Presentation Strategy by Venue: What to Show, and What to Hold Back

The two earlier posts in this series were about craft. The first covered what I look for when I watch a talk, and the second covered how to build one, slide by slide. Both of them assume a question has already been answered, and I have noticed that most students never actually stop to answer it: who is in the room, and what do they have the right to walk out with?

That question changes the deck more than any design choice will. The same project, presented well, looks genuinely different in a Friday seminar than it does in a fifteen-minute conference session — and the two conference formats, full paper and extended abstract, are not the same talk either. Students get this wrong in both directions. They bring a stiff, over-formal conference talk to a seminar where nobody wanted one, and then they bring a seminar’s openness to a conference room full of people working on the same problem they are.

So this post is about calibration. Three venues, three strategies, and one rule underneath all of them: what you show should be what that room has earned, and nothing past it.

The school seminar: think out loud, mostly about other people’s work

A lab meeting or a departmental seminar is, in the main, a room of fellow students and a few faculty who know your area only roughly. Nobody in that room is evaluating whether your contribution is publishable. They are there to learn something and to help you think.

That has a concrete consequence: the center of gravity of a seminar talk is the literature review, not your results. You are the person in the room who has read the most about this topic in the last three months. The most valuable thing you can do is give everyone else that map — what the field is arguing about, which strands have converged, where they stop. Spend the bulk of the time there, and spend it synthesizing rather than listing (the failure mode described in the first post is at its most tempting in exactly this format).

Your own work belongs in the talk, but in proportion: a sketch of the direction you are taking, some preliminary progress, one or two things that are not working yet. That last category is the one students skip, and it is the one the format is actually best at. A seminar is the cheapest place you will ever get to say “here is where I am stuck” to ten people who might know the answer. A polished talk that hides the stuck parts wastes the only real advantage the venue has.

What this means practically:

• Do not dress it up as a conference talk. No formal contributions slide, no results section engineered to look conclusive, no defensive framing. The register is closer to a well-organized discussion than a performance.
• Preliminary is fine, and say that it is. Label it. “This is one pilot participant, n equals one, so read it as direction, not evidence” costs you nothing in that room and buys you honesty you will need later.
• Leave time and room for the discussion. If nobody interrupts you, the talk was probably too closed.
• Open questions are content, not weakness. Ending on three things you have not figured out is a perfectly good seminar ending.

The one thing I will push back on is the opposite failure — treating “it’s just a seminar” as permission not to prepare. The register is informal; the thinking is not. Sloppy claims are still sloppy in a small room, and this is the room where your colleagues learn how carefully you think.

Conference, full paper: present the paper, and nothing above it

If your paper has been accepted, peer-reviewed, and is going into the proceedings, your job on stage is narrow and clear. Present what is in the paper. The claims you defend on stage should be the claims that survived review, stated at the same strength they are stated in the text.

The temptation here is to oversell — to open with a sweeping vision, to promise more generality than the study supports, to let “we demonstrate a framework for X” quietly become “we have solved X.” I want to name where that instinct comes from, because it is not random: that is grant-proposal rhetoric, and it is the correct rhetoric in a proposal. A proposal is an argument about a future that has not happened yet, and it is supposed to be ambitious. A paper presentation is a report on something that already happened, and the standard is accuracy, not ambition. Borrowing the proposal voice for a paper talk reads, to the people in the room who have done this work, as either inexperience or spin.

The audience at a conference session is the audience most able to check you. Some of them reviewed papers in your track. Some of them have run the experiment you are describing. If you inflate a claim, someone in that room usually knows, and the first question in Q&A will find the seam. Meanwhile, the cost of accuracy is close to zero: nobody has ever thought less of a talk for stating a limitation plainly.

So, concretely:

• Stick to the paper’s scope. Methods as run, results as measured, limitations as written.
• Do not add unpublished results on the fly because they make the story rounder. If it did not go through review, it does not get presented as though it did.
• State limitations before Q&A does. Naming the weak point yourself converts an attack into a discussion.
• Future work is one slide, framed as a direction, not a promise. “We are extending this to X” is fine. A detailed roadmap of unpublished plans is not — see the next section for why.

