Articles & Case Studies
Semantic Contact Search vs Exact Title Match
In a DiscoLike test across 300 medical pain clinics, semantic contact search for "executives and senior decision makers" returned 745 contacts across 274 job titles. The 10 titles ChatGPT recommended would have matched only 89, missing 88% of the list.
One overlooked choice in contact search can leave you with at least 3x fewer relevant contacts: exact title matching or “contains” instead of semantic search.
Here is a real search example.
The Test: 10 ChatGPT Job Titles vs One Audience Description
A client defined the audience for an outreach campaign as “executives and senior decision makers.”
I asked ChatGPT for the top 10 job titles it would use to find that audience.
The list was reasonable:
CEO, Chief Executive Officer, Owner, Founder, President, Medical Director, Chief Medical Officer, Chief Operating Officer, Managing Partner, and Executive Director.
I took those 10 titles and checked what would happen if we used them as “title contains” searches.
Semantic Search Results: 745 Contacts Across 274 Titles
We ran semantic search across 300 companies in the medical pain clinics vertical using the client’s title description.
It returned 745 contacts across 274 unique job titles.
The title variation is incredible.
How Many Contacts Exact Title Matching Misses
Those 10 exact title match searches would have returned only 89 of them. So 656 contacts, 88% of the list, would have been missed.
Expand the list to 20 reasonable titles and you will still miss roughly three quarters of the contacts.
And a lot of the missed titles aren’t obscure.
Doctor, MD, Physician, Managing Director, Chief Financial Officer, Director of Business Operations, plus hundreds of other title variations.
In this case, searching by the description of who we wanted returned 8.4x more contacts than the 10 titles ChatGPT recommended.
Why Describing Your Audience Beats Listing Job Titles
I’ve been noticing this more as we work with real prospecting data. The way people describe who they want to reach is often much cleaner than the job titles those people actually use.
Why Most Contact Providers Can’t Do Semantic Matching
Budget unlimited contact providers, and even more expensive providers making high coverage claims, don’t have this kind of semantic matching.
It’s not easy to build. You need to vectorize titles, run similarity searches, burn GPUs, and maintain infrastructure that a lot of vibe coded products simply don’t have.
The data might already be there, or might even be cheap. The problem is being able to extract the right people from it.
Ping me if you want to see the raw data and independently confirm the numbers, or just try DiscoLike’s Persona Lookalike contact search.
Frequently Asked Questions
What is semantic contact search?
Semantic contact search finds people by matching a description of who you want to reach, such as “executives and senior decision makers,” instead of a fixed list of job titles. Titles are vectorized and matched by similarity, so variations like Physician, MD, or Managing Director are found without being listed.
How many contacts does exact title matching miss?
In a search across 300 medical pain clinics, 10 exact titles matched 89 of the 745 relevant contacts semantic search found, missing 88%. Expanding to 20 reasonable titles still misses roughly three quarters.
Why not just add more job titles to the search?
Because real titles vary far more than any list. The 745 contacts in this test held 274 unique job titles, and many of the missed ones were common titles like Doctor, Physician, and Chief Financial Officer, not obscure variants.
Do most contact data providers support semantic title matching?
Most do not. Some approximate it with an aliasing dictionary that maps known title variants, which still misses a lot. True semantic matching requires vectorizing titles, running similarity search on GPUs, and maintaining that infrastructure.
Related posts:
