设计多合作伙伴项目的 KPI

当 40 个合作伙伴报告同一项 KPI 时,任何定义不清之处都会被汇总。本文介绍精简的共同核心、每项 KPI 的参考表、如实反映情况的汇总规则,以及附有填写示例的免费模板。

CpThe quarterly figures from forty partners arrive the week before the board meeting. One line adds up the young people trained, and it looks like a good quarter. Then the questions start. One partner counted every session attended, another only those who finished the course, a third included its own volunteers. The total is correct arithmetic, and it means almost nothing./C CpNobody did anything wrong; the program never wrote down which question it was asking. That is the first problem of KPIs for partner programs: every loose definition gets added up. This guide covers a small shared core of KPIs, a reference sheet for each one (a free template with a filled-in example), aggregation rules that tell the truth, the checks to run first, KPIs that follow a partner's stage, and how many is too many./C H2Why KPIs for partner programs are different/H2 CpThe OECD's evaluation glossary defines key performance indicators as the subset of indicators considered most important to the goals, monitored in an ongoing way, and notes that any indicator is "always approximate only" and needs interpretation [1]. With one partner, you can simply ask what they counted. With forty, three things change:/C Ul LiStrongDefinitions become contracts./Strong Each partner settles every ambiguity its own way, and the total mixes them. INTRAC lists what adding up across a portfolio requires: the same definitions, every relevant project collecting the data, similar timescales, the same tools and consistent quality [2]./Li LiStrongThe burden multiplies./Strong One more indicator is one more number for you, and forty for your partners, every round./Li LiStrongPartners are not alike./Strong Size, context and maturity differ, so the same number means different things in different places./Li /Ul H2A small shared core, plus each partner's own/H2 CpWhat works in most multi-partner programs is two layers. A Strongshared core/Strong of five to eight KPIs that every partner reports with the same definition: the only numbers you add up. Then two or three Strongpartner-specific indicators/Strong each, chosen by the partner for its context and read partner by partner./C CpThe IFRC's monitoring and evaluation guide describes the trade-off: standard indicators have accepted definitions and can be compared across projects, but can be unrepresentative of the local context, so a mix is often best [3]. INTRAC names the risk of standardizing everything: at worst, projects collect information that is of no use to them [2]. The partner-specific layer is also where the best indicators for implementing partners come from; some later earn a place in the core./C H2Outputs, outcomes and what a partner controls/H2 CpIn the OECD glossary, outputs are the products and services an intervention delivers, within the control of the implementing team; outcomes are their short and medium-term effects [1]. In a multi-partner program, that has two consequences:/C Ul LiStrongOutputs add up; outcomes resist it./Strong INTRAC finds aggregation quite possible at output level and extremely difficult at outcome level, even where industry-standard definitions exist [2]. Outcome KPIs need your tightest definitions./Li LiStrongOutcomes depend on context./Strong A youth employment partner in a district with few jobs will not match one in the capital, however good its course. Compare each partner with its own baseline first./Li /Ul CpA sound core holds both: outputs that show the work is happening, and one or two outcomes that show whether it matters./C H2Write every KPI down: the KPI reference sheet/H2 CpEvery core KPI needs a written definition before the first partner reports. The IFRC's M&E plan records, for each indicator, a detailed definition, its sources, the methods and timing of collection, the people responsible and the intended use. The definition should state precisely how the indicator is calculated, such as the numerator and denominator of a percentage, and how it is split by sex, age or other variables. People are named by job title too, so the indicator survives staff changes [3]. The template below adds three fields that matter when many partners report: the aggregation rule, the data quality checks and a version log./C TableCaptionThe fields of a KPI reference sheet, with the filled-in example's answers/Caption Thead Tr Th scope="col"Field/Th Th scope="col"What it settles/Th Th scope="col"In the example/Th /Tr /Thead Tbody Tr TdDefinition/Td TdEvery term a partner could read two ways/Td TdA graduate attended at least 80% of sessions; work means at least 10 hours a week/Td /Tr Tr TdNumerator and denominator/Td TdWhat is counted, out of what/Td TdGraduates not reached at six months stay in the denominator/Td /Tr Tr TdSource and