What Recruiters Actually Look For in Your Charts
I have watched hiring managers open a portfolio and skip straight past the fancy dashboards to read the one-line takeaway under each chart. That line decides whether they scroll further. If it reads "Sales by region," they move on. If it reads "Northeast sales grew 22% from Q1 to Q2, driven by a 30% jump in the Chicago territory after the March promo," they stop and read the project. The takeaway is the part they grade first, and most candidates never write one.
The second thing recruiters grade is whether you chose the right chart, not the prettiest one. A line chart for time series, a bar chart for comparing categories, and a scatter plot for relationships. I have rejected portfolios where a candidate used a 3D exploded pie chart for a funnel, because it says they learned a tool demo, not how data communicates. Chart choice is the visible proof of whether you think like an analyst.
Before you build any chart, write the single sentence the chart must communicate. If you cannot say it in one sentence, you are not ready to draw it. That sentence becomes your title, and it should contain a number. "Churn is highest in the first 60 days" is a start; "Customers who churn within 60 days are 3x more likely to have skipped onboarding" is a finding.
Judgment Beats Tool Count, Every Time
A portfolio listing Tableau, Power BI, Excel, SQL, and Python does not tell a recruiter you are a good analyst. It tells them you can click buttons. What tells them you are good is a project where you explain why you chose a bar chart over a line chart, or why you aggregated monthly instead of weekly. I recommend writing one short paragraph per project explaining the choices, because that paragraph is where judgment shows up. It is also the part candidates almost always leave out.
In my experience, the strongest portfolios contain three to five projects, not twenty. Quality collapses past that point. Each project should have a messy input, a clean output, and a recommendation a business could act on tomorrow. A recruiter would rather see one e-commerce funnel project done honestly than five copied tutorial dashboards. If you want to build the underlying analysis skills first, the chart selection guide for Excel covers when each chart type actually earns its place.
Five Project Types That Get Recruiters' Attention
Not every project is worth the space. These five types show up repeatedly in interviews because they map to real analyst work. Pick the three or four that fit the roles you want.
Project 1: A Sales Distribution Map With a Clear Insight
A choropleth map of sales by state is common, but it only earns points if it leads somewhere. Use a public retail dataset with region, state, order date, and revenue columns. The task is to answer one question: where is growth hiding?
Load a regional sales table with columns state, region, order_date, and revenue. Before you chart anything, compute revenue per state and rank it. You are looking for a state that ranks high on revenue but low on growth, or vice versa. That mismatch is your story.
Build the choropleth with states shaded by revenue. Then annotate the two or three states that break the pattern, like a small state with outsized growth. The map becomes a vehicle for the finding instead of decoration.
Finish with a sentence like "Texas holds 18% of revenue but grew only 4% year over year, while Tennessee grew 31% off a small base." That sentence is what gets quoted in an interview.
Project 2: An E-commerce Funnel With the Drop-Off Named
Funnel charts are everywhere, so a funnel alone is meaningless. The version that gets attention names the drop-off and connects it to a cause. Use a sessions table with stage timestamps: landing, product page, add to cart, checkout start, purchase.
Use COUNT(DISTINCT session_id) at every stage so one user counted multiple times does not inflate the numbers. Then compute the conversion rate between each pair of stages, not just the end-to-end rate.
Chart the funnel and label the biggest percentage drop. In most retail datasets that is add-to-cart to checkout. Add a one-line cause hypothesis, like "The shipping estimator appears after checkout start, so cart abandonment spikes when shipping is revealed."
Break the funnel down by source (organic, paid, email, referral). State one concrete action, such as moving the free-shipping threshold above the visible cart total. A recommendation you can say out loud is what separates this from a tutorial copy.
Project 3: A/B Test Results Plotted Honestly
A/B test results are a favorite in interviews because they test whether you understand uncertainty. The classic dataset is a checkout-page experiment with control and treatment variants, each with a conversion rate. Plotting the raw numbers without a confidence interval is a red flag.
Calculate conversion rate per variant, then lift = (treatment rate - control rate) / control rate. With 3.8% against 3.2%, the lift is 18.75%. Record the sample size, because it determines whether that lift is real.
A bar chart with 3.8% and 3.2% side by side looks conclusive. Add 95% confidence interval error bars, and the overlap shows the real story. If the intervals overlap, the right sentence is "we cannot yet say the treatment wins."
End with the recommendation: run the experiment longer, or ship it, or kill it. In my experience, recruiters ask about this project more than any map or dashboard, because it proves you will not overclaim in front of stakeholders.
Project 4: A Customer Segmentation You Can Name
Segmentation projects show you can turn a big table into named groups. Use a transactions table with customer_id, order_date, and order_amount, and segment by recency, frequency, and monetary value. The point is not the clustering algorithm, it is whether the segments make business sense.
Compute recency (days since last order), frequency (order count), and monetary value per customer. Bucket each into High, Mid, or Low, then name the combinations. A "new high-value" customer (recent, few orders, big spend) is a segment a marketer can act on.
Plot frequency against monetary value and color by recency. A scatter plot with a color legend reads better than a 27-cluster dendrogram. Add a two-line profile per segment, like "Champions: order monthly, spend above $120, churn rarely."
