Personalized Event Experiences Using AI

What personalization is, and is not

Personalization is relevance, not flattery. Using a first name in an email is the smallest form. The useful form is showing someone the three sessions out of forty that fit what they told you they came for.

It is also not magic. A recommendation can only be as good as the information behind it. If the attendee told you nothing, the best any system can do is suggest what is popular.

• Rules-based

  You decide: attendees who ticked 'finance' see the finance track first. Simple, predictable and easy to explain.

• AI-based

  A model weighs several signals, such as stated interests and past behavior, and ranks suggestions. It can notice patterns you did not write a rule for.

• Self-service

  The attendee builds their own view, for example a personal agenda. Not AI, and often the most trusted version.

  Personalized agenda in the Eventdex app

The data personalization runs on

Collect only what you will use. Each question at registration costs a little goodwill and a little completion, so ask for what changes the experience.

What to tailor, in order of payoff

Start where an attendee's time is most scarce: choosing among sessions and choosing whom to meet.

1. 1

   Sessions

   Suggest sessions from stated interests and role. For multi-track events this is the biggest benefit, because attendees cannot attend everything.

   Conference agenda guide
2. 2

   People

   Match on goals and interests and tell each person why the match was suggested, so they trust it.

   What is event matchmaking
3. 3

   Exhibitors and sponsors

   Point attendees to the booths that fit their interests, which also gives sponsors visitors who care.

4. 4

   Messages

   Send reminders about what the attendee chose, not a general broadcast to everyone.

5. 5

   Answers

   A chat assistant that knows the attendee's role can answer with what applies to them, such as the exhibitor's setup times or the speaker's room.

   AI event chat assistant

Asking for data fairly

People share more when they understand what they get in return, and when the request is plain. Put the explanation next to the question, not in a policy nobody reads.

Before we collect personalization data for [event name]:

[ ] Every question has a use we can name in one sentence.
[ ] The form says why we ask: "We use this to suggest sessions and people."
[ ] Optional questions are marked optional.
[ ] We say who can see the answers (organizers, sponsors, other attendees).
[ ] We say whether a third-party service processes the data.
[ ] Attendees can change or remove their answers.
[ ] We do not use sensitive details for matching unless the attendee agrees.
[ ] Our privacy policy is linked from the form.
[ ] A named person owns questions and requests about data: [name]

What Eventdex describes

These are the personalization features Eventdex's pages describe. They are reported as written, and the last item says what is missing.

• Smart attendee recommendations

  The AI event platform page says recommendations use session history, preferences and networking behavior to suggest content, exhibitors and connections.

  AI event platform

• AI business matchmaking

  Attendees and exhibitors are paired using profiles, industries, interests and event objectives.

  Business matchmaking app

• Registration paths and a personal agenda

  The registration software page describes personalized registration paths and session selection, and the mobile app page describes a personalized agenda that guests create from their preferences.

  Event registration software

• What the pages do not describe

  They do not describe how recommendations are scored, how a model is trained, or per-attendee personalized email content. Ask in a demo and get the answer in writing.

Where personalization goes wrong

Most failures are about trust and relevance, not technology.

• Too little data

  Recommendations from an empty profile are generic. Ask two or three good questions at registration rather than none.

• Creepy specifics

  Showing people that you know how they moved around the venue, without telling them, damages trust. Be open about what you use.

• A filter bubble

  If everyone only sees what matches their past, nobody finds the unexpected session. Leave room for browsing.

• No explanation

  'Suggested for you because you chose finance' is trusted more than an unexplained list.

How to tell whether it worked

Compare against your own last event or against a group that did not get the feature. Pick measures before the event, and do not claim a result you did not measure.

• Use of the suggestions

  How many attendees saved a suggested session or accepted a suggested meeting.

• Attendance spread

  Whether session attendance is better matched to room size than last time.

  Event attendance tracking

• Attendee comments

  Ask directly: 'Did the suggestions help you choose?' in the post-event survey.

  Post-event survey guide

• Other measures

  Pick from the event KPIs list the ones your goal depends on.

  Event KPIs