The Scaffolded CFP: how to design a call for proposals that gives you the lineup you actually want

Most CFP problems aren't selection problems. They're shaping problems.

By the time you're reviewing four hundred submissions, the damage is done. Your funnel was unshaped, so your lineup is unshaped. Fifteen submissions on the same hot topic. Six excellent talks crammed into one bucket, zero in another. The same circuit speakers as last year, because nothing about your call signaled to anyone else that they should bother.

If your CFP keeps producing the wrong lineup, the talent pool isn't the problem. The shape of your funnel is.

A scaffolded CFP flips the work. You build the shape of the program before opening submissions. Not session titles. Topical buckets that span what the conference needs to cover. You make those buckets visible. Submitters see what you need.

Now the funnel does work for you. Some applicants slot into needed buckets. Some see a crowded bucket and reconsider whether their angle adds anything. Some submit outside the buckets because their idea is genuinely exceptional, and you keep that lane open. Signal-to-noise jumps. Not because you scored harder. Because you shaped earlier.

This post is the method. Tool-agnostic. You can run it in a Notion doc and a Google form. Easier in Gythr because we built the scaffolded CFP into the module, but the principle is what matters.

Three types of speakers, three different behaviors

Every CFP gets three types of submitters. The scaffolded CFP serves all three differently. Worth understanding before you build the buckets.

Circuit speakers. Same talk every conference. They've polished it. They know it works. They're not tailoring it to your event because their value is the talk, not the adaptation. With an unshaped CFP, circuit speakers dominate. They're known names, they submit instantly, their abstracts are tight, reviewers default to "they'll be fine." With a scaffolded CFP, you can see them clearly. Their submission slots into one bucket only and they make no attempt to address what you said you needed. That's not a knock on them. It's information. Sometimes you want the circuit speaker. Sometimes you don't.

Dynamic speakers. These are the ones who can give five different talks depending on what you need. Their problem is that with an unshaped CFP, they're guessing. They write the abstract they think will get accepted, which means they hedge toward what's worked before. With a scaffolded CFP, they don't have to guess. They see the buckets, see what's underserved, craft to fit. The lineup gets sharper because the most adaptable speakers are aiming at the right targets.

Emerging voices. Speakers who haven't done a major conference yet. Often the freshest content, often the least-polished pitch. With an unshaped CFP, they look at the existing lineup and assume the crowded categories are the only categories that matter. They don't apply, or they apply to the wrong bucket. With a scaffolded CFP, they can see "this category needs voices" and feel licensed to bring theirs. Some of the best lineup decisions I've watched come from this. The under-known speaker who submitted to a bucket nobody else covered, who turned out to be the talk people remembered.

Three types. Three different behaviors when the funnel is shaped. Build the scaffold knowing all three are on the other end of it.

Build the scaffold before you open the call

Three to five hours of team work. Output: a one-page document. Here's the work.

Write the scope as a paragraph, not a list. Who's coming. What they're navigating. What they need to hear from your stage that they aren't hearing elsewhere. Not the mission statement. The actual thing. If you can't write that paragraph, you can't scaffold a CFP yet. Stop and figure that out first.

Find what's been over-served and under-served. If you've run this conference before: which topics had three sessions when one was enough? Which had zero when they should have had two? Year one? Look at three peer conferences in your space and ask the same question.

Draft 4 to 7 buckets. Fewer than four and everything slots in. You've shaped nothing. More than seven and it becomes a tag list, which submitters skim past. Four to seven is enough.

Get the specificity right. Each bucket should be specific enough that you can name three different sessions that would fit it, broad enough that you can imagine ten different angles within it. "AI" is too broad. Every submission slots in. "Building inclusive AI products in the federal government" is too narrow. Nobody applies. "AI ethics in product decisions" is about right. Test: would a thoughtful speaker know exactly what this bucket is asking for and feel like they have room to be original within it?

Stress-test with two or three people who'll attend. Not your team. Audience-side people. Show them the buckets, ask what's missing, what's overlapping, what would you submit to. Their reaction is data. If three people independently say "I don't know what this means," rewrite it. If everyone slots their hypothetical submission into the same bucket, you're under-shaped.

Decide what's public and what stays internal. Buckets get published. Reasoning behind the buckets, what you over-served last year, what sponsors are funding, which voices you're centring this cycle, stays internal. Submitters see the shape. They don't need to see the math.