The paper talk is the one venue where you are, in a narrow sense, safe: the work is public, timestamped, and citable. Use that safety. Be precise, be concrete, and let the work be the size that it actually is.

Conference, extended abstract or presentation-only: be deliberate about what leaves the room

This is the format where I want you to slow down, and it is the one students handle most carelessly, because it feels like the low-stakes version. It is not. An extended abstract or presentation-only slot typically means: the work is early, the full paper is not written or not accepted, and nothing you present is protected by a publication record. You are standing in front of an audience of people working on adjacent problems, showing them what you are doing before you have planted the flag.

That does not mean be paranoid, and it does not mean be vague — a talk that says nothing wastes everyone’s afternoon, including yours. It means be deliberate. Decide, before you build the deck, which cards are on the table and which are not.

My working split:

Show	Hold back
The motivation and the problem — that framing is your contribution to the conversation	Your full research plan, stage by stage
The gap, stated clearly	Every unexecuted experiment, with conditions and measures spelled out
The results you actually have, however preliminary	Specific untested hypotheses you have not begun to test
Enough method to make the results interpretable	Novel implementation details that are the real technical advance
One general direction for where this is heading	A dated roadmap of what you will do next and in what order

The line I use: show what you have done; be sparing about what you are going to do. Completed work, even small, is yours — you did it, you can talk about it, and presenting it establishes that you were there first. Unexecuted plans are a different thing entirely. A plan is cheap to copy and fast to execute for a group with more people and more equipment than you have. Handing over a complete experimental design, at a venue with no proceedings to timestamp your priority, is the one genuinely avoidable risk in this whole series of posts.

A few specifics that come up often:

• “Future work” on a preliminary talk should get vaguer, not more detailed. This is the opposite of what students instinctively do. When the work is early, a future-work slide with five bulleted, fully specified studies on it is a gift to the room. One sentence of direction is enough.
• A framework or architecture diagram is usually safe; the unpublished implementation detail inside it is not. Show the boxes. You do not owe the room the contents of every box.
• Distinguish what you are still deciding from what you have decided. If you are comparing three possible measures and have not picked one, that is internal deliberation, not presentation content.
• Q&A is where this leaks. You can rehearse a careful deck and then give the whole plan away in ninety seconds because someone asked a friendly question and you enjoyed answering. “That’s the next study, and I’d rather not get ahead of the data” is a completely normal, completely professional thing to say at a conference. Practice saying it.
• If someone is genuinely interested, take it offline. Collaboration is good. It should just start with a conversation you chose to have, not with a slide you left up for four hundred people.

None of this is about secrecy for its own sake. It is about sequencing. Present the finished thing; protect the unfinished thing until it is finished too.

A quick calibration table

	School seminar	Conference, full paper	Conference, extended abstract
Audience	Peers, mostly outside your niche	Specialists, some reviewed your track	Specialists, working nearby
Center of gravity	Literature synthesis	Method and results, as published	The results you actually have
Your own work	A sketch plus preliminary progress	Full, exactly at paper scope	Selective, completed work only
Register	Thinking out loud	Precise and factual	Deliberate and controlled
Claim strength	Hedged, clearly labeled preliminary	Exactly as written in the paper	Conservative, under-claim rather than over
Future work	Open questions, invite discussion	One slide, direction not promise	Deliberately general
Main risk	Over-formalizing; not preparing	Overselling past the evidence	Giving away unexecuted plans

The question underneath all three

Every one of these venues is answering the same question differently: what does this specific room have the right to walk out with?

A seminar room has earned your map of the literature and your honest confusion, because it is there to help you. A full-paper session has earned every claim in your paper, exactly as reviewed, because the work is public and the audience is qualified to check it. An extended-abstract session has earned your framing and your finished results, and has earned nothing at all about the study you have not run yet.

Get that question right and the deck mostly builds itself. Get it wrong and no amount of the craft in the first two posts will save the talk — you will be presenting an excellent version of the wrong thing.