What is digital marketing analytics?什么是数字营销分析？
Digital marketing analytics is the disciplined collection, definition, comparison, and interpretation of digital marketing data so a team can evaluate performance and make testable decisions about channels, campaigns, audiences, creative, and budget. It connects activity metrics such as impressions and clicks to business outcomes such as qualified pipeline, orders, contribution margin, or retained customers.
数字营销分析是对数字营销数据进行有纪律的采集、定义、比较与解释，使团队能够评估表现，并围绕渠道、活动、受众、创意和预算作出可检验的决策。它把展示、点击等活动指标连接到合格商机、订单、贡献利润或留存客户等业务结果。
A dashboard is only one output. A credible marketing analytics process begins with a decision, specifies the eligible population and measurement window, reconciles data across systems, chooses a method whose assumptions fit the question, and records what evidence would change the conclusion. It distinguishes what was observed from why it may have happened. Attribution models describe how credit is assigned under a rule; experiments address whether an action caused an incremental effect.
仪表板只是输出之一。可信的营销分析流程从决策开始，明确合格总体和衡量窗口，跨系统对账，选择假设与问题相符的方法，并记录哪些证据会改变结论。它区分“观察到了什么”和“为什么可能发生”。归因模型描述在某种规则下如何分配功劳；实验则用于判断某项行动是否造成增量效果。
Why marketing analytics matters—and where it stops营销分析为何重要，以及它的边界在哪里
Teams search for marketing analytics when platform reports disagree, campaign totals do not reconcile to revenue, or a channel appears efficient until returns, margin, sales qualification, or retention are included. The discipline creates a common measurement language for marketing, finance, sales, product, and data teams. That shared language is useful for deciding whether to continue a campaign, change a message, shift budget, investigate an audience, or run an experiment.
当平台报表彼此矛盾、活动总数无法与收入对账，或某渠道在计入退货、利润、销售资格与留存后不再高效时，团队会需要营销分析。它为市场、财务、销售、产品与数据团队建立共同的衡量语言，用来决定是否继续活动、修改信息、调整预算、调查受众或开展实验。
Compare campaigns and channels, locate funnel loss, explain a metric change, evaluate audience or creative patterns, and monitor experiments.
比较活动与渠道、定位漏斗流失、解释指标变化、评估受众或创意模式，并监控实验。
Estimate scenarios, set measurement plans, choose a reporting grain, and inform budget choices with explicit uncertainty and constraints.
估算情景、制定衡量计划、选择报告粒度，并在明确不确定性与约束的前提下支持预算选择。
A correlation, path, or attributed conversion does not reveal what would have happened without the marketing exposure.
相关性、路径或被归因的转化无法说明没有营销曝光时会发生什么。
Market research studies markets, needs, attitudes, and competitors using primary and secondary evidence. Marketing analytics primarily studies performance data from marketing operations and customer journeys.
市场研究用一手和二手证据研究市场、需求、态度与竞争者；营销分析主要研究营销运营和客户旅程中的表现数据。
Marketing analytics versus web analytics: web analytics concentrates on website and app acquisition, behavior, and conversion. Marketing analytics can include that evidence but also joins advertising cost, email, CRM stages, sales outcomes, offline campaigns, experiments, and finance data. Product analytics focuses on in-product behavior and value; customer journey analytics follows interactions across touchpoints. The fields overlap, but the decision owner and outcome determine which lens leads.
营销分析与 Web Analytics 的区别：Web Analytics 侧重网站和应用的获客、行为与转化；营销分析可以纳入这些证据，同时连接广告成本、邮件、CRM 阶段、销售结果、线下活动、实验与财务数据。产品分析侧重产品内行为与价值，客户旅程分析则跟踪跨触点互动。它们会重叠，但应由决策负责人和结果指标决定主要视角。
Prepare marketing analytics data before building a dashboard搭建仪表板前，先准备营销分析数据
Start with a measurement brief and a source inventory. A brief names the decision, owner, population, target behavior, time window, comparison, and acceptable uncertainty. The inventory records where spend, exposure, interaction, identity, consent, conversion, revenue, margin, and retention live. The Google Analytics data-collection overview explains how website and app interactions become reports; it is one source in a broader marketing system, not a complete cross-channel ground truth.
