Charles A. Hollenbeck

Multi-subject instructional content system

West Coast University · Design for Six Sigma (DMADV) · 240+ published assets, 16-person team

An academic resource center had no instructional content at all: no repository, no production process, and no route by which student demand could reach anyone able to answer it. Subject expertise, information management capability, and a publishing platform all existed, connected by nothing.

Over eighteen months I led a cross-functional team of librarians, subject faculty, information management staff, and academic leadership to produce 83 written guides, 21 explainer videos, and at least 136 faculty-produced subject lectures across six disciplines. Topic selection ran off a monitored stream of student search inquiries. Production was distributed to six subject experts who together delivered 1.31 times my own output while carrying full teaching loads.

I could not have specified what a microbiology or pathophysiology video should contain. I could specify when it was due, what it should look like, and how its topic would be chosen. Separating the what from the when and how is what let one coordinator direct production across six disciplines he was not qualified to teach. The collection remains live and institutionally maintained more than two years after my departure.

Read the full portfolio (PDF)See the guides

Asynchronous writing feedback service

West Coast University · DMAIC · 8 campuses, 16 degree programs

A multi-campus university had no scalable way to give students written feedback on drafts. I designed and launched a drop-off service across eight campuses and sixteen degree programs, staged campus by campus, running to a 24 to 48 hour turnaround standard with a distributed team of eight to twelve instructors.

The service worked. What it could not do was prove it. The intake instrument captured a single timestamp, so the turnaround commitment the entire service was built around was never measurable from the operating record. I reported the measurement gap instead of asserting a result the data could not support.

Resident activity programming capacity

State-regulated assisted living residence · DMAIC · 706 sessions across 7 months

A residence serving seventy assisted-living residents ran a limited activity program carried largely by one coordinator. I spent a week observing before changing anything, surveyed residents, talked with them about what they wanted, and rebuilt the schedule.

Programming rose 56 percent, from 2.26 to 3.51 sessions per day, and held for six months. The increase was additive, not a reallocation. Where the capacity came from is the interesting part: not longer days and not more staff. Early mornings rose sevenfold, mid-afternoons more than tripled, weekends more than doubled. The building had the hours all along and nobody had scheduled into them.

What I could not establish was whether the program reached more residents or gave the same residents more to do, because the attendance record attributes counts to a calendar cell, not to a session. I reported that as the project's primary constraint.

Contracted tutoring supplier quality

Vendor management · DMAIC · 1,292 sessions, 514 students, 652 tutors

An institution had contracted a third-party tutoring provider and had no way to assess whether it was delivering. I built the analysis from the vendor's own session records.

One live session in five ended within five minutes, at a rate stable across nine months and therefore systemic rather than incidental. Failure concentrated in the institution's own core subjects, not in general education, meaning the supplier performed adequately at what any institution teaches and poorly at what this one existed to teach. Three hypotheses died along the way, including a modality confound that had been inflating apparent performance; once stratification removed it, the defect rate rose rather than fell.

Constructed-response scoring reliability

Standardized assessment · DMAIC · Method demonstration on constructed data

Human raters applying an ordinal rubric produce variation. Where two raters disagree by two points or more, adjudication follows, score release slows, and a student's classification depends partly on who received their response. A generative model with documented parameters and seeds produces 1,000 double-scored responses and a repeated-anchor study across ten raters, two trials, and two phases.

A factorial experiment produced a result I did not expect: an expanded exemplar library achieves almost nothing on its own, and only works when combined with a mechanism that makes raters revisit it. One target was missed by two tenths of a point. The document records the decision taken in response, declines to lower the threshold to meet it, and carries it forward as an open item.

This project uses simulated data throughout. It is grounded in my work as a constructed-response scoring professional and demonstrates method; it does not report work performed for any employer.

Written guides

83 guides across six disciplines

Topic selection ran from a monitored stream of student search inquiries, not from what instructors assumed students needed. The collection is still live and institutionally maintained.

Open the guides

Video library

21 numbered explainer videos

Each video answers one question in the time a student is willing to give it. The full library is hosted by the university and requires an institutional login. Three samples are below.

The library at West Coast University

How can I apply my academic writing skills to the workplace?

What is critical analysis, and what are the pitfalls of relying on AI?

What is the difference between qualitative and quantitative research?

AI-assisted microlectures

Course video series, 2026

Short video lessons rebuilt from the written guides, made for my own composition courses and shown here as samples of a process that runs at scale. Each script is rewritten for spoken delivery rather than read off the original guide, the visuals are held to what a viewer needs at that point in the lesson, and a topic is adapted to the program the students are in when that changes what the lesson has to do. The written collection behind them runs to 83 guides, and the same process converts any of them.

When should a multilingual writer keep a word in another language?

A lesson for multilingual students on when another language belongs in an academic paper. It separates code-switching from borrowing, works through the Spanish word sobremesa as a case where translation costs the reader something, and sets the test that decides the question: what the original language adds, and whether the reader has been given enough context to follow it. The premise is that multilingualism is a rhetorical resource rather than a problem to be corrected out of the draft.

How does academic writing carry into a healthcare workplace?

The nursing and healthcare version of the workplace-writing question answered more generally in the library above. It follows four habits, clarity, organization, precision, and audience awareness, out of the classroom and into the documentation, handoffs, and patient instructions a clinician writes daily, and turns on a pair of requests to a colleague, one vague and one specific, that shows what precision buys the reader.

What is technical writing, and who is it for?

Defines technical writing by what it has to do for a reader rather than by how it sounds, names the four document families a graduate is most likely to write, and turns on the difference between writing for an expert, a new user, and a decision-maker who each need the same information for different reasons.

The presenters and voices in these videos are AI-generated, and differ from one video to the next. Scripting, instructional design, and editorial review are my own.

Objective: Analysis: Perception, Proofs, and Precision

Kendall Hunt, 2026 · 112 pages

This is not a composition textbook. It follows information through three stages, reception, processing, and transmission, and names what goes wrong at each one: biases as information comes in, complexes while it is held, fallacies when it goes back out. It then runs that same sequence through a second axis, the conscious level, the subconscious level, and memory itself, where information is written, stored, and recalled.

Students who can locate a problem on that grid gain something most of them never get from a rhetoric text, which is the ability to watch their own reasoning while it is happening. The practical payoff is the thesis statement. It is the single hardest thing to teach in composition, and it gets easier the moment a student understands a thesis as a claim that has survived its own processing and not merely a sentence to be produced on demand.

Publisher page

The Argument Engine

Argumentation theory · Monograph in submission

A framework for evaluating arguments in the ordinary case, where the evidence will not resolve into proof and a decision still has to be made.

The Borrowed Sciences

History and sociology of expertise · Monograph in submission

What happens when United States security institutions adopt social science methods, and why some of those borrowings survive while others quietly fall out of use.