Data storytelling for institutional research: turning numbers into narratives

A graduation rate is a sentence with the ending cut off. Data-informed advocacy is the skill of finishing it for the person in the room.

Wilson BrightWilson Bright
June 9, 2026
10 mins read
Share:
Table of Contents

Introduction

Picture the moment. The data request that took three days to fulfill finally lands in a leadership meeting. Someone reads it aloud: "Our graduation rate is 62%." A few heads nod. Someone else says, "Okay, good to know." The conversation moves on. Nothing happens.

That is the failure mode that no amount of analytical rigor fixes. The number was right. The query was clean. The methodology was sound. And it changed nothing, because a data point is only ever as good as how it is interpreted and communicated. A graduation rate dropped into a room without context is a sentence with the ending cut off.

This post is an argument about the skill that closes that gap. Call it data-informed advocacy, or data storytelling, or just finishing the sentence. It is the part of the job that does not show up in the SQL and rarely shows up in the training, and it is the part that decides whether your institution acts on what you found. Here is the uncomfortable version: if you do not tell the story behind your data, someone else will, and they will get it wrong.

Same number, finished sentence

Take that 62% and finish the sentence. (To be clear, 62% here is an illustrative figure, not any one institution's reported number. It tracks closely with the real national benchmark: the 6-year graduation rate for first-time, full-time bachelor's-seeking students at 4-year institutions was 64% for the fall 2014 cohort, per NCES.) The bare number says one thing. The finished sentence says something a decision-maker can act on:

"Six out of ten students, many of them first-generation and Pell-eligible, cross the finish line and enter the workforce."

Same data point. Completely different object. The first is a status line. The second is a claim about who your institution serves and what it delivers. Notice that the finished sentence is not spin; every word of it is defensible from the same dataset. What changed is that the analyst chose to surface the part of the data that matters to the person listening, instead of handing over the raw figure and hoping they would do that work themselves.

That choice is where advocacy starts. And it is also where the difficulty starts, because the right ending to the sentence is different for every audience.

One number, four audiences

AudienceWhat they actually care aboutHow you finish the sentence
Campus leadershipStudent success momentum and where the institution is gaining or losing groundSix in ten finish, and the first-to-second-year gain this cohort is the lever we should be talking about.
State agency or system officeWorkforce alignment and affordability across the public missionOur completers enter the regional workforce at competitive earnings, and the cost-per-completion compares favorably to peers.
AccreditorsAssessment, evidence, and continuous improvement over timeHere is the rate, here is the trend across three cohorts, and here is the intervention we tied to the movement.
Employers and regional partnersWhether graduates stay local and arrive ready to workA majority of our completers stay in the region, which is your future hiring pipeline.

The number in the leftmost mental column never changed. The sentence did, four times. That is not a trick of presentation. It is the recognition that the same fact carries a different decision for each person who hears it, and your job is to hand each of them the version that connects to the choice in front of them. Do that well and the data starts to move. Hand everyone the raw 62% and you are back to nodding heads.

Three lessons for data advocacy

1

Know your audience and finish the sentence

Stop dropping the data bomb

The instinct is to deliver the number and let it speak for itself. Numbers do not speak for themselves; that is the whole problem. Before the data leaves your hands, decide who is receiving it and what decision they own, then finish the sentence in their language. The same graduation rate is a momentum story for a president, a workforce story for a legislator, and an evidence story for an accreditor. Leading with the finished sentence is not dumbing down the analysis. It is doing the last and hardest mile of it.

2

Not all data is worth the squeeze

Ship directional, skip the page of caveats

A request that warrants a directional answer does not warrant a week of precision. Before you spend that week, ask the decision rule out loud: will more precision change the decision? If a stakeholder needs to know whether a number is roughly 60% or roughly 90%, a same-day directional figure ends the conversation, and the page of methodological caveats you were about to append only buries the signal. Reserve the deep, every-decimal rigor for the requests where the decision genuinely turns on it. This is the same scoping muscle that good request intake builds: deciding up front how much rigor a question actually deserves.