Leave an "outside the scaffold" lane. Some of the best talks come from people who looked at your buckets, decided their idea didn't fit, and applied anyway. The scaffolded CFP isn't a closed funnel. It's a shaped one. The next section covers how to weight that lane.

Surface the scaffold to submitters

The buckets do nothing if submitters can't see them or don't trust them. Two design choices matter here.

Where the buckets live on the form. Top. Before the abstract field. Before the bio. The first thing a submitter does is read the buckets and pick one. This is what trains them: this CFP is shaped, you're choosing your lane, your job is to make a strong case for the lane you picked.

What each bucket says. Each bucket needs a name and a one-line description. The name is searchable shorthand. The description is the actual frame. "AI Ethics in Product Decisions" is the name. "Talks about how teams are making real product decisions when AI is part of the stack: what got built, what got cut, what changed in how decisions get made" is the description. Submitters read the description and self-evaluate. About a third will pick the wrong bucket anyway. That's fine. Reviewers can re-bucket on the back end.

The crowding question. Some events show submitters the relative crowding of each bucket as the CFP runs. ("This bucket has 40 submissions. This one has 4.") We've watched this go both ways. Crowding visibility helps emerging voices feel licensed to fill underserved buckets. It also creates a small risk that good submissions to crowded buckets self-censor when they shouldn't. Our recommendation: surface crowding only after a certain volume of submissions (don't show it the first week), and frame it as "where we still need voices," not "where the competition is thin." If you're not sure, leave it off the first time you run a scaffolded CFP and add it next cycle if your funnel needs more shaping.

The outside-the-scaffold lane. Always include "Other / Outside the scaffold" as a bucket option. Make it explicit: "If your idea doesn't fit the buckets above and you think it's exceptional, submit here. Tell us why it should be on this stage even though we didn't ask for it." The framing matters. You're not punishing submissions to this lane. You're asking them to do extra work because they're asking you to expand your scope.

Most submissions to the outside lane will not be exceptional. A small number will be the best submissions you receive. Treat the lane like a pull request review: high bar, but read every one carefully. The cost of accepting one bad outside-lane talk is one bad slot. The cost of rejecting one great one is the talk that would have made the conference memorable.

How this works in Gythr

Everything above runs in any tool. The reason it's easier in Gythr is that the scaffolded CFP isn't a workaround. It's how the module is built.

When you set up a CFP in Gythr, you define your buckets in the CFP settings. Submitters see them on the public submission form. Reviewers see them in the kanban. Accepted talks inherit the bucket so the program shape is visible at every stage of the workflow: submission, review, scheduling, post-event reporting. The bucket isn't metadata you add later. It's the spine of the CFP.

The AI in the module exists for the same reason the rest of the module exists. To remove friction, not to perform. Five things it does today, all in the background:

• Surfaces near-duplicate submissions so reviewers stop reading the same talk twice
• Generates topic prompts from your conference theme so you have a starting point when scaffolding the buckets
• Pulls quick facts from each abstract so reviewers can get the gist without re-reading the long form
• Drafts preliminary placement in the agenda once a talk is accepted
• Auto-tags submissions so the kanban stays browsable as the volume grows

None of these are "talk to AI about your CFP." None of them write the conference for you. They sit underneath the work, removing the specific frictions that make CFPs painful at scale. We don't add AI to a module unless we can name the friction it removes. If we can't name it, we don't ship it.

The next thing we're building is the V2 of this method. AI-assisted topic scaffolding. You brief Gythr on your event (audience, scope, last year's gaps, what your sponsors are funding, where you want to broaden) and Gythr drafts the buckets for your team to refine. Same scaffolded CFP method as this post describes, less of the three-to-five-hour offsite. Rolling out Q3 2026. If you want early access, email me.

One more thing

The biggest thing the scaffolded CFP changes isn't the lineup. It's the posture you have toward the talent pool. An unshaped CFP says "tell us what you have." A shaped CFP says "we know what we need, show us how you'd contribute." The first is passive. The second is the actual work.

Try it on the next CFP you run. Even if you do nothing else differently. Even if you skip the AI and the crowding visibility and just internally write the buckets before opening, the lineup you get back will be measurably different. We've watched it happen. Try it once. Tell us how it goes.