You already collect the data that should decide your speaker. Three layers: your past events, your registration fields and live market signal.
Link copied
You already collect the data that should decide your speaker. You just do not use it for that.
Every registration form asks for job title, company and sector. Every post-event survey asks what people thought of the sessions. Every year the same spreadsheet records who attended, who stayed to the end and who left after lunch. Then, when the time comes to book next year's keynote, the decision gets made from a shortlist someone liked the look of.
This is not carelessness. It is a gap in the workflow. The data sits in the registration platform, the survey lives in someone's inbox and the speaker conversation happens in a different meeting entirely. Nobody joins them up.
The Amex GBT 2026 Global Meetings & Events Forecast shows how far the collection side has moved: 40% of planners expect to deploy AI-powered apps delivering personalised schedules and smart networking suggestions, and 35% cite personalised communications and attendee matchmaking as their planned uses. Improving the attendee experience is the top priority for a second consecutive year.
So the industry is getting sophisticated about personalising the agenda. The choice of who stands on the stage is still, in most organisations, made on instinct.
The usual first move is to reach for a sector report. These are worth reading and they set the agenda properly. They just cannot set your calendar, for three reasons.
Publication lag. A corporate report is typically written on six to nine months of data. By the time it lands, the field has moved. The picture is accurate and it belongs to the past.
The averaging effect. A national finding flattens the difference between sectors. Adoption in financial services and adoption in retail dissolve into the same average. You are booking a speaker for your sector, not for the mean.
Format blindness. Reports produce subjects, not formats. Whether a theme lands as a keynote, a workshop or a programme running across the year is not a question a report answers.
The report tells you what is being discussed. Your own data tells you what your room needs.
Most organisations run a post-event survey and then read only the headline score. The useful material is underneath it.
Three things are worth pulling out:
Session attendance against the schedule. Which sessions held the room and which emptied. A session that lost a third of its audience in the first fifteen minutes is telling you something about level, not about the topic.
The questions people actually asked. Q&A transcripts and chat logs are the most honest data in the building. Nobody types a question to be polite. If four people asked variations of the same question, that is next year's session brief.
The gap between rating and recommendation. A speaker can score well on delivery and poorly on relevance. Those two numbers moving apart is the clearest signal that the content was good and aimed at the wrong room.
None of this requires new tooling. It requires reading what you already have with the speaker decision in mind.
Registration fields are collected for logistics and then never read again. They describe your audience better than any persona document.
Four fields carry most of the weight:
Field | What it decides |
Job title and seniority | The level the content should pitch at, and whether one level is enough |
Sector or business unit | Which examples will land and which will feel borrowed |
Years in role | Whether the audience needs orientation or challenge |
Session preferences at sign-up | Where genuine appetite sits, before anyone has been in the room |
The seniority spread matters more than the average. An audience running from graduate to board member is not a "mixed audience", it is two audiences sharing a room, and it usually needs either a different format or a different speaker level. Working out which is a decision you can make from data you already hold, weeks before you approach anyone.
This also connects directly to the brief. Everything in the table above belongs in the audience profile section of the speaker brief; most briefs leave that section thin because the data was never pulled together.
The third layer is the one you cannot generate internally: what is actually being booked right now, and how far ahead.
This matters for two practical reasons. First, availability. A theme that is being booked heavily will have its recognised names committed further out than your planning cycle assumes. Second, format. The same subject moves through the market in different shapes over time, starting as a keynote and later being requested as a workshop or a series, and that shift tells you how familiar your audience is likely to be with the material already.
This is a reasonable thing to ask an agency directly. Which themes are moving, at what lead time, and in what format. A bureau that cannot answer it is quoting from a catalogue rather than reading a market. That is one of the questions worth asking before you sign.
Each layer answers a different question, and the order matters.
Layer | Question it answers | Where it comes from |
Market report | What is the sector discussing? | Published research |
Your past events | What did this audience respond to? | Survey, attendance, Q&A |
Your registration data | Who is actually in the room? | Sign-up fields |
Live market signal | What is available, when, in what format? | Your agency |
Run them in that sequence and the shortlist tends to build itself. The report narrows the subject. Your event history rules out the levels that have not worked. Registration data fixes the pitch. Market signal tells you whether the name you want is realistically bookable in your window, or whether the format should change.
What this replaces is the conversation that starts with a name and works backwards to a justification. When that conversation happens, the budget usually goes up and the fit usually goes down, because the reasoning is running in reverse.
The final decision between shortlisted speakers still takes judgement. But judgement applied to evidence is a different exercise from judgement applied to a hunch.
1. Open your last post-event survey and read the free-text answers rather than the score. Mark every comment that mentions level, relevance or pace.
2. Export your registration data and count the seniority spread. If it runs across more than two levels, decide now which one is primary.
3. Pull the Q&A log or chat transcript from your last event. Group the questions. The largest group is a session brief.
4. Ask your agency which themes are moving in your sector and at what lead time. Book that conversation before you build the shortlist, not after.
→ Talk to us about what your event data is telling you
---
Cassie Kozyrkov, who pioneered decision intelligence as Google's first Chief Decision Scientist, makes the same argument one level up: the value of data sits in the decision it changes, not in the collecting. Watch the full talk.
Four sources, in order. Published sector research sets the subject. Your own post-event data shows what this audience has responded to before. Your registration fields describe who will actually be in the room. Live market signal from your agency tells you what is bookable, when and in what format. Most organisations use only the first and the last.
With the material you already have. Attendance figures against the schedule show which sessions held the room. Q&A logs and chat transcripts show what the audience actually wanted to know. Both exist for most events without anyone having designed a survey. Add two questions to your next survey: how relevant was this session to your role, and what did you expect that you did not get.
By making "mixed" specific. Count the seniority spread from your registration data rather than describing it. If the room runs from new joiners to directors, that is two audiences, and the choice is either a different format or a different speaker level. That decision can be made from sign-up data weeks before you approach anyone.
Before you build the shortlist. Reading your own data takes an afternoon; the market signal conversation takes half an hour. Doing both after a shortlist exists tends to produce justification rather than selection, because the names are already anchored.
Yes, for the subject. Published research is where you learn what your sector is discussing and why. What it cannot do is tell you the level, the format or the timing for your particular room, because it reports an average across a period that has already passed.
Which themes are moving in your sector, at what lead time, and in what format. Those three answers tell you whether the name you have in mind is realistically available in your window and whether your audience is likely to have encountered the material already. An agency that answers only with a list of names is reading a catalogue rather than a market.
After Dinner Speakers
Contents