Give each segment a concrete play, such as an email cadence for Champions and a re-engagement offer for at-risk high spenders. Segments without an action are just clustering; segments with an action are analysis.
Project 5: An Executive Dashboard That Leads With One Number
A dashboard gets hired when it answers the boss's question in five seconds. The template that works is one headline KPI, one trend line, one comparison chart, and one table with filters. If you are new to the tool, the Power BI first report guide and the Tableau beginner path both walk through a clean four-panel setup.
Choose the single metric the executive cares about most, usually revenue or active users. Put it at the top with the period-over-period change, like "Revenue $2.4M, +14% vs last month." One number, not twelve.
Below the KPI add a 12-month line chart for the trend, one bar chart comparing segments or regions, and a small table you can filter by date or product. Resist adding a fourth chart; restraint is the signal.
Screen-record yourself narrating the dashboard for two minutes. Say the finding out loud, point to the chart that supports it, and state the decision. A narration beats a screenshot because it proves you can talk to a stakeholder.
Chart Choice Principles That Read as Senior
The chart rules are not complicated, but breaking them is how you signal junior. Use a line chart for anything over time, a bar chart for comparing categories, a scatter plot for relationships between two numbers, and a histogram for distributions. Use a pie chart almost never, and never a 3D version. A table is often the most honest choice when you have exactly five numbers to compare.
The second principle is to remove everything that does not carry meaning. Drop gridlines when the values are obvious, remove the legend when there is one series, and label axes in plain English. I usually set the color palette to two or three muted tones and reserve a bright color for the one series I want the eye to land on. If a chart needs a paragraph of explanation to make sense, the chart is wrong, not the reader.
Sort your bar charts by value, not alphabetically. A sorted chart lets the reader see the ranking at a glance, which is what a bar chart is for. I have lost count of the portfolios that alphabetized the axis and hid the whole point behind a wall of text.
Tableau, Power BI, or a Custom Web Page?
The tool choice matters less than you think, and it should follow the roles you want. If the target job postings name Power BI, publish in Power BI. If they name Tableau, use Tableau. If you are applying to startups or want to show front-end skills, a custom HTML page with a charting library reads well, because it proves you can ship something anyone can open without a license.
Whichever you choose, the delivery format stays the same: a landing page per project with a title, a one-sentence finding, the chart, a short write-up of your method, and the data source. I recommend a simple site over a wall of links to public dashboards, because a hiring manager can scan a site in seconds and would have to click into each dashboard to learn anything. If you are still early on the dashboard side, the guide to building a data analyst portfolio covers the broader project set beyond visualization.
The Presentation Format That Reads as Professional
Structure each project the same way so a recruiter knows where to look. Open with the title and the one-sentence finding. Follow with the chart, then a short method note listing the dataset, the tool, and the cleaning steps. Close with the recommendation. Keep the whole project to a two-minute read, because that is roughly the attention budget a hiring manager will give it.
Put the project that matches the target role first. If the posting leads with dashboards, open with the executive dashboard. If it leads with experimentation, open with the A/B test. Do not keep the same order for every application. And match the resume: the resume bullet "Improved funnel conversion analysis by 14%" should point to the same project in the portfolio, or a recruiter will notice the disconnect.
For each project, write the title, the one-line finding, and a short method note. Keep the writing plain and specific. The landing page is the first thing a recruiter reads, so spend the most time here.
List where the data came from, what you removed or corrected, and how you handled missing values. One sentence per step is enough. This note is what answers the inevitable "how did you clean this?" question.
End with the decision and one honest limitation, like "the sample only covers 90 days" or "the funnel does not include mobile traffic." A stated limitation reads as senior, and interviewers respect it.
Reorder the projects for the specific application. The strongest match for the role goes at the top, even if it is not the one you enjoyed most. Recruiters read the first project and judge the rest against it.
Recompute the headline number in your takeaway before you ship. A portfolio with a wrong percentage in the first project is worse than no portfolio, because it signals the analysis cannot be trusted.
What To Leave Out of a Portfolio
Leave out tutorial datasets copied verbatim, dashboards with no takeaway, and charts chosen for looks over clarity. Also leave out the tool-logo wall and any project you cannot explain in two minutes. A hiring manager would rather see four focused projects than twelve shallow ones, and I have seen strong candidates undone by padding.
Before you send the link, open the portfolio on your phone and on a borrowed laptop. Half of the portfolios I review break on mobile or require a login to view the charts. If a recruiter hits a paywall or a blank frame, they will not chase the project. Publish the dashboard publicly or export the charts as images, and put a PDF fallback link on each page.
Your Next Three Actions This Week
Pick one of the five project types and build it this week, start to finish. Write the one-sentence takeaway before you open the charting tool. Then record a two-minute narration of the finished project. That is the minimum viable portfolio entry, and it is enough to send to one real application to test how it lands.
After the first project is live, review the chart choice rules once more, add a second project that covers a different skill, and get the data source and limitation notes written down. Then apply to one role with the portfolio link in the first line of your application. The portfolio earns its keep when it makes a recruiter stop scrolling, and the only way to find out is to send it.