frequency/Td TdWhere the number comes from, and when/Td TdCourse register and a follow-up call; quarterly/Td /Tr Tr TdAggregation rule/Td TdHow partners' values are added up/Td TdPooled rate, plus the partners at 50% or more/Td /Tr Tr TdQuality checks/Td TdWhat is checked before adding/Td TdA numerator above its denominator goes back to the partner/Td /Tr Tr TdVersion log/Td TdWhat changed, when and why/Td TdEducation and training added in version 1.1/Td /Tr /Tbody /Table H2The KPI reference sheet template (free download)/H2 CpThe template has two A4 pages: how to use it, the indicator, its definition and data collection (page 1); the aggregation rule, quality checks, limitations, version log and sign-off (page 2)./C FigureImg src="/blog-media/88/kpi-reference-sheet-page-1.png" alt="Page 1 of the blank KPI reference sheet: how to use it, the indicator, its definition with numerator and denominator, and data collection" width="620" height="877" FigcaptionPage 1 of the template: the indicator, its definition and how the data is collected./Figcaption /Figure CpA href="/blog-media/89/kpi-reference-sheet.pdf"Download the KPI reference sheet template (PDF, 2 pages)/A. Print one per KPI, or copy the fields into your own M&E plan./C Div id="aw-blog-cta" class="aw-blog-cta" CpStrongDefining KPIs for more than a handful of partners?/Strong In Advancer.World you define each KPI once per program and record it per partner, next to where that partner stands on the program's stages. Daily snapshots keep the history, so the trend is there when you report. Free to start./C CpA class="lp-btn lp-btn-primary lp-btn-lg" href="/Account/Register"Sign up now/A/C /Div H2Example: one KPI, fully defined/H2 CpHere is the sheet filled in for a core outcome KPI of the Youth Pathways Program, in which Horizon Child Fund works with 40 local partners that run employment courses for young people. Both are fictional./C FigureImg src="/blog-media/90/kpi-reference-sheet-example-page-2.png" alt="Filled-in example, page 2: the aggregation rule with the pooled rate and a 50% threshold ticked, data quality checks, limitations and the version log of a fictional youth employment KPI" width="620" height="877" FigcaptionPage 2 of the example (fictional organizations): how the KPI is added up, checked and versioned./Figcaption /Figure CpA href="/blog-media/91/kpi-reference-sheet-example.pdf"Download the filled-in example (PDF, 2 pages)/A. Three choices are worth copying:/C Ul LiStrongGraduates nobody could reach stay in the denominator./Strong Otherwise the partner that only reaches its success stories reports the best rate./Li LiStrongThe sheet says what it must not be used for./Strong Ranking partners or deciding funding, because local labor markets differ more than partners do./Li LiStrongThe version log explains every break./Strong When version 1.1 added education and training, the log said so, and nobody mistook a new definition for progress./Li /Ul H2Adding it up: same data, three stories/H2 CpOnce the definitions hold, you decide how to add partners' values together. It sounds like arithmetic, and it is where most portfolio reports go wrong, because the same data supports opposite headlines./C FigureImg src="/blog-media/92/kpi-aggregation-three-stories.png" alt="Sample data for ten fictional partners: the total in work or training rose from 528 to 763, the average of partner rates rose from 53% to 63%, partners at the 50% threshold fell from 8 to 4, and the pooled rate went from 53% to 50%" width="1600" height="1240" FigcaptionIllustration with sample data: ten fictional partners, one KPI, three ways of adding it up, three different headlines./Figcaption /Figure CpIn this sample, young people in work or training rose 45 percent, mostly because one partner more than doubled its intake. The average of the ten partners' rates rose ten points, because four small partners did very well. The partners at the 50 percent threshold halved. All three are true. The honest summary is the pooled rate, down from 53 to 50 percent, with the four partners that slipped below the line named./C CpAveraging partners' percentages misleads twice over. It gives a partner with 20 graduates the weight of one with 900. And, as UNDP's handbook puts it, "averages tend to hide disparities" [4]: two partners at 90 percent and two at 10 average 50, like four at 50. Show the count next to every rate, too: with 20 graduates, one young person moves the rate five points. Levels and stages should not be averaged at all; count partners per level, as in our A href="/blog/partner-capacity-assessment-tool"partner capacity assessment guide/A./C CpOne more trap, from my years in operations research: a pooled rate can fall while no partner gets worse, if the growth happens at partners with lower rates. Before you ask who declined, ask who grew./C