A useful marketing portfolio is a focused set of case studies that shows the problem, your specific contribution, what shipped, what changed, and the evidence behind each claim. Instead of writing “improved landing-page conversions,” identify the page and measurement period, state the conversion rate and denominator, explain your role, and link to an approved source. Start with a few relevant projects you can explain accurately, not a large gallery.
A portfolio complements a résumé; it does not replace one. No project count, format, or portfolio is a universal hiring requirement. Canva reported in January 2025 that a Sago survey of more than 10,000 respondents across 10 countries found that 72% of surveyed hiring managers preferred candidates with a portfolio. That is a reported survey finding, not an industry-wide benchmark or proof that a portfolio secures interviews. Read Canva’s survey report.
The practical goal is evidence readiness: you should be able to explain what happened, who did what, how the result was measured, what can be shared, and where the supporting record lives.
What a marketing portfolio needs to prove
A marketing portfolio is selected proof of your decisions, contribution, deliverables, and outcomes. It should help a reader distinguish what you personally did from what a team did, and what the evidence supports from what you cannot establish.
Three well-supported projects are a practical starting point, not a rule. Choose examples that match the work you want to do and demonstrate different relevant skills. A short bio, target specialization, selected work, and clear contact route make the portfolio easier to navigate, but the case studies should remain the main proof.
Permission is separate from anonymization. A project may be technically possible to disguise while still being restricted by a client agreement, employer policy, or NDA. Removing a logo does not create permission to publish.
Build each case study around an evidence contract
Before writing polished copy, record the claim, its source, your role, and whether the material is approved for the intended audience. Use five parts:
- Context: What business or audience problem was being addressed, and during what period?
- Contribution: What did you decide, create, test, coordinate, or analyze? Name shared responsibilities plainly.
- Deliverables: What actually shipped? Identify the relevant asset versions or approved links.
- Outcome: What changed? Separate observed results from estimates, models, feedback, and qualitative outcomes.
- Evidence and constraints: What source supports each claim, and what cannot be shared or confidently attributed?
For example: “I researched the audience, rewrote the landing page, and coordinated a test. The growth team owned analytics setup. In March, the page converted 48 of 1,000 sessions; the comparison period and source export are recorded in the evidence register.” This wording identifies contribution without claiming sole ownership or implying causation.
A portfolio metric is only as useful as its source, scope, comparison basis, and permission to disclose it.
Examples can clarify structure without proving performance. Carlinda Lee’s portfolio combines an introduction, tools, quantified highlights, and featured work. Those figures are portfolio-owner claims without independently supplied analytics, so use the page as a structural example only. Cal State Fullerton provides a student-oriented framework requiring a résumé, achievements, credentials, and LinkedIn information. It is an academic program requirement, not a general employer standard. View the university framework.
Capture evidence before drafting
Gather evidence while project files and stakeholders are still accessible. A spreadsheet or controlled document register can be enough to start. Keep the register as an index, source files in their existing system of record, and the public case study as a separate approved version.
ClickUp Docs is one possible drafting workspace. ClickUp describes collaboration, comments, revision history, permissions, embedded content, and export options. It does not provide a verified portfolio-specific template, automatic analytics validation, or a published portfolio evidence schema. Check current product details before relying on plan-specific features. If you need help structuring a workspace, consider ClickUp consulting.
Useful register fields include project ID, brief, target audience, role, contributors, deliverable references, metric observations, reporting period, attribution window, source reference, permission status, redaction status, and approval status. Keep records at distinct levels:
- Case-study record: One person, project, and portfolio version.
- Asset record: One source artifact or version, such as a landing page or social post.
- Metric observation: One object, metric, measurement period, attribution window, source, and capture run.
- Approval record: One decision by a reviewer for a specific case-study version.
This separation prevents an aggregate campaign outcome from being attached to an individual post or page. If intake becomes automated, use stable source IDs where available and enforce uniqueness with a database constraint or transactional upsert. A search-then-create check can race when two submissions arrive together. If the source has no stable ID, retain its URL or file reference, capture time, content hash where practical, and a manually assigned provenance ID. These are proposed implementation controls, not documented ClickUp features. For workspace implementation support, see ClickUp setup and automation.
A structured record can make drafting and review more reliable. The following is an illustrative editorial model, not a ClickUp schema. The capture run distinguishes repeated imports of the same underlying observation.
{
"case_study_id": "cs_proj_2026_014_v1",
"person_id": "person_087",
"project_id": "proj_2026_014",
"source_asset_id": "analytics_export_014_v1",
"metric_observation_id": "obs_lp014_cv_202603_run01",
"object_type": "landing_page",
"object_id": "lp_014",
"metric_name": "conversion_rate",
"value": 4.8,
"unit": "percent",
"numerator": 48,
"denominator": 1000,
"measurement_start": "2026-03-01",
"measurement_end": "2026-03-31",
"attribution_window": "session",
"capture_run_id": "run_2026_04_02_01",
"source_ref": "analytics_export_014",
"permission_status": "pending"
}Check what each metric actually measures
State the object, metric definition, unit, period, source, and comparison behind every numeric claim. A conversion rate needs a numerator and denominator. A percentage change needs both a baseline and a current value, with comparable periods and definitions. Label observed, modeled, estimated, and illustrative results distinctly.
