Data Storytelling and the Execution Premium Across Six Roles

When I started my career as a data analyst a couple of decades ago, what separated a standout analyst from an ordinary one was easy to name. You could write VBA macros in Excel. You could write SQL that did not collapse on a large join. You knew which system held the numbers and how to get them out. Those were the skills that got you into the room. Data storytelling was not on the list.

They are still useful. But AI now does most of that work, and it does it faster than I ever did. The premium has moved — towards storytelling, towards data lineage mapping, towards understanding why two systems report different numbers for the same quantity. It has moved to a broader set of capabilities.

That is a flattering thing to believe about the profession you came up in, which is why I wanted it tested. We put the question to Curie, our deep research capability, across six knowledge-work functions: Data Analysis, Software Engineering, Product Management, Financial Analysis, HR Business Partnering and Supply Chain Analysis. The evidence base is Draup's live platform data as of July 2026: millions of professional profiles and hundreds of thousands of active job postings. The pattern held, and in several functions it was sharper than I expected.

A balanced market that is quietly splitting in two

One cross-cutting number frames everything that follows. The Talent Hiring Difficulty Index for all six roles in the United States sits in a narrow band between 4.7 and 5.9 on a ten-point scale. That is a balanced market. Supply is abundant, and median experience in these pools runs from seven years for Data Analysts to fourteen for Product Managers and HR Business Partners.

So the question facing a hiring manager is not whether you can find a Data Analyst or a Financial Analyst. It is whether you can find one who is actually good. The index measures aggregate availability, and it is deceptive at exactly the moment it matters. Inside each role the market is bifurcating: the mechanical-execution segment is oversupplied and commoditised, the judgment-and-narrative segment undersupplied and pulling away. The index does not yet capture that split. The skill taxonomy does.

SQL is the floor. Data storytelling is the premium.

The Data Analyst is where the inversion is most visible, because the role's old value proposition was almost entirely technical. Draup's supply-side data shows SQL in 316,011 analyst profiles, the single most common skill in the pool, followed by Data Analysis (285,862), Python (269,967), Power BI (240,158) and Excel (224,913). An analyst who leads with SQL as a headline skill in 2026 is a driver listing the ability to operate a steering wheel.

Demand has not fallen away — 27,248 Data Analyst postings in March 2026 alone, on monthly volumes of 19,000 to 23,000 across the preceding year. What has changed is what those postings ask for. The skills classified as emerging, meaning they are growing in job-description demand faster than they appear in candidate profiles, read as a statement of the new premium: Data Storytelling, Exploratory Data Analysis, DataOps practices, Data Architecture, Use Case Identification, Generative Model Development. The declining skills are the mechanical layer AI is absorbing: MySQL, SSIS, SSRS, QlikView, Visio, Scala.

Then there is the ratio that made the case for me. Storytelling appears in 15,093 analyst profiles. SQL appears in 316,011. For every analyst who can construct a story from data, twenty-one can query it. That gap is the premium, and it recurs across each of the five capabilities the data points to: narrative construction; statistical depth past the regression itself, where Hypothesis Testing appears in only 6,603 profiles; detection of data that is simply wrong; lineage tracing, with Metadata Management in 7,134 profiles; and the confidence to challenge the question before answering it. AI will produce a beautifully formatted wrong answer from a corrupted upstream feed. The analyst who can trace the number back through three systems and three transformations catches it before the CFO takes it to the board.

The same inversion, five more times

Every function shows the same shape. The saturated skill is held by hundreds of thousands; the premium capability by a fraction.

