Most of the writing about e-learning development treats the subject matter expert as a source: you extract the content, you build the course. That is not how it works when the course is any good. The instructional designer and the expert are partners for the length of the project, and the partnership has a shape, with stages that repeat across projects and predictable places where it strains. I want to describe that shape using the project I am in the middle of right now, because a real build, unfinished and with its problems still showing, is more useful than a tidy one.
The project is a math curriculum for South Dakota’s State Personnel Development Grant, delivered on the state Department of Education’s Canvas and Catalog platform. I was brought on as the learning experience designer, not the content expert. The expert is an assistant professor of math education at Northern State University who has been on the SPDG math work since its beginning, through the advisory group, the revision group, and the supplemental documents, and is now writing and filming the course. My job is to get that expertise onto the platform in a form that teachers can learn from in the hours they actually have.
What does each role do?
A learning experience designer is responsible for the shape of the learning: how content is sequenced and chunked, how it is assessed, what the page looks like and how it reads, whether a teacher on a phone at nine at night can find what they need. On this project that has meant deciding how many videos go on a page, what a check for understanding should look like, how the glossary works, and what happens to the course after I am gone.
The subject matter expert is responsible for the content being right, current, and taught in the way the field would recognize as good practice. On this project that means the concrete-representational-abstract sequence, schema instruction, and explicit instruction, and the demonstrations of how they combine across K-8 mathematics. That knowledge does not come from anywhere else. I can build the container, but I cannot fill it.
What follows is the process I use to keep those two roles working together. I think of it in five stages, and I will walk through each with what actually happened.
Define
The first meeting sets scope, audience, objectives, timeline, and deliverables, and it is where I ask the questions that will be expensive to ask later. For the math course the audience was K-8 teachers across the state, the delivery platform was already decided, and the course had a target length of fifteen hours. That number came from the literacy course that preceded it, and the honest answer to “how long is this course” was that nobody knew. We agreed that time-on-platform data from the first cohort would tell us, which is another way of saying the first-year participants would be piloting it.
Two other things got settled early that shaped everything after. External evaluators would need pre and post assessment data captured cleanly, with a true first attempt on the post, because the grant reports to the Office of Special Education Programs. And I would be building a tool to move content from Google Docs into Canvas, and that tool would be handed to the state at the end so the next course would not need me. Both of those are Define-stage decisions even though they feel like Develop-stage details. If you do not know what the evaluators need before you design the assessments, you design them twice.
Design
Design is where both of us sketch a blueprint and then merge them. The expert’s blueprint was the content sequence: ten modules, opening with the strands of mathematical proficiency, then introducing the three instructional approaches separately before showing how they combine, then five video-heavy modules demonstrating the models across number concepts and the four operations, then assessment, then a final module on person-centered IEPs written by a special education colleague.
My blueprint was about what a module looks like on the platform. The videos average eight to ten minutes, with some as short as two and the later ones running longer as the mathematics gets more complex. We decided to put roughly two videos on each page, grouped by topic, with a short framing note on the page about what to look out for and where to take notes. That decision came from the literacy course, where longer pages with more videos had been harder to return to. The content itself lives in Google Docs, and we agreed on a convention where the expert marks figure and video placement with curly brackets so the build tool knows where things go. That convention is the closest thing we have to a shared language, and it removed a whole category of back-and-forth.
Demonstrate
Before building everything, build one thing and look at it together. The first module through the Doc-to-Canvas pipeline showed us what the page structure looked like with real content in it, what the video framing notes needed to say, and where the tool was rough. It was rough. It still is. But the point of the stage is to find out whether the design decisions survive contact with the actual material, and a prototype module answers that faster than any planning document.
The assessment direction came out of this stage too. The course as originally scoped had no checks for understanding and no pre or post. We agreed on scenario-based multiple choice questions, asking which model is most appropriate in a given teaching situation, with images labeled in the question stem rather than as clickable answers, since the answers are teaching decisions rather than pictures. I recommended pools of ten to fifteen questions per module with five served at random. The group was wary of mastery cutoffs that would force a module retake, and I agreed. Where it landed, later in September, was simpler than what I had proposed: one post-test per module using all ten of the expert’s items, unlimited attempts with the highest score kept, and no separate pre-test unless the evaluators ask for one. What the literacy pilot had taught everyone was that the feedback on each answer, explaining why each option is right or wrong, mattered more than the score, so that became the priority.