先建立衡量简报和数据源清单。简报写明决策、负责人、总体、目标行为、时间窗口、比较方式与可接受不确定性；清单记录花费、曝光、互动、身份、同意状态、转化、收入、利润与留存分别位于何处。Google Analytics 数据采集概览说明网站和应用互动如何进入报告；它只是更广泛营销系统中的一个来源，并非完整的跨渠道事实来源。
| Data class数据类别 | Typical fields常见字段 | Critical validation关键验证 |
|---|---|---|
| Media and campaign媒体与活动 | Spend, impressions, clicks, campaign, ad group, creative, audience, platform conversion花费、展示、点击、活动、广告组、创意、受众、平台转化 | Currency, timezone, account scope, refunds, attribution window, API backfills币种、时区、账户范围、退款、归因窗口、API 回填 |
| Web and app网站与应用 | Sessions, events, landing pages, source/medium, consent, conversions, pseudonymous IDs会话、事件、落地页、来源/媒介、同意状态、转化、化名 ID | Tag coverage, duplicate events, channel rules, bots, cross-domain and identity behavior标签覆盖、重复事件、渠道规则、机器人、跨域与身份行为 |
| Email and lifecycle邮件与生命周期 | Send, delivery, open, click, unsubscribe, journey, suppression, campaign member发送、送达、打开、点击、退订、旅程、抑制、活动成员 | Machine opens, deliverability, holdouts, reused links, local time机器打开、送达能力、留出组、复用链接、本地时间 |
| CRM and salesCRM 与销售 | Lead, account, opportunity, stage timestamps, owner, source, qualified and won outcomes线索、账户、商机、阶段时间、负责人、来源、合格与赢单结果 | Deduplication, stage history, offline imports, ownership changes, source overwrites去重、阶段历史、线下导入、负责人变更、来源覆盖 |
| Orders and finance订单与财务 | Orders, net revenue, discount, return, tax, cost, margin, customer and product IDs订单、净收入、折扣、退货、税、成本、利润、客户与商品 ID | Cancellation timing, recognition rules, currency conversion, test orders, late updates取消时间、确认规则、汇率转换、测试订单、迟到更新 |
Minimum analysis package: a clear question, exports or governed read-only sources, a data dictionary, stable join keys, metric definitions, timezone and currency rules, campaign and release calendar, privacy and consent context, and at least one trusted total for reconciliation. If a field or rule is unknown, record it as an assumption rather than silently selecting a default.
最小分析包：明确问题、导出文件或受治理的只读数据源、数据字典、稳定连接键、指标定义、时区与币种规则、活动与发布日历、隐私和同意背景，以及至少一个可用于对账的可信总数。若字段或规则未知，应记录为假设，而不是静默采用默认值。
How to do digital marketing analytics in seven repeatable steps如何用七个可重复步骤开展数字营销分析
- Frame the decision before the metric. Replace “show campaign performance” with “Should we renew this paid-search campaign for the next four weeks, for which segments, and under what cost ceiling?” Name the owner, choice, population, window, outcome, guardrails, and deadline.先界定决策，再选择指标。把“展示活动表现”改成“未来四周是否续投该付费搜索活动、针对哪些分群、成本上限是多少？”写明负责人、选择、总体、窗口、结果、护栏和截止时间。
- Write metric and source contracts. Define numerator, denominator, unit, eligibility, event logic, attribution window, currency, timezone, exclusions, owner, and revision history. Record which system owns spend, conversion, customer, and revenue facts.编写指标与数据源契约。定义分子、分母、单位、资格、事件逻辑、归因窗口、币种、时区、排除项、负责人和修订历史；记录哪个系统负责花费、转化、客户与收入事实。
- Audit collection and joins. Compare platform totals to raw exports, conversions to server or CRM outcomes, and campaign keys across systems. Inspect duplicates, missing tags, backfills, consent exclusions, identity merges, offline imports, and tracking changes before calculating a rate.审计采集与连接。把平台总数与原始导出比较，把转化与服务端或 CRM 结果比较，并核对跨系统活动键；计算比率前检查重复、漏标、回填、同意排除、身份合并、线下导入与埋点变更。
- Build a baseline and segment deliberately. Plot outcomes, costs, and key rates over time with releases and campaign changes marked. Segment only by attributes that can change the decision: channel, campaign, creative, audience, geography where appropriate, device, lifecycle stage, new versus returning customer, or account type.建立基线并有目的地分群。绘制结果、成本和关键比率随时间变化，并标记版本与活动变更；只按可能改变决策的属性分群，例如渠道、活动、创意、受众、适用时的地区、设备、生命周期阶段、新老客户或账户类型。
- Choose the method by question. Use descriptive trends for monitoring, funnels for ordered transitions, cohorts for time-aligned behavior, attribution for governed credit allocation, experiments for incrementality, and forecasting for explicit scenarios. A complex model is not automatically a better fit.按问题选择方法。用描述性趋势监控，用漏斗研究有序转化，用同期群比较按时间对齐的行为，用归因执行受治理的功劳分配，用实验检验增量，用预测评估明确情景。复杂模型并不自动意味着更合适。
- Stress-test the finding. Change the time window, maturity rule, attribution setting, identity logic, join rule, denominator, and outlier treatment. Compare platform, analytics, CRM, and finance totals. Look for seasonality, selection bias, small samples, survivorship, overlapping campaigns, and Simpson’s paradox.进行稳健性检验。改变时间窗口、成熟规则、归因设置、身份逻辑、连接规则、分母与异常值处理；比较平台、分析系统、CRM 与财务总数；检查季节性、选择偏差、小样本、幸存者偏差、活动重叠和辛普森悖论。
- Turn evidence into an owned action. Separate observation, inference, alternatives, recommendation, uncertainty, and next test. Assign an owner, change date, target metric, guardrails, stop condition, and review date. Preserve the query, input version, definitions, and result so another analyst can reproduce the decision.把证据转为有负责人的行动。分别记录观察、推断、替代解释、建议、不确定性与下一步检验；指定负责人、变更日期、目标指标、护栏、停止条件与复核日期；保留查询、输入版本、定义与结果，使其他分析师能够复现决策。
Choose marketing analytics metrics by decision and funnel stage按决策与漏斗阶段选择营销分析指标
There is no universal “best marketing metric.” A metric is useful when its definition is stable, its population matches the decision, its latency is understood, and a plausible action can change it without sacrificing a more important outcome. Use a small measurement set: one outcome, two or three diagnostics, and guardrails. Platform-reported conversions can help optimize within a platform, while finance or CRM outcomes may be more appropriate for cross-channel allocation.
不存在通用的“最佳营销指标”。只有当定义稳定、总体与决策匹配、延迟已知，并且某项行动可以在不牺牲更重要结果的情况下改变它时，指标才有用。应采用精简衡量集：一个结果指标、两三个诊断指标和若干护栏。平台报告的转化适合平台内优化，而跨渠道分配通常更适合使用财务或 CRM 结果。
| Decision决策 | Outcome结果指标 | Diagnostics诊断指标 | Guardrails and cautions护栏与注意事项 |
|---|---|---|---|
| Continue or pause a campaign继续或暂停活动 | Incremental orders, qualified pipeline, contribution margin, or another governed business outcome增量订单、合格商机、贡献利润或其他受治理业务结果 | Spend, reach, CTR, conversion rate, cost per outcome, frequency花费、触达、CTR、转化率、单结果成本、频次 | Returns, brand safety, lead quality, overlap, sample size退货、品牌安全、线索质量、重叠、小样本 |
| Shift channel budget调整渠道预算 | Marginal profit or incremental customers at the proposed spend level拟议花费水平下的边际利润或增量客户 | Response curve, saturation, blended CAC, payback, attributed value响应曲线、饱和度、综合 CAC、回收期、归因价值 | Shared audiences, attribution bias, fixed costs, capacity constraints共享受众、归因偏差、固定成本、产能约束 |
| Choose a creative or message选择创意或信息 | Qualified response or experiment lift for the intended audience目标受众的合格响应或实验提升 | Viewability, completion, click, landing engagement, assisted conversion可见度、完成、点击、落地页参与、辅助转化 | Unequal delivery, novelty, fatigue, placement mix, multiple testing投放不均、新奇效应、疲劳、版位组合、多重检验 |
| Improve a journey改进旅程 | Completion, qualified lead, purchase, activation, or retained value完成、合格线索、购买、激活或留存价值 | Step conversion, time to next step, path, error, abandonment步骤转化、下一步耗时、路径、错误、放弃 | Eligibility, optional paths, identity, cross-device gaps资格、可选路径、身份、跨设备缺口 |
Metric hierarchy: activity metrics describe delivery; diagnostic metrics locate change; outcome metrics represent business value; guardrails detect damage. Vanity metrics are not inherently useless—reach and clicks can diagnose delivery—but they become misleading when presented as the final business result. Rate metrics always require an explicit denominator, while cost metrics require consistent currency, allocation, and time rules.