TableCaptionWhich aggregation rule for which kind of KPI/Caption Thead Tr Th scope="col"Kind of KPI/Th Th scope="col"Example/Th Th scope="col"Add up with/Th Th scope="col"Avoid/Th /Tr /Thead Tbody Tr TdCount of unique people or things/Td TdYoung people enrolled/Td TdSum, after ruling out double counting/Td TdAdding counts defined differently/Td /Tr Tr TdPercentage or rate/Td TdGraduates in work at six months/Td TdPooled rate: all numerators over all denominators/Td TdAveraging partners' percentages/Td /Tr Tr TdLevel, rating or stage/Td TdA partner's stage in MEAL and data/Td TdThe number of partners at each level/Td TdAveraging levels into a score/Td /Tr Tr TdYes or no/Td TdSafeguarding focal point in place/Td TdShare of partners with a yes, and which/Td TdCounting a missing answer as a no/Td /Tr Tr TdTarget or threshold/Td Td50 percent or more in work/Td TdShare of partners at or above it, with names/Td TdA share without the names behind it/Td /Tr /Tbody /Table CpWhatever the rule, report the program figure, how partners spread around it and the list behind it: every number on a board slide should lead back to the partners who produced it./C H2Check the data before you add it up/H2 CpAdding up turns small errors into confident totals. The IFRC asks for procedures to check and clean data and to treat missing data, and names the usual causes of unreliable data: typing errors, duplicate entries, inconsistent data, and data deleted or lost [3]. UNDP's handbook names what is at stake: reliability (consistent measurement) and validity (measuring what was intended) [4]. Five checks cover most of it:/C Ol LiStrongCompleteness./Strong Every partner due has reported. Missing is marked as missing, never as zero./Li LiStrongConsistency./Strong No numerator is larger than its denominator, and the splits by sex or age add up to the total./Li LiStrongPlausibility./Strong A jump of more than 20 points since the last round gets a question, not a silent correction./Li LiStrongDuplicates./Strong The same person counted by two partners, or twice by one./Li LiStrongVerification./Strong A few records per partner per year, checked against the register or the call log./Li /Ol CpRun them as questions to the partner, not as an audit. A flagged value is often a definition problem, and the fix belongs on the reference sheet, for everyone./C H2Match KPIs to the stage a partner is in/H2 CpIf your program runs on a maturity path, the right KPIs change as a partner moves along it: a KPI that matters at stage 2 may tell you nothing at stage 4. The example program has four stages:/C TableCaptionKPIs that carry information at each stage of the example program/Caption Thead Tr Th scope="col"Stage/Th Th scope="col"What the partner is building/Th Th scope="col"KPIs that tell you something/Th /Tr /Thead Tbody Tr Td1 Starting/Td TdA course that runs well/Td TdEnrollment by sex and age; course completion/Td /Tr Tr Td2 Delivering/Td TdA follow-up call for every graduate/Td TdShare of graduates reached; the work-or-training KPI begins, as a baseline/Td /Tr Tr Td3 Connecting/Td TdLinks with employers/Td TdThe work-or-training KPI against its target; placements offered/Td /Tr Tr Td4 Leading/Td TdLasting, decent work/Td TdGraduates still in work at 12 months; the partner's own indicators/Td /Tr /Tbody /Table CpTwo rules follow. Do not hold a partner to an outcome it is not yet set up to influence: a placement rate from a stage 1 partner measures its district, not its work. And retire a KPI once nearly every partner meets it: course completion matters at stage 1, but at stage 4, with every partner above 90 percent, it separates nobody and informs no decision. Both rules depend on knowing each partner's current stage, and there are A href="/blog/partner-progress-assessment"three ways to keep it current/A: partners assess themselves, the work they complete moves them, or your team decides. For describing the stages themselves, see our guide to A href="/blog/how-to-build-a-maturity-model"building a maturity model/A./C H2How many KPIs is too many?/H2 CpDo the multiplication first. Six core KPIs, each split four ways and reported quarterly by 40 partners, make 3,840 numbers a year. Twelve KPIs split eight ways make 15,360. Each one is typed by someone at a partner and checked by someone on your team. The IFRC's advice is to monitor only what is necessary and sufficient, because "extra information is more often a burden than a luxury" [3]. UNDP's handbook puts credibility, not the volume of data, at the heart of good indicators [4]./C CpMy test for every KPI has three questions: who looks at it, which decision does it change, and what will we do if it drops? If nobody can answer all three, it is not core. That usually leaves five to eight shared KPIs, two or three per partner, and a few that switch on and off with the stage./C

