not require naming whose project, which product or which material cannot be discussed.

这组边界也不会一次划定就永久有效。市场更成熟以后，原来必须特殊处理的部分，可能已经可以收进可配置的范围。产品自己变了，旧的全球内核可能不再合适。政策一变，昨天还能全球一致的流程，今天就必须在当地拆开。组织能力提升以后，总部可以学会原来只能当地掌握的判断，当地也可以承接原来只能由总部发布的部分。

把边界写成永久政策，和完全不划边界，是同一种偷懒。前者假装已经想完了，后者假装每次都可以重新吵一遍。真正要保持的，是复查边界的习惯：什么变了，哪一层判断需要重做，谁有新的证据。

在需要长期维护的产品里，这种移动很常见。数据覆盖变了，规则对地图或定位的要求变了，或者一个区域的使用者从少数专业用户变成大量日常用户，原来放在“必须特殊处理”里的部分，可能要重新放回可配置的结构。反过来也一样：原来以为可以全球共用的假设，会在新的使用方式里露出缺口。

全球产品能力并不是把所有市场做成一样，也不是把每个市场都做成例外，而是建立一种能够持续判断“一致与差异边界”的能力。

The boundary does not stay valid once it is drawn. As a market matures, something that had to be handled as a special case may fit inside what can be configured. When the product itself changes, the old global core may no longer be the right one. A policy change can force apart, locally, a flow that was globally consistent yesterday. As the organization gets better, headquarters can learn judgments that only the local team could make, and the local team can carry work that only headquarters used to release.

Writing the boundary as permanent policy, and refusing to draw one at all, are the same kind of shortcut. The first pretends the thinking is finished. The second pretends the argument can be restarted every time. What has to be kept is the habit of reviewing the boundary: what changed, which layer of judgment has to be redone, and who now has new evidence.

In a product that has to be maintained for a long time, this movement is ordinary. Coverage of the data changes. Rules for maps or positioning change. A region’s users shift from a small professional group to everyday use at scale. Something that sat in “must be handled as a special case” may need to move back into what can be configured. The reverse happens too: an assumption that looked globally shared shows a gap in a new pattern of use.

A global product capability is not making every market the same, and it is not making every market an exception. It is the ability to keep judging where consistency ends and difference begins.