Smarter Nearby Filtering: Using Midpoints for Multi-Location Posts

When I built the Nearby feature in my side project Rekkoku, I ran into a small but critical design challenge.

Each post in Rekkoku can contain up to 7 places from Google Maps.

So the goal was simple:

"From the user's current location, show posts that are geographically closest."

Seems easy until you realize:

Which location should I compare the user's position to?

﻿

The Problem

If I calculated the distance from the user's location to every single place in every post that would be:

•Too heavy on performance

•Difficult to scale

•Messy to query

Even worse, each post contains multiple locations.

Should I compare all 7? Just the first one? The average?

﻿

The Insight: Use a Midpoint

I needed one clean coordinate per post, a single point to represent all 7 locations in that post.

So I took a step back and thought:

"What if I just calculate the midpoint?"

Like in high school geometry find the average of all the latitudes and longitudes.

That becomes the center point of the post.

﻿

Implementation Flow

Here's how I did it:

1.When a post is created, I extract the latitude and longitude from each Google Maps link.

2.I calculate the average of all lat