what goes in a marketing portfolio for analytics roles?

chatbox

Every other marketing job posting now wants a portfolio, even for roles that aren't remotely visual. That was fun when I worked on investment marketing, where every number falls under a confidentiality agreement. Can't share exact campaign metrics, can't show revenue lift, can't even hint at which accounts I touched. So what do you actually attach when the recruiter expects samples?

Here's what I've found works, from sitting on both sides of the table:

Build the portfolio around methods, not outcomes. Use synthetic data or open datasets. For example, run a clustering analysis on an R dataset and write up how it would feed a segmentation strategy. That shows you know segmentation isn't targeting, and that you've thought about why last-click attribution is a trap.
Include code with generic variable names. brand, customer, campaign are fine. It's the same reason I'd share a Screaming Frog crawl logic without exposing a client's URL list. The hiring manager wants to see your process, not your former employer's secrets.
Show your understanding of measurement pitfalls. Write about GIGO, spurious correlation, and why CUPED matters in A/B tests. A portfolio that digs into those issues tells me more than a dashboard full of vanity metrics. I'd rather read a theoretical walkthrough that demonstrates real knowledge than see a polished dashboard that treats correlation as causation.
Keep it easy to review. A PDF with a clean structure works. So does a simple single-page site. Don't overthink the medium. Just make sure the work is organized so someone can skim the methods and then go deep where needed.

You can also create a case study on a fake brand. Make up a scenario, walk through your measurement plan, show the analysis code, and explain what you'd recommend. That gets around confidentiality completely while still proving you can do the job.

What did you end up submitting? I'm curious to see how others handle this.

adcraft

Most portfolios I see are garbage - screenshots of dashboards nobody asked for, vanity metrics dressed up as "strategy." If you're in analytics and strategy, showing the output is a liability. NDAs exist. Competitive intel is real. Handing over a client's segmentation map isn't "transparent," it's a hiring risk.

What I actually want to see: the thinking. How you structured a clustering analysis to pressure-test a segmentation hypothesis. Where you caught a spurious correlation before it became a budget decision. whether you understand that last-click attribution is a lie we tell ourselves to feel good about ROAS.

Fake data is fine. r's built-in datasets are fine. a notebook walking through CUPED variance reduction on synthetic A/B test data? that tells me more than a 50-slide deck of someone else's quarterly review.

Code with generic variable names (brand, customer, spend) - clean, commented, reproducible. No proprietary outputs. If you can explain why you chose a hierarchical model over k-means for that segmentation problem, and where the assumptions break, you're already ahead of 90% of the "growth hackers" listing "data-driven" in their headline.

Theoretical depth beats performative scale every time

avadata

Asking for a portfolio in strategy or analytics is a massive red flag - it signals the hiring team doesn't understand the function. i've sat on both sides of the table at early-stage startups, and the "send me your portfolio" request almost always comes from founders who conflate output with outcomes. they want pretty decks, the role requires SQL pipelines, attribution modeling, and the ability to tell finance why CAC payback shifted 14 days last quarter.

Content and social? Sure, a swipe file makes sense. growth, performance, analytics? The work lives in BigQuery, dbt models, and internal dashboards nobody outside the org should see. Sharing that isn't just impractical - it's a NDA violation waiting to happen.

if a job description for a Senior Analytics Lead lists "portfolio required" under qualifications, I close the tab. They're not hiring a strategist, they're hiring a PowerPoint designer

copycat

That tension between showing results and honoring NDAs is real. performance marketing lives in spreadsheets and dashboards that legally belong to someone else - but the thinking behind those numbers? That's yours to share.

Frame case studies around the problem, the hypothesis, the test structure, and what you learned. "We needed to lower CAC by 20% in a saturated channel. here's the audience segmentation approach I built, the creative angles we tested, and why the winning variant worked." Swap exact figures for directional outcomes - "cut acquisition costs roughly in half" tells the story without exposing proprietary data.

hiring managers care about how you think, not just what the spreadsheet said. Show the architecture, not the address

chatbox

Depends on the audience, honestly. for technical hiring managers, a website with live tagged events and GTM containers shows more than any PDF could. for client-facing work, a clean PDF one-pager per project keeps it scannable. Which format are you leaning toward?

harper-local

I like to call garbage in and out “you’re what you eat.”

What’s the risk of last click attribution? And why not call attribution a journey?