When attendees complain about matchmaking, they usually blame the algorithm. "It suggested my competitor." "Everyone on my list is a student." More often than not, the problem was set weeks earlier, in an onboarding form someone put together in an afternoon.
Matching can only work with what people tell it. This guide covers the two inputs that matter most, the interest list and the onboarding questions, and how to test them before a single attendee sees them.
Two inputs, two jobs
It helps to separate what you ask into two buckets.
Interests and goals drive the match. They come from a fixed list, so one person's answers can be compared directly with another's. Two people either both picked "Payments" or they didn't.
Onboarding questions shape the profile and tell you, the organizer, who is in the room. They can be open text, single choice or multiple choice. Some are written for other attendees to read. Some are only for you.
Mixing the two is the most common mistake. A free-text question such as "What are your interests?" produces answers that can't be compared reliably. An interest list with 80 entries turns a matching input into a survey nobody finishes.
Build a short interest list
An interest only helps matching if it splits the room. "Innovation" at a technology conference does not. Nearly everyone will tick it, so it tells the system nothing about who should meet whom.
Long lists create a second problem. If people pick twelve interests each, everyone shares a few with everyone else and every pair looks roughly as good as every other. Sensible scoring stops rewarding overlap after a point for exactly this reason. In Conecta, shared interests are worth up to 40 points and the score stops rising at five shared interests, so extra overlap beyond that changes nothing.
Practical guidance:
- Aim for something like 15 to 30 interests in total, grouped into four to six categories. A smaller event can manage with fewer.
- Write each interest so an attendee can answer "is this me?" in two seconds.
- Use your audience's words, not your marketing team's. "Hiring backend engineers" beats "Talent".
- Cut any interest you expect most of the room to select.
A before and after
Say you are running a fintech conference. A first draft might read: Innovation, Digital transformation, Fintech, Payments, Digital payments, Banking, AI, Machine learning, Blockchain, Crypto, Regulation, Compliance, Startups, Investment, Networking, Partnerships.
Look at it closely. "Fintech" and "Innovation" describe the whole event. "Payments" and "Digital payments" are the same thing. "AI" and "Machine learning" overlap heavily. "Networking" is a goal, not a topic, and "Investment" is probably a goal too.
A tighter version, by category:
| Category | Interests | Goals allowed |
|---|---|---|
| Sector | Retail banking, Payments, Lending, Insurance, Wealth management | Networking, Trading, Investing |
| Function | Product, Engineering, Risk and compliance, Sales and partnerships | Networking, Recruiting, Mentoring |
| Technology | Fraud detection, Open banking APIs, AI in credit decisions, Digital identity | Networking, Trading |
| Company stage | Early-stage startup, Growth-stage company, Bank or corporate | Networking, Investing, Mentoring |
That is 16 interests across four categories, and each one divides the room into people it describes and people it doesn't. The goals column is covered further down.
Keep categories mutually exclusive
Each category should answer a different question. "Which sector are you in?" and "What is your job?" are different questions. "Technology" and "Trends" are not, which is how "AI" ends up listed twice.
Rules of thumb:
- Every interest belongs to exactly one category.
- Each category looks at people from its own angle: sector, function, technology, company stage, region.
- If you can't state the question a category answers, merge it into another one.
Clean categories make the list quicker to scan, and they make a shared interest mean something. Two people who share a sector and a technology have a specific reason to talk. Two people who both ticked "Innovation" under "Trends" don't.
Topics are not goals
The interest list says what someone cares about. It doesn't say what they want. A founder and an investor might both pick "Lending", one looking for money and the other looking for deals. Two founders might pick the same thing and be chasing the same investors.
That is why goals need their own input, attached to each interest. Attendees choose an interest, then say what they want from it: Networking, Trading, Recruiting, Investing or Mentoring. The organizer decides which goals each interest allows. In the table above, "Early-stage startup" allows Investing and Mentoring alongside Networking but not Trading, while the technology interests only allow Networking and Trading.
Restricting goals per interest does two things. It stops combinations that make no sense, such as Recruiting on "Insurance" as a sector, and it makes the choice faster because people see two or three options instead of five. The reasoning behind putting goals first is in why goals matter more than shared interests, and the matchmaking features page shows how organizers edit categories, interests and allowed goals.
Decide what is required
Every required question costs you some completions. Every optional one costs the attendee nothing, and many people skip it. Decide question by question:
- Required: the inputs matching depends on, and the question that writes the profile summary.
- Optional: anything you want only for your own reporting, and anything personal.
- Removed: anything you already collected at registration, and anything nobody will read.
A sensible limit is three to five onboarding questions on top of interests. If you need more data for sponsors or reporting, collect it at registration, where people expect forms.
Choose the question type for how you will use the answer. Short answer for the profile summary. Multiple choice when you want one answer you can group by, such as company size. Checkboxes when several answers are valid, such as the regions someone covers.
Write the About question carefully
In Conecta, the first short-answer question becomes the "About" text on each profile. Whatever you ask there is what every other attendee reads when deciding whether to meet someone, so it deserves more thought than any other field.
Generic prompts produce generic answers. "Tell us about yourself" tends to return a job title and a line of company copy. A prompt like "What are you working on right now, and who would be useful for you to meet here?" asks for something another attendee can act on.
Compare two answers to that second prompt:
Head of Partnerships at a payments company. Passionate about innovation.
Building card issuing for small lenders in Central Europe. Looking for banks open to a pilot and an engineer who has been through a PCI audit.
The second person will get better meetings, and the prompt is what made that answer more likely. Keep it under 20 words, make it required, and if your form allows a placeholder, use one that shows the kind of answer you want.
Test with ten fake profiles
Before you open onboarding, run a dry test. It takes about an hour and catches most problems while they are still cheap to fix.
- Write ten personas that represent your real audience. For a startup event: a seed investor, a later-stage investor, a first-time founder, a founder raising a Series A, a corporate innovation manager, a recruiter, a senior engineer, a mentor, a student and a service provider.
- Fill in onboarding as each one, choosing interests and goals the way that person realistically would. Make one of them lazy: required questions only, two interests.
- Read each persona's recommendations and ask whether that person would want to meet the top three.
- Look for the edge cases. Does anyone have an empty list? Does everyone see everyone? Does the seed investor see the other investor at the top?
- Fix and repeat. Change the interest list, the allowed goals or the questions, then run the same ten personas again.
Ten profiles will give thin results, and that is useful. If your tool hides weak matches, a persona with an empty list is telling you the interests are too fine-grained or the allowed goals too narrow. If every persona sees the same names in the same order, the list is too broad.
Make sure the test profiles are removed or hidden before real attendees arrive. A fake investor at the top of a founder's list on launch day costs you trust you won't get back. Once the setup holds up, the how it works page shows how it carries through to what attendees see.
Before you open onboarding
- No duplicate interests, and none that most of the room would pick.
- Each category answers a different question.
- Each interest allows only the goals that make sense for it.
- Required questions are limited to what matching and profiles need.
- The About question asks what people are working on and who they want to meet.
- Ten test personas get recommendations you would be happy to defend.
In short
Good matchmaking starts with a short list of interests that split the room, sorted into categories that don't overlap, with goals attached to each one. Keep the questions few and make the About prompt do real work. Then test the whole thing with ten honest personas before anyone else sees it. If you are planning the wider networking programme as well, our guide to improving conference networking covers the agenda and the floor. The algorithm matters, but it can only rank what you gave it.