3

Start with why

Emotions drive decisions, data justifies them

Simon Sinek built the Golden Circle (why, how, what) on the claim that people respond to purpose before particulars. In his framing, the "why" speaks to the limbic brain that governs emotion and decision-making, while the "what" speaks to the rational neocortex. Treat that as his model rather than settled neuroscience, but the practical takeaway holds for IR: numbers do not move people, stories do. People decide on the why and then reach for the data to justify the decision. Lead with the human stake, then let the data make the case airtight. Sinek first delivered this at TEDxPugetSound in 2009, and it became one of the most-watched TED talks of all time, which tells you something about how hungry people are for the why.

Start with why, made concrete

Here is what "start with why" looks like when there is real money on the table. Georgia has a program called the Tuition Equalization Grant, created by the General Assembly in 1971 to help Georgia residents attend in-state private colleges. The legislature sets the award amount each session, which means every year the grant is, in effect, up for renegotiation. Programs like that live or die on whether anyone can make the case for them in a way a legislator feels.

The data case is straightforward: X residents served, Y dollars per student, Z completions in fields the state says it needs. (In the 2022 session, the General Assembly even considered narrowing eligibility to nursing and STEM majors to chase workforce priorities; that legislation did not pass, but it shows how readily funding conversations get framed around workforce.) All of that is true and necessary. It is also, on its own, the kind of thing legislators shrug at.

What follows is an illustration of how the why lands, not a documented case. Imagine the grant is a modest sum, and lawmakers are about to let it slide, until a student who waited tables to stay enrolled writes to them about exactly how many tables that grant represented: how many late shifts it covered, how many semesters it kept her from dropping. Suddenly the small number is not small. The data established that the program is real and measurable. The framing made it matter. The why gave it a pulse. And the human face made it impossible to ignore.

Layer it in order: data to prove it is real, framing to show why it matters, a why to give it a pulse, and a human face to make it impossible to look away from. Skip the last three and you have a spreadsheet nobody acts on.

The power and fluency grid

"Know your audience" is good advice and useless until you make it concrete. Here is a grid I find more practical than personas: plot the person on two axes. One axis is how much power they have to decide. The other is how fluent they are with data. Four quadrants, four different ways to finish the sentence.

QuadrantWho they areHow to reach them
Low power, low fluencyA community leader or local advocate who cares but cannot read a regressionLead with a story and one clean infographic. The narrative is the payload; the chart is the proof.
Low power, high fluencyA fellow analyst or research partnerThis is your ally. Test the message on them before it goes up the chain. They will find the hole an executive would have found in public.
High power, low fluencyA decision-maker who controls the budget but skimsNo chart that needs a legend. Make the takeaway the title of the chart, so the headline is unmissable even at a glance.
High power, high fluencyA data-literate executive or board memberGo deeper with a dashboard and the ROI, and still frame it with mission and impact. Fluency does not mean they stopped caring about the why.

The grid is a reminder that fluency and power are independent. A brilliant analysis aimed at the wrong quadrant fails the same way a sloppy one does. The high-power, low-fluency quadrant is where careful IR work most often dies on the table, because the analyst defaults to the depth they would want and the decision-maker checks out before the insight arrives. Make the takeaway the chart title and you have met them where they read.

Where the data hides in plain sight

Finishing the sentence often means reaching past your own warehouse. A graduation rate becomes a workforce story only when you can say what completers earn and whether they stay. Most of that data is public and already collected. It is sitting in plain sight.

  • IPEDS for enrollment and completion. The graduation-rate component counts completers within 150% of normal time at the same institution, divided by the adjusted cohort. Worth knowing its limitation out loud: it only counts first-time, full-time entrants, which undercounts transfer-in and part-time students.
  • College Scorecard for earnings and debt by program. It publishes institution- and program-level (4-digit CIP) data on completion, debt at graduation, loan repayment, and earnings measured 4 years after the credential, compared to high school graduates nationally and within state.
  • Census Post-Secondary Employment Outcomes (PSEO) for earnings by institution and major. It reports graduate earnings at the 25th, 50th, and 75th percentiles at 1, 5, and 10 years out, by matching transcript records to a national jobs database.
  • American Community Survey and Census data for the regional demographics and net migration that tell you whether your graduates stay.
  • Federal Student Aid loan-portfolio data for the affordability and repayment side of the story.
  • WICHE "Knocking at the College Door" projections for the demographic cliff hiding in plain sight: U.S. high school graduates are projected to peak in 2025 and then decline about 13% through 2041, with future classes more Hispanic and Multiracial.

