# One Number, Two Stories: What a State Agency and a University Can Teach Us About Counting Accurately *SDP Fellow Molly Tuck at South Carolina First Steps was tasked with understanding how many students in the state were Kindergarten-ready. At Claflin University, SDP MSI Fellow Rahat Khan needed to provide a clear number for how many freshmen were enrolled. Both questions should have an easy, accurate answer. But in reality, there's much more to it.* November 12, 2026 - [ Blog ](/news-categories/blog) Imagine you work at a state education agency. Leadership asks, "How many students in our state are kindergarten-ready, and who are they?" Now imagine you work at a university, and the Dean needs the final number of incoming freshmen. Both questions sound simple. Both should have one answer, from one source, in minutes. In practice, that's rarely what happens. Two organizations at opposite ends of the education pipeline, South Carolina First Steps and Claflin University, ran into the same problem: the number depended on who you asked. The reasons were very different, and so were the fixes. **Two Questions, One Hidden Problem** *Early childhood: "How many kids are kindergarten-ready?"* South Carolina First Steps is a state agency and nonprofit offering a wide range of early childhood services, including First Steps 4K (free, full-day 4K for four-year-olds) and local partnership programming for families and caregivers. Molly Tuck, an SDP Fellow at First Steps, set out to learn which combination of these services is associated with stronger Kindergarten Readiness Assessment scores. Answering that would show where gaps exist, who is being served, and how many children are truly being reached. But she couldn't get to those questions yet. First she had to find out how many children were actually in the data. *Higher education: "How many freshmen did we enroll?"* Claflin University is the oldest historically Black college in South Carolina, with a student body of 1,755 students and 40 undergraduate majors, six graduate degrees, and 10 online degree programs. Rahat Khan, an SDP MSI Fellow, found that data was available and reporting was active. Yet the incoming freshman cohort number changed depending on which office answered the request. On the surface, the two situations look unrelated. Underneath, both organizations had data they couldn't fully trust. **Why the Numbers Didn't Add Up** *Tuck's problem: the same child, entered twice* When Tuck tried to connect data between First Steps 4K and the local partnership programs, she found that data entry errors were creating duplicate records. A child might be entered as "John Doe" on one form and "Jon Doe" on another. Same child, two records. That led to two serious problems: - Inflated counts. Programs looked larger than they were. - Broken tracking. A child split across two records can't be followed over time, which defeats the purpose of a longitudinal data system. *Khan's problem: the same student, defined differently* Claflin's issue wasn't typos. It was ownership and meaning. Responsibility for the data was diffuse, and definitions varied across offices. Nobody had formally been assigned to resolve differences, so validation happened under deadline pressure. When a high-stakes report was requested, staff spent significant time tracing numbers back to their sources and debating which version was authoritative. That eroded confidence in what the numbers meant. Though both Molly and Rahat had different problems, both ended in the same place: a number no one could confidently stand behind. **Fix #1: Teaching the Software to Recognize "Pretty Sure"** Exact matching, known as deterministic matching, only links records when identifiers match perfectly. Real-world data is full of typos, nicknames, and inconsistent entries, so "John" and "Jon" would never connect. Tuck used probabilistic matching instead. Rather than asking, "Are these records identical?", it asks, "How likely is it that these records describe the same person?" In other words, \*we're pretty sure this is the same child.\* **Fix #2: Getting Everyone to Agree on What "Correct" Means** Claflin couldn't solve its problem with a better algorithm. The cohort number differed because people defined and handled the data differently, so the fix had to be about people and process. *Step one: a structured self-assessment.* Seven departments examined their own practices across five areas: ownership, definitions, access, quality, and reporting alignment. The results made visible what informal conversations hadn't. Most departments couldn't clearly identify a data owner, definitions varied across units, and few kept documented glossaries. The assessment didn't uncover entirely new problems. What it did was give staff a shared view of conditions they had previously experienced separately. Instead of anecdotes and general frustration, leaders had documented evidence of gaps in ownership and meaning. *Step two: a Data Governance Charter.* Built from what the departments themselves identified, the charter spelled out who would define, maintain, and validate key data, and how disagreements would be resolved. The aim wasn't to put one department in control. It was to establish clear accountability, much like assigning parts within an ensemble so everyone knows their role. Claflin first applied the charter to data elements where unresolved ownership or meaning had visible consequences, which tested whether the roles and decision processes worked in practice. As responsibilities became clearer, departments no longer had to renegotiate authority every time a question came up. **Two Tools, One Goal** It's tempting to see these as separate stories, one technical and one organizational, but they reinforce each other. Both organizations diagnosed the problem before fixing it: Tuck found the duplicates before analyzing outcomes, and Khan used a self-assessment to document gaps before writing the charter. Both also aimed for a number leaders could trust, and neither stopped at a one-time cleanup. [Tuck pushed for a centralized data system so duplicates are prevented at the source and built tools so staff could use clean data in real time](https://sdp.cepr.harvard.edu/preschool-supply-and-demand-toolkit), while Kahn's charter [makes accountability permanent so staff spend less time tracing numbers and more time acting on them.](https://sdp.cepr.harvard.edu/key-of-c) Neither approach is enough alone. Clean matching without agreed-upon definitions leaves you arguing about what counts, and clear definitions without clean records leave you with inflated numbers. **Why This Matters Beyond** A true count is the foundation for every other question you want to ask: how many, who, and where the gaps are. Without a validated baseline, it's easy to mistake data errors for real gaps, or to produce findings that stakeholders reject because they don't match what they know to be true. **The Bottom Line** Before you can say how many children are kindergarten-ready, or how many freshmen enrolled, you have to be sure each person is counted once and counted consistently. Molly Tuck's linkage guide and Claflin's governance charter are two very different tools solving the same core problem: building trust in the numbers. Whether you're counting preschoolers or college freshmen, the answer to "how many?" is only as good as the data processes and the people behind it. \--- Access the SDP Preschool Supply and Demand Toolkit for more about Molly and her colleagues’ work on early education data: Review the MSI Practice Guides for more on Rahat Khan and his colleagues’ work on data governance in postsecondary education: See also:- [ SDP Blog ](/tags/sdp-blog) - [ Data Governance ](/topics/data-governance) - [ Early Education ](/topics/early-education) - [ Postsecondary Success ](/topics/postsecondary-success) - [ Data Systems and Culture ](/resource-topic-category/data-systems-and-culture) - [ Early Childhood Education ](/impact-areas/early-childhood-education) - [ Postsecondary Success ](/resource-topic-category/postsecondary-success) Share on:- [ Facebook ](#) - [ Twitter ](#) - [ Linkedin ](#)