Campaign evidence
G07 · TOPIC · campaign-evidenceGCRI dossier · service level R0
GCRINexus CampaignsEstablished
Topic dossier · Nexus Campaigns
Campaign evidence
A campaign produces activity that is easy to count, such as impressions, shares and attendance, and outcomes that are hard to establish, such as understanding, trust or safer practice. Campaign evidence is the discipline of measuring each level honestly and claiming no more than the design supports. This page sets out how reach, learning and outcomes are measured, why attribution is difficult, how metrics distort practice, and which evidence funders, agencies and evaluators can rely on.
DossierG07 · topic
- 01In briefDefinition and scope
- 02The question4 mechanisms
- 03Key terms3
- 04FieldH3 field plate
- 05Indicators8
- 06PipelineRun graph
- 07Sources7
- 08Decision holders4
- 09Methods2
GCRI · Established · service level R0
01In brief
The short answer.
Campaign evaluation is the measurement of what a campaign reached, what people understood and what they did differently, claiming no more than the design supports. Reach and impressions are easy to count; understanding and safer practice are hard to establish, and attributing them to a campaign needs a comparison group.
02Geography
Digital demand and connectivity in Indonesia.
Urban demand centres and the hubs that anchor digital and data infrastructure, shown for Indonesia as an example geography for this subject.
GCRI · Nexus Campaigns · fieldDESIGN FIGURE
03Console
Campaign wave tracker, in outline.
The working screen behind this subject: its structure, terms and states. Names are placeholders and values illustrative.
Campaign wave tracker
Campaign evidence · Wave 2 · Campaign area North
- People reached14076Active
- Consent complete, % of participants97Normal
- Knowledge score change0.8pointsNormal
- Open feedback items12Watch
Wave in fieldGCRI · Nexus Campaigns
Preparedness actions by area
Messages under test
| Prepare a water reserve | A | 61 | Keep |
| Know your warning levels | B | 57 | Revise |
| Check on neighbours | A | 74 | Keep |
Safeguarding and consent
- S-1Minor enrolled without guardian recordPause and verify1 d
- S-2Withdrawal requestData deleted, logged5 d
04The question
Campaign evidence: how the risk develops.
Learning metrics and demonstrable outcomes. The dossier below sets out how the risk develops, what is measured, where the data come from, how it is modelled and who decides.
Mechanism
Reach is not understanding
Exposure, attention, comprehension, intention and action are separate levels, and evidence at one does not establish the next. A logic model states which level each measure speaks to, so that impressions are not reported as learning or intentions as behaviour.
Mechanism
Attribution needs a comparison
Change during a campaign coincides with news, prices, seasons and other programmes. Without a comparison group, a credible counterfactual or a pre-registered design, a before-and-after difference cannot be credited to the campaign, however large it appears.
Mechanism
Metrics bend practice
When a measure becomes a target it stops measuring well. Campaigns optimised for engagement reward provocative content, platforms change how they count views, and easy metrics crowd out harder evidence. The choice of metric shapes what the campaign becomes.
Mechanism
Evidence travels poorly
Effects found in one country, medium or period often do not replicate elsewhere. Positive evaluations are published more often than null ones, so reviews of the literature can overstate what campaigns achieve. Context has to travel with every result.
05Key terms
Terms used on this page.
Each term as this page uses it. Where a standard-setting body defines the term, the wording follows that body.
| Term | Definition |
|---|---|
| Deduplicated reach | The number of distinct people a campaign reached, counting each person once across channels. |
| Pre-registration | Publishing an evaluation's design and analysis plan before results are known. |
| Null result | A finding of no detectable effect, which still informs future campaigns when published. |
06Indicators
What is measured, and how well.
Each indicator carries its unit, its cadence, the kind of source behind it and the evidence label a value would carry. No values are shown: a programme supplies them under its own authority.
| Indicator | Unit | Cadence | Source type | Label |
|---|---|---|---|---|
| Deduplicated reach | persons or accounts reached | per campaign | Platform and site analytics | ESTIMATED |
| Message recall | percent of a sampled audience recalling the message | per survey wave | Post-campaign survey | REPORTED |
| Comprehension gain | difference in test scores before and after | per cohort | Pre and post assessment | OBSERVED |
| Intention to act | mean score on a validated scale | per survey wave | Survey instrument | REPORTED |
| Observed protective action | proportion of a sample observed taking the action | per study | Observation or administrative records | OBSERVED |
| Effect estimate | standardised difference with confidence interval | per evaluation | Controlled or quasi-experimental evaluation | ESTIMATED |
| Pre-registration status | registered or not, with registry and date | per evaluation | Trial and study registries | DECLARED |
| Cost per person reached | currency per person | per campaign | Budget and reach records | ESTIMATED |
07Canvas
From sources to decisions.
The topic as a computational graph: named sources feed the methods, an evidence review gates the outputs, and the decisions stay with the authorities that hold them.
Pipeline canvas · Campaign evidence as a run graphDesign, not a record
Fig. 02Nodes carry their type, two metric rows and a state; connectors show what feeds what; dashed frames name the owners; the timeline is the cadence. Design of the pipeline, not a record of a run.
08Sources
Where the evidence comes from.
Named public datasets and institutional records, each with its licence class. Sources by agreement are used only under their terms and never republished.
| Source | What it gives | Licence class |
|---|---|---|
| 3ie Development Evidence Portal | Impact evaluations and systematic reviews of development interventions, including communication | Free to read |
| Campbell Collaboration systematic reviews | Reviews of social interventions with stated methods and certainty of evidence | Open access, CC BY |
| AEA RCT Registry | Pre-registered randomised trials in the social sciences | Free to read; metadata openly available |
| OSF Registries | Pre-registered study designs and analysis plans | Free to read; metadata CC0 |
| Meta Content Library and API | Public content and engagement data from Facebook and Instagram, the successor to CrowdTangle | Access by application for qualified researchers |
| OECD DAC Evaluation Resource Centre | Evaluation reports from development agencies | Free to read |
| Platform transparency reports | Reach, moderation and enforcement data published by platforms, including under the EU Digital Services Act | Free to read; reuse per platform terms |
09Decisions
Who decides, and what the evidence serves.
Nexus Campaigns prepares evidence. The decisions below are taken by the authorities named, under their own rules, and never by the platform.
| Who decides | The question | Evidence this page serves |
|---|---|---|
| Funders and foundations | Whether to continue, scale or stop funding a campaign | Outcome records by level; effect estimates with intervals; quality of the evaluation design |
| Public health and communication agencies | Which channels and formats to use in the next campaign | Comprehension and recall evidence; cost per person reached; context of each result |
| National audit offices and evaluation units | Whether public spending on campaigns met its stated aims | Logic models; pre-registration records; observed action measures |
| Platform research and transparency teams | Which data to make available to independent researchers | Documented researcher access needs; gaps in published transparency data |
Outputs this topic produces
- Campaign outcome records separating reach, understanding, intention and observed action
- Logic models stating what each measure can and cannot show
- Pre-registered evaluation plans, with null results published alongside positive ones
- Learning summaries recording the context each result came from
10Core questions
What experts ask.
- 01
What should change?
Knowledge, intended action, trusted sources or behaviour, chosen to match the campaign's aim.
- 02
Compared with what?
The baseline and, where feasible, a comparison group, so change can be attributed with care.
- 03
How long does it last?
Whether effects persist after the campaign or fade within weeks.
- 04
Who was missed?
Which groups the campaign did not reach and why.
11Mechanisms
What drives the outcome.
Reach is easy to count and weakly linked to effect. Impressions and views say little about understanding. Evaluation research distinguishes outputs (what was delivered), outcomes (what changed for people) and impact (longer-term change attributable to the campaign).
Attribution is the hard part. Other messages, events and policies move the same measures. Without a baseline and some form of comparison, a campaign may claim credit for change it did not cause or miss effects it did have.
Publication bias distorts learning. If only successes are reported, the field learns little. Recording null and negative results is as valuable as recording successes, and it protects the next campaign team from repeating an approach that has already failed elsewhere.
Measures also shape behaviour. A team rewarded for reach will optimize for reach. Agreeing measures of understanding or action before launch keeps attention on the public purpose.
12Methods and limits
How evidence is gathered, and where it stops.
| Method | What it shows | Its limit |
|---|---|---|
| Baseline and follow-up surveys | Change in knowledge or intentions | Self-report and attrition |
| Comparison designs | Effect relative to a group not reached | Often hard to arrange ethically or practically |
| Administrative data | Behaviour such as service uptake | Many confounding influences |
| Qualitative follow-up | Why the campaign worked or did not | Not generalizable on its own |
13Methods
How it is studied, and what the methods cannot say.
14Questions
Questions and answers.
Why not just report reach?
Reach shows how many people may have seen something, not whether they understood, believed or acted on it. Reporting it as impact overstates what is known.
What is a counterfactual?
An estimate of what would have happened without the campaign, usually from a comparison group or a design that rules out other explanations. Without one, change cannot be attributed to the campaign.
Why publish null results?
Because an evidence base built only from successes overstates what campaigns achieve and misleads the next funder. A null result is information about what did not work in that context.
What replaced CrowdTangle?
Meta closed CrowdTangle and moved researcher access to the Meta Content Library. Access is by application, and coverage and definitions differ, so comparisons across the change need care.
Are disappointing results published?
Yes. Outcome records report results against the measures agreed before launch, including those that fall short.
Does every campaign need a control group?
No. Comparison designs are used where they are ethical and feasible. Where they are not, the record states the limits of the claim.
Can I see the data behind a result?
Aggregate results are published with their method. Individual responses are never published, in line with each campaign's consent protocol.
Related
Around Nexus Campaigns.
Nexus Campaigns
Bring your expertise to this field.
The current service level is R0: information and enquiries. An enquiry receives a reference and names the next step; applications, payments, compute allocations and registrations open only when their team, terms and review capacity exist.
Forgot Password
Reset Password