在 Advancer.World 中管理 KPI

Advancer.World 源于 UNICEF、Airbus 等组织项目的数字化转型。项目由阶段和主题组成,阶段之间的步骤承载着实际工作:举措、调查问卷,以及用于佐证数据的文件上传。

Advancer.World 项目设计:青年发展路径项目,包含四个阶段——起步、实施、联结和引领,涵盖课程交付、雇主关系、后续跟进、MEAL 和安全保障
Advancer.World 中的示例项目(示例数据):从起步到引领的四个阶段,涵盖项目交付、就业联系和学习。

每个项目只需定义一次 KPI,并按合作伙伴分别记录,因此所有合作伙伴都依据相同的 KPI 进行报告。合作伙伴可以使用 22 种语言中的任意一种,查看自己的进展情况以及下一阶段的要求,但看不到其他合作伙伴的数据。每日快照会保留历史记录,供 Excel 报告和 AI 撰写的 Word 图表报告使用。在 Gold 和 Enterprise 方案中,REST API 和 MCP 服务器可以读取合作伙伴的评分卡、各子主题的进展情况及其 KPI 数值,以及项目趋势。免费方案提供的五个 KPI 足以让您与少数合作伙伴试行一套共同的核心指标。

常见问题

KPI 和指标有什么区别?

所有 KPI 都是指标,但并非所有指标都是 KPI。KPI 是被认为对实现目标最为重要、并持续监测的一类指标 [1]。在多合作伙伴项目中,共同核心指标就是您的 KPI;其他指标则可根据合作伙伴和团队的需要使用。

实施合作伙伴是否应该自行选择指标?

部分指标可以由他们自行选择。共同 KPI 需要与合作伙伴商定统一定义;合作伙伴特定的指标则由合作伙伴自行决定。国际红十字会与红新月会联合会建议,由实际使用监测与评估计划的人共同制定该计划,这也有助于提高数据质量 [3]。

如何汇总各合作伙伴的百分比?

先将所有分子相加,再将所有分母相加,然后用分子总和除以分母总和。切勿对各合作伙伴的百分比取平均值;报告结果时,还应按名称列出达到或超过您设定门槛的合作伙伴。

从您最不信任的 KPI 开始

开始行动不需要新建一套框架。选出您最不信任的核心 KPI,与两三家合作伙伴一起填写该 KPI 的参考表,并请他们各自统计相同的五个案例。差异会告诉您需要记录什么。然后再处理下一个 KPI。

统一定义,每位合作伙伴,每一轮。 设置项目的核心 KPI,邀请合作伙伴,并基于同一份数据在 Excel 中生成报告,或使用 AI 撰写 Word 报告,同时保留历史记录。合作伙伴无需付费。免费开始使用。

立即注册

参考资料

  1. OECD,《可持续发展评估与基于结果的管理关键术语词汇表(第二版)》,OECD Publishing,巴黎,2023 年,https://alnap.org/documents/20488/OECD_glossary_key_terms_eval_rbm_sd-fr-es.pdf(访问于 2026 年 10 月)。
  2. Nigel Simister,INTRAC,《总结组合层面的变化:组织层面的成果框架》,M&E 论文 10,2016 年 1 月,https://www.intrac.org/resources/paper-10-summarising-portfolio-change-results-frameworks-organisational-level/(访问于 2026 年 10 月)。
  3. 国际红十字会与红新月会联合会,《项目/计划监测与评估(M&E)指南》,日内瓦,2011 年,https://www.ifrc.org/sites/default/files/2021-09/IFRC-ME-Guide-8-2011.pdf(访问于 2026 年 10 月)。
  4. 联合国开发计划署,《发展成果规划、监测与评估手册》,纽约,2009 年,http://web.undp.org/evaluation/handbook/documents/english/pme-handbook.pdf(访问于 2026 年 10 月)。

© 2026 Mourinho Solutions。保留所有权利。

← 返回博客