• Data Analyst — Saturated skill: SQL · Profiles: 316,011 · Premium capability: Data Storytelling · Profiles: 15,093 · Ratio: 21:1
• Software Development Engineer — Saturated skill: Java · Profiles: 1,805,761 · Premium capability: Systems Architecture · Profiles: 147,706 · Ratio: 12:1
• Product Manager — Saturated skill: Product Management · Profiles: 279,308 · Premium capability: Stakeholder Management · Profiles: 55,764 · Ratio: 5:1
• Financial Analyst — Saturated skill: Microsoft Excel · Profiles: 218,599 · Premium capability: Business Partnering · Profiles: 5,145 · Ratio: 42:1
• HR Business Partner — Saturated skill: HR Business Partnership · Profiles: 182,498 · Premium capability: Courageous Conversations · Profiles: 1,260 · Ratio: 145:1
• Supply Chain Analyst — Saturated skill: Supply Chain Management · Profiles: 88,838 · Premium capability: Supply Chain Resilience · Profiles: 302 · Ratio: 294:1

The ratios are not random. They track how much of each role's pre-AI value sat in mechanical execution. The Product Manager, where judgment was always central, shows the most compressed ratio; Strategic Thinking is already the top soft skill in that pool at 317,132 profiles, ahead of Leadership. A/B Testing and Data-Driven Decision Making, at roughly 25,000 profiles each, have become table stakes. The PM who leads with "I'm data-driven" is describing the floor.

At the other end sit the roles whose execution layer was most dominant. Java appears in 1,805,761 engineer profiles; GitHub Copilot, the tool that writes Java, appears in 50,576. Postings now ask for Copilot (28,576 mentions), Cursor (17,754), Prompt Engineering (13,445), LangChain and retrieval-augmented generation. The market is not asking for better Java. In Financial Analysis, Business Partnering — a skill that barely existed in the taxonomy a decade ago — now appears in 8,732 job descriptions against 5,145 profiles.

The number I keep returning to is the HRBP one. Courageous Conversations appears in 1,260 profiles against 182,498 carrying HR Business Partnership. The HRBP who can tell a senior leader the truth about their management style, or a CEO the truth about the culture, is doing the part of the job no system will take. I did not expect a skill taxonomy to surface something that reads like a character trait, and it was the finding that stayed with me.

What the pay curves are actually measuring

Compensation confirms the bifurcation with unusual consistency. Every one of the six roles shows a P90/P10 ratio between 2.0 and 2.3. Software Engineers run $96,366 to $135,650 to $200,692 on an evidence base of 6.2 million reported salaries. HR Business Partners run $80,346 to $111,421 to $161,527. Data Analysts run $63,390 to $88,140 to $146,029. These are not commodity markets.

The shape of the pay-by-experience curve matters more than the spread. Pay is essentially flat through the first four to seven years, accelerates sharply between years five and fifteen, then plateaus. The acceleration is not bought with more SQL, more Java or more Excel — those hours accumulate steadily throughout. It tracks domain knowledge, interpretive judgment and organisational credibility. The plateau after year fifteen suggests that seniority without differentiated capability stops earning anything extra.

Audit the job descriptions before the next requisition

The most useful action available in the next thirty days is unglamorous: read your job descriptions across these six functions and take the mechanical execution skills out of the screening criteria. A requisition that leads with five years of SQL or proficiency in Excel selects for the floor of the post-AI talent market and pays market rates for it.

Three changes to hiring practice follow. Screen for the premium capabilities directly — give a candidate a dataset with a deliberate error and ask them to find it; ask them to explain a complex finding to the non-technical person in the room; ask them to challenge the premise. Treat the technical screen as a threshold rather than a differentiator. And look for the emerging-skill signals by name: Data Storytelling, Business Partnering, Courageous Conversations, Supply Chain Resilience, Systems Architecture.

For L&D functions, the emerging skill list for each role is not a forecast. It is a live signal of what employers have started requiring before supply caught up, which makes it a ready-made curriculum. For individuals, the counterintuitive part is that the path to the P90 band is not more tools. It is deeper domain knowledge in the industry you serve, statistical literacy past the textbook, and the ability to make a room of non-specialists act on what you found. The analyst who finds the error but cannot convince the CFO it matters has not added value. The HRBP who diagnoses the culture problem but never has the conversation has not solved it.