指标层级：活动指标描述投放，诊断指标定位变化，结果指标代表业务价值，护栏指标发现损害。所谓虚荣指标并非天然无用——触达和点击可以诊断投放——但若被当成最终业务结果就会误导。比率必须明确分母，成本指标则必须使用一致的币种、分配与时间规则。
Marketing analytics methods: reporting, attribution, experiments, and models营销分析方法：报告、归因、实验与模型
| Method方法 | Best use最佳用途 | What it does not establish不能证明什么 |
|---|---|---|
| Descriptive reporting描述性报告 | What happened, where, when, and for which governed segment发生了什么、在哪里、何时、涉及哪个受治理分群 | Why it happened or what would happen under another action为何发生，或采用另一行动会怎样 |
| Funnel, path, and cohort analysis漏斗、路径与同期群分析 | Where journeys differ, stop, or repeat for comparable groups可比群体的旅程在哪里分化、停止或重复 | Motivation or causal effect without further evidence在缺少进一步证据时的动机或因果效应 |
| Rule-based or modeled attribution规则或模型归因 | Consistent credit allocation among observed touchpoints for reporting or optimization为报告或优化在已观察触点之间一致分配功劳 | The counterfactual incremental value of a channel渠道的反事实增量价值 |
| Controlled experiment or holdout受控实验或留出组 | Incremental effect under a credible assignment and analysis design在可信分配与分析设计下估计增量效应 | Automatic generalization to other audiences, periods, or spend levels自动推广到其他受众、时期或花费水平 |
| Marketing mix or response modeling营销组合或响应模型 | Aggregate channel response, scenario planning, and saturation over longer periods较长周期内的汇总渠道响应、情景规划与饱和度 | User-level journeys or certainty beyond model assumptions and data variation用户级旅程，或超出模型假设与数据变化的信息确定性 |
| Forecasting预测 | Expected ranges under stated trends, seasonality, and scenarios在明确趋势、季节性与情景下估计预期范围 | That the chosen intervention will cause the forecasted outcome所选干预一定会导致预测结果 |
Google Analytics documents multiple attribution scopes and models in its official attribution overview. Use those reports consistently, but do not confuse assigned credit with causal proof. When channel incrementality matters, pre-register an experiment or credible holdout, define interference and contamination risks, and report uncertainty. When experimentation is impractical, triangulate: compare attribution, cohort patterns, time-series evidence, and business totals, while keeping the conclusion appropriately qualified.
Google Analytics 在官方归因概览中记录了多种归因范围与模型。应一致使用这些报告，但不要把分配的功劳误认为因果证明。需要判断渠道增量时，应预先登记实验或可信留出设计，定义干扰与污染风险并报告不确定性。若无法实验，可三角验证归因、同期群模式、时间序列证据与业务总数，同时对结论保持适当限定。
Marketing analytics example: a campaign that looks efficient营销分析示例：一个看似高效的活动
Hypothetical example: a subscription business compares paid social and paid search during a four-week acquisition campaign. The numbers below are invented only to demonstrate the workflow; they are not customer data or an industry benchmark. Platform reports show 1,200 attributed trials for social at a lower cost per trial than search, so the first dashboard suggests moving budget to social.