One caution that doubles as a permission slip: just because the data are not perfect does not mean they are not usable. PSEO and the Scorecard field-of-study data are explicitly experimental and partial. PSEO only covers institutions with data-sharing agreements and pools multiple award years. The Scorecard's earnings figures carry their own coverage caveats. Knowing exactly why a dataset is imperfect is what lets you use it responsibly: you cite the limit, you frame the number as directional, and you finish the sentence anyway. Waiting for perfect data is how the story goes untold.

Where Clema fits in

None of this advocacy happens if the analyst never gets out of the warehouse. In our whitepaper drawn from 50-plus interviews with IR professionals across 19 states, the recurring theme was not a shortage of analytical skill. It was a shortage of time, because ad-hoc requests consume the capacity that interpretation and storytelling would otherwise use. You cannot finish the sentence on a number you are still busy producing.

That is the case for answering data requests in minutes instead of weeks. Not so the team can field more requests, but so the freed time goes to the work that actually moves people: choosing the audience, finishing the sentence, finding the human face. The alternative is giving away free labor on the Nth decimal of a number nobody will act on regardless. The same logic runs through our best practices for handling data requests: reclaiming team capacity is the precondition for advocacy, not a nice-to-have beside it.

Every row in your dataset is a student. The whole point of advocacy is to keep that true when the data reaches the room.

See what 50-plus IR teams told us

Our whitepaper, drawn from interviews with IR professionals across 19 states, lays out where institutional research capacity goes and how teams reclaim it for the interpretation work that drives decisions.

Read the whitepaper

Acknowledgment

Thanks to Carolyn Mata of the Council of Independent Colleges, whose keynote on data-informed advocacy at the South Carolina Association for Institutional Research conference (SCAIR 2026) shaped much of the thinking in this piece.

Frequently asked questions

What is the difference between data reporting and data-informed advocacy?

Reporting hands over the number: "our graduation rate is 62%." Advocacy finishes the sentence for the person receiving it: "six in ten students, many first-generation and Pell-eligible, finish and enter the workforce." Same data, but advocacy surfaces the part that connects to the decision in front of the listener. Reporting informs; advocacy is built to make someone act.

How do you frame the same number for different audiences?

Decide who is receiving it and what decision they own, then finish the sentence in their language. The same graduation rate is a student-success momentum story for campus leadership, a workforce and affordability story for a state agency, an assessment and continuous-improvement story for accreditors, and a local-hiring-pipeline story for employers. The number never changes; the ending of the sentence does, every time.

What is the decision rule for how much precision a request deserves?

Ask it out loud: will more precision change the decision? If a stakeholder only needs to know whether a figure is roughly 60% or roughly 90%, a same-day directional answer ends the conversation, and the page of caveats you were about to attach only buries the signal. Reserve every-decimal rigor for the requests where the decision genuinely turns on the exact number.

How should I tailor a message using power and fluency?

Plot the person on two axes: their power to decide and their fluency with data. Low power, low fluency gets a story and one clean infographic. A high-fluency analyst is your ally to test the message on. A high-power, low-fluency decision-maker needs the takeaway as the chart title, no legend required. A data-literate executive can go deeper, but still frame it with mission and impact.

Which federal datasets help finish the sentence on a graduation number?

IPEDS for enrollment and completion, College Scorecard for earnings and debt by program (4 years after the credential), Census PSEO for earnings by institution and major at 1, 5, and 10 years out, the American Community Survey for regional demographics and migration, Federal Student Aid for loan-portfolio data, and WICHE projections for the demographic outlook. Most are public and already collected.

How do you use imperfect or experimental data responsibly?

Just because the data are not perfect does not mean they are not usable. PSEO and College Scorecard field-of-study data are explicitly experimental and partial; PSEO only covers institutions with data-sharing agreements and pools multiple award years. The fix is to know exactly why a dataset is limited, cite that limit, frame the figure as directional, and finish the sentence anyway. Waiting for perfect data is how the story goes untold.

Ready to get started?

Reclaim Your Team's Capacity

See how Clema can help your IR team handle routine requests automatically