For a hypothetical landing page, 48 conversions from 1,000 sessions is 4.8%. If the comparable baseline was 40 from 1,000 sessions, the change is 0.8 percentage points, or a 20% relative increase. Those are different descriptions. Neither proves that the rewrite caused the change unless the test design supports that conclusion.
A campaign’s leads, a landing page’s conversion rate, and a post’s saves are different observations. Give each its own object, period, source, and attribution window. Never attach a channel total to an individual asset.
Use a decision gate before publishing: block a conversion-rate claim if its denominator is unknown, and block a percentage-change claim if the baseline or comparison period is missing. When exact figures are confidential, use only a client-approved range, indexed change, or directional result. If even that is restricted, describe the method, deliverables, and decisions without implying an unverified outcome.
Choose examples that show different kinds of marketing work
Select projects for relevance to the target role and the evidence you can explain. The examples below are evidence patterns, not proven hiring formulas. Each row includes a practical intake and fallback path.
| Scenario | Evidence intake | Validation and action | Fallback |
|---|---|---|---|
| SEO or content | Published asset, audience, goal, role, dated analytics export, and target query or content brief. | Separate the asset’s result from site-wide organic traffic. State the period and comparison, then link the claim to the relevant observation. | If analytics are unavailable, show the brief, research, edits, and approved deliverable without a performance claim. |
| Social media | Authorized post capture, platform, post ID or URL, publish date, metric period, and metric definition. | Check whether the figure is post-level or account-level. Archive the approved copy before publication and label the object clearly. | If a link disappears, use an authorized saved copy and record its source and capture date. |
| Paid media or email | Campaign report, creative or message, audience and budget context where permitted, attribution window, and ownership notes. | Separate campaign outcomes from individual creative results. State who owned setup, budget, copy, analysis, and approval. | When data cannot be shared, present the permitted creative and contribution, or arrange an approved private walkthrough. |
Student assignments, volunteer projects, personal channels, and speculative work can demonstrate relevant thinking when client experience is limited. Label the work accurately. For speculative work, identify the brief as self-initiated and do not imply that a real brand commissioned or shipped it.
Use AI for a bounded drafting job
AI can summarize approved notes, organize a first draft, or flag missing dates and role details. It should not supply missing metrics, infer sole ownership, turn correlation into causation, or decide whether client material may be published.
Use deterministic rules for required fields, allowed status values, known denominators, duplicate keys, source references, and permission gates. Use AI for summarization and question generation. A person must approve numbers, attribution, client naming, redactions, and any disclosure required by the employer or client.
Document the AI-assisted task, input type, human decisions, fact-check method, editing steps, and final owner. Canva reported that 73% of surveyed hiring managers believed candidates should disclose generative-AI use in application materials. That is survey context, not a universal rule; follow relevant employer or client expectations.
A bounded request might ask AI to summarize only supplied notes and return missing evidence as questions. If the notes contain a current conversion rate but no baseline, the draft should say “comparison not supplied” rather than calculate an improvement. Every factual statement should then be checked against its source before approval.
Choose a format and publish only approved work
- Website: Useful for public, linkable work that you can update. Check mobile display, accessibility, broken links, and whether embedded content still loads.
- PDF: Useful for a tailored application or pitch. Add a version date and review it when evidence or permissions change, since copies can become stale.
- Private walkthrough: Suitable for approved work that cannot be made public. It still requires permission and controlled sharing; a private presentation is not permission to disclose restricted material.
Use clear publication states as a proposed process: private_raw, redacted_internal, approved_public, and expired_or_withdrawn. Keep original evidence separate from the redacted draft and public version. Before release, check client permission, NDA obligations, contributor attribution, personally identifiable information, screenshot contents, metric support, and link access. A client’s public-facing asset does not automatically authorize disclosure of internal performance data.
Final quality gate and common questions
- The target role or specialization is clear.
- Each case study states context, contribution, deliverables, outcome, and evidence.
- Metrics identify the object, unit, period, and comparison where relevant.
- Team contributions and AI assistance are described accurately.
- Permission, redaction, PII, and screenshot contents are checked.
- Links work, and a named owner has a future review date.
How many projects should a marketing portfolio include?
There is no universal count. Start with a small set of relevant projects you can substantiate, then add examples when they demonstrate a distinct skill, audience, or type of work.
Can I build a portfolio without client experience?
Yes. Use clearly labeled class, volunteer, personal, or speculative work. State the brief, intended audience, your contribution, and what was actually tested or delivered.
Should my portfolio be a website or PDF?
Choose based on use. A website is updateable and linkable; a PDF can be tailored and dated for a particular application. Some marketers maintain both, but neither is required for every role.
Can I include confidential client work?
Only when permitted. If public approval is unavailable, keep the work private or describe only the skills and deliverables you are allowed to disclose. Anonymization does not override an NDA.
What if I no longer have the original analytics?
Do not reconstruct a number from memory. Look for an authorized export or ask the analytics owner to confirm it. If the evidence cannot be recovered, present the deliverable and process without the unsupported metric.
Review the published version when a link breaks, a result becomes outdated, or an approval changes. A dated source reference and named portfolio owner make that maintenance practical.