假设示例：某订阅业务在四周获客活动中比较付费社交与付费搜索。以下数字仅为说明工作流而虚构，不是客户数据或行业基准。平台报告显示，社交渠道归因到 1,200 次试用，单次试用成本低于搜索，因此第一版仪表板建议把预算转向社交。
The analyst first reconciles trial IDs to the product database and removes duplicates created by repeated browser and server events. Next, the team joins CRM qualification and eight-week paid retention. Social still produces more trials, but a larger share is ineligible for the target market and fewer become retained paid accounts. Search has higher acquisition cost but higher qualified conversion and contribution margin. Neither comparison proves incrementality because channel audiences and intent differ.
分析师先把试用 ID 与产品数据库对账，并删除浏览器与服务端事件重复生成的记录；随后连接 CRM 资格和八周付费留存。社交仍带来更多试用，但其中更大比例不符合目标市场资格，转为留存付费账户的比例也更低。搜索获客成本更高，但合格转化率与贡献利润更高。由于渠道受众和意图不同，这些比较都不能证明增量。
The decision is not “social failed.” The team keeps a bounded social budget for discovery, tightens eligibility and landing-message alignment, and tests incremental retained accounts with contribution margin as the outcome. Search remains funded under a marginal cost ceiling. The analysis record stores source extracts, identity and attribution rules, exclusions, query versions, and the review date. This example shows why campaign performance analysis should move from platform activity to a governed outcome without pretending observational data answered the causal question.
决策不是“社交失败”。团队保留受限的社交探索预算，收紧资格与落地页信息一致性，并以贡献利润为结果检验增量留存账户；搜索则在边际成本上限内继续投入。分析记录保存数据源导出、身份与归因规则、排除项、查询版本和复核日期。这个示例说明，活动表现分析应从平台活动推进到受治理结果，同时不能假装观察数据已经回答因果问题。
Validate marketing analytics findings before changing spend改变预算前，先验证营销分析结论
Before a digital campaign result drives spend, verify that campaign IDs, UTMs, landing pages, event names, and conversion keys retain the same meaning from collection through reporting. Reconcile platform, analytics, CRM, and finance totals separately, then explain why clicks, sessions, people, leads, and customers differ. Retest the conclusion after changing attribution windows, identity rules, cohort maturity, and duplicate-event handling.
在数字活动结果影响预算前，应确认活动 ID、UTM、落地页、事件名和转化键从采集到报告始终含义一致。广告平台、分析系统、CRM 与财务总数需要分别对账，并解释点击、会话、用户、线索和客户数量为何不同；还应改变归因窗口、身份规则、同期群成熟度和事件去重方式，重新检验结论。
- Wrong denominator: using all visitors when only eligible exposed users could convert, or mixing people, sessions, accounts, and opportunities.分母错误：只有合格曝光用户可以转化，却使用所有访客；或混用用户、会话、账户和商机。
- Duplicate conversion: browser, server, CRM, and offline imports represent the same outcome without a stable deduplication key.转化重复：浏览器、服务端、CRM 与线下导入表示同一结果，却没有稳定去重键。
- Attribution-window mismatch: platforms, analytics, and CRM use different lookback windows, clocks, scopes, and credit rules.归因窗口不一致：平台、分析系统与 CRM 使用不同回溯窗口、时钟、范围与功劳规则。
- Mixing gross and net value: revenue ignores returns, discounts, taxes, cost of goods, payment failures, or cancellations.混用毛值与净值：收入忽略退货、折扣、税、商品成本、支付失败或取消。
- Selection and survivorship: comparing mature and immature cohorts, reachable and unreachable audiences, or retained users without accounting for entry conditions.选择与幸存者偏差：比较成熟与未成熟同期群、可触达与不可触达受众，或不考虑进入条件就比较留存用户。
- Correlation presented as cause: treating an attributed path, time trend, or high-value audience as evidence that the marketing action created the outcome.把相关性当成因果：把归因路径、时间趋势或高价值受众当成营销行动创造结果的证据。
Digital measurement also needs privacy controls inside the data flow. Map which tags, cookies, SDKs, server events, and advertising uploads run under each consent state; propagate suppression and deletion downstream; restrict audience exports; and set retention by purpose. The UK ICO guidance on cookies and similar technologies is one official reference, while the applicable requirements depend on jurisdiction and implementation.