Develop
Development is the long stage, and it is where the partnership is tested by the calendar. The expert has a full-time faculty role. I was scheduled to need materials by the start of September, which is the busiest stretch of the year at my day job. We set a priority order so that if something slipped, the right thing slipped: question pools first, then the glossary, then the note-taking document.
The glossary uses the HTML abbreviation element so that a term shows its definition on hover, appears on first use on each page, and reads correctly to a screen reader. The expert supplies the terms and definitions; I supply the mechanism. The note-taking document mirrors the one from the literacy course: key terms, fill-in-the-blank learning points, drawing space for the video modules, reflection questions per section, and an application question per module. It is not collected and not required, and it stays a document rather than becoming a platform feature, which is one less thing to maintain.
Accessibility is where the division of labor is clearest. The expert writes alt text directly below each image in the Google Doc, kept under about two hundred characters, and the videos get both a transcript and a separate written description of what is happening on screen, since a transcript of someone demonstrating a fraction model does not tell you what the model looked like. I have been asked more than once whether I could generate the alt text. I could. I would rather the person who chose the image describe it.
As I write this, most of the module content is in hand, the question pools are done for a few modules, and the build tool works but needs attention every time it runs. Review and quality assurance time is the biggest risk to the launch date. That is usually true, and it is usually the thing that gets cut.
Deliver
The last stage is the exit meeting: final review, delivery decisions, and the handoff. On this project the launch is set for October 1, three weeks from the day I am writing. The evaluators use their own survey tool for the periodic feedback the grant requires, and they will supply links to embed rather than having those forms rebuilt inside the platform. They also want observer access for time-on-platform data, which is what will finally answer the fifteen-hour question.
The handoff is the part I care most about, and the part the original version of the five stages did not say enough about. I am no longer the administrator of the state’s Canvas account, and the design pieces I built for it are still in place and still being used. The build tool is being finished in a form that generates a course package file for manual upload, rather than talking to the platform directly, specifically so it can be handed to the state and run by someone who is not me. The Deliver stage covers more than the course going live; it also means the designer leaving in a way that does not take the process with them.
Where does the partnership strain?
Four things go wrong regularly, and they went wrong here in recognizable ways.
The most common is language. The designer talks about chunking and checks for understanding; the expert talks about schema instruction and the CRA sequence. The fix on this project was not a glossary between us but a convention: the curly brackets in the document, the two-videos-per-page rule, the labeled-image question format. Once those existed, most conversations were about content rather than about how to say things.
Depth comes up next. An expert who has spent years on a subject wants all of it in the course, and a designer is trying to protect a fifteen-hour target that may not be real. The way through was to let the module structure carry the argument: the operations modules each move from general concepts through whole numbers, integers, fractions, and decimals, and anything that did not fit that progression was a candidate for a supplemental document rather than a page.
Then there is time. Both of us have other jobs. The priority order was the practical answer, and the honest one, because it admitted in advance that something might not be finished. The thing that will actually be short is review time, and no priority order fixes that.
Vision is the last, and the least trouble here. We did not disagree much about how the content should be presented, but we did have to work through whether the assessments should gate progress. The group’s instinct was against mastery cutoffs and toward feedback, and the literacy pilot had evidence for that instinct. When the designer and the expert differ, the tiebreaker is what the last cohort said.
What holds it together?
None of this works without a few plain agreements. Decide early who has the final say on what: the expert on whether the mathematics is right, the designer on how it is sequenced and assessed. Be honest about time before the calendar forces the conversation. Write down the decisions and the reasons, partly for the project and partly for whoever inherits it, which on this project is the state. And accept that the plan will change when the material arrives, because the prototype exists to change it.
I will know in a few months whether the fifteen hours was close, whether the question pools were the right size, and whether the build tool survived its handoff. If you are on the expert side of one of these partnerships and it looks different from where you sit, I would like to hear that. You can reach me at licht.education@gmail.com, and there are more tools, articles, and resources at bradylicht.com.
Downloads
The PDF below is the 2024 version of this article. Its five-stage process is the same, but the examples in it are illustrative rather than drawn from a real project and are out of date; the article above is the current version.