数字衡量还需要把隐私控制嵌入数据流。应梳理不同同意状态下运行的标签、Cookie、SDK、服务端事件和广告上传，把拒绝与删除要求传递到下游，限制受众导出，并按用途设置保留期限。英国 ICO 关于 Cookie 与类似技术的指南是一项官方参考，具体要求取决于司法辖区和实施方式。
Use InfiniSynapse for governed, multi-source marketing analysis使用 InfiniSynapse 开展受治理的多源营销分析
InfiniSynapse can help investigate a digital journey after the tracking plan and business outcome are defined. Combine governed channel exports, web or app events, CRM stages, and finance outcomes, then use natural-language analysis to trace where campaign keys stop matching or conversion quality changes. The analyst still approves definitions, reviews joins, and decides whether an observed channel difference warrants a test.
在跟踪方案和业务结果已经明确后，可用 InfiniSynapse 调查数字旅程。把受治理的渠道导出、网站或应用事件、CRM 阶段和财务结果组合起来，再通过自然语言分析定位活动键在哪里失配、转化质量从哪里发生变化。指标定义、连接复核以及是否需要实验，仍由分析师决定。
Include a campaign and UTM map, event dictionary, deduplication key, consent fields, conversion maturity window, channel cost extract, and trusted source totals. This package lets the analysis follow one campaign from impression or visit to qualified customer value without silently mixing platform definitions.
数据包应包含活动与 UTM 映射、事件字典、去重键、同意字段、转化成熟窗口、渠道成本导出和可信源总数。这样才能沿着单个活动从曝光或访问追踪到合格客户价值，避免无意中混用平台定义。
Analyze digital campaign data with InfiniSynapse使用 InfiniSynapse 分析数字活动数据For example, ask for landing-page conversion and eight-week retained value by device and governed acquisition channel, while reconciling event counts to analytics and customer totals to CRM. Request an exception table for missing campaign IDs, duplicated conversions, and unmatched customer keys. A precise request exposes tracking defects and audience-quality differences that a generic channel ranking would hide.
例如，可以按设备和受治理获客渠道比较落地页转化率与八周留存价值，同时把事件数同分析系统对账、客户数同 CRM 对账；并要求列出活动 ID 缺失、转化重复和客户键未匹配的异常表。精确的问题能够暴露跟踪缺陷和受众质量差异，而笼统的渠道排名会掩盖这些问题。
Build a marketing analytics operating rhythm建立营销分析运营节奏
Match the review cadence to digital execution. Check broken tags, missing campaign parameters, delayed imports, and abrupt spend shifts each day. Review pacing, creative fatigue, landing-page behavior, and funnel leakage each week. Reconcile qualified customers, refunds, and margin after cohorts mature, then revisit consent behavior and attribution settings whenever tracking or channel configuration changes.
复核节奏应贴合数字执行：每天检查标签失效、活动参数缺失、导入延迟和花费突变；每周查看投放进度、素材疲劳、落地页行为和漏斗流失；同期群成熟后再核对合格客户、退款与利润。跟踪或渠道配置变化时，还要重新检查同意行为和归因设置。
- Version definitions: effective dates and change logs prevent a renamed event or stage from masquerading as customer behavior.版本化定义：生效日期与变更日志可避免把事件或阶段重命名误认为客户行为变化。
- Separate facts and judgment: label observed data, model output, inference, recommendation, and assumption in every decision note.区分事实与判断：在每份决策记录中标注观察数据、模型输出、推断、建议与假设。
- Prefer decision thresholds: define what value, uncertainty, sample, or guardrail breach changes the action before viewing the result.优先定义决策阈值：查看结果前先规定什么数值、不确定性、样本或护栏突破会改变行动。
- Keep an analysis ledger: store question, owner, source versions, code or query, definitions, result, decision, and review outcome.保留分析台账：保存问题、负责人、数据源版本、代码或查询、定义、结果、决策与复核结果。
For broader platform selection and implementation tradeoffs, see the Data Analytics Software guide. For the wider cross-channel discipline that connects digital activity with offline spend, CRM, and business outcomes, continue to the marketing analytics guide.
关于更广泛的平台选择与实施权衡，可参考数据分析软件指南。如需把数字活动与线下花费、CRM 和业务结果放进更完整的跨渠道体系，请继续阅读营销分析指南。
Digital marketing analytics frequently asked questions数字营销分析常见问题
Digital marketing analytics is the disciplined collection, definition, comparison, and interpretation of digital marketing data so a team can evaluate performance and make testable decisions about channels, campaigns, audiences, creative, and budget.
数字营销分析是对数字营销数据进行有纪律的采集、定义、比较与解释，使团队能够评估表现，并围绕渠道、活动、受众、创意和预算作出可检验的决策。
Begin with a campaign decision and a downstream business outcome. Map campaign parameters to events and customer keys, validate browser and server collection, reconcile channel costs, wait for the required conversion maturity, and assign each finding to a budget action, tracking repair, or experiment.
先明确活动决策与下游业务结果，再把活动参数映射到事件和客户键，验证浏览器与服务端采集，核对渠道成本，并等待所需的转化成熟期；最后把发现落实为预算行动、跟踪修复或实验。
Use platform delivery and click metrics to diagnose execution, but judge performance with qualified conversion, retained customer value, contribution margin, or another agreed business outcome. Always state the entity, eligibility rule, time window, and cost boundary behind the number.
平台投放与点击指标适合诊断执行情况，但表现判断应使用合格转化、留存客户价值、贡献利润或其他约定业务结果。每个数字都应说明实体、资格规则、时间窗口和成本边界。
Web analytics focuses on website and app traffic, acquisition, behavior, and conversions. Marketing analytics is broader: it can combine web analytics with ad platforms, email, CRM, sales, cost, experiment, and revenue data to evaluate marketing decisions.
Web Analytics 侧重网站和应用流量、获客、行为与转化。营销分析范围更广，可把 Web Analytics 与广告平台、邮件、CRM、销售、成本、实验和收入数据结合，用于评估营销决策。
No. A platform or analytics model distributes credit among recorded touchpoints; it does not reveal what would have happened without the channel. Use a suitable holdout or incrementality design when the budget decision requires a causal estimate.
不能。平台或分析模型只是在已记录触点之间分配功劳，无法说明没有该渠道时会发生什么。当预算决策需要因果估计时，应采用合适的留出或增量实验设计。
Connect governed campaign, event, CRM, and value data with stable keys and source totals. InfiniSynapse can help compare funnels and surface unmatched records in natural language, while the analyst confirms tracking quality and approves any resulting campaign change.
使用稳定键和源总数连接受治理的活动、事件、CRM 与价值数据。InfiniSynapse 可通过自然语言帮助比较漏斗并发现未匹配记录，分析师则负责确认跟踪质量并批准由此产生的活动调整。
Official sources and further reading官方来源与延伸阅读
- Google Analytics: how data collection and reporting workGoogle Analytics：数据采集与报告如何工作 — official background for website and app measurement.网站与应用衡量的官方背景。
- Google Analytics attribution overviewGoogle Analytics 归因概览 — official definitions of attribution scope, models, and reporting behavior.归因范围、模型与报告行为的官方定义。
- Google Analytics BigQuery Export schemaGoogle Analytics BigQuery 导出 Schema — official field-level reference for exported event data.导出事件数据的官方字段级参考。
- UK ICO guidance on cookies and similar technologies英国 ICO 关于 Cookie 与类似技术的指南 — official privacy guidance relevant to some measurement implementations.与部分衡量实现有关的官方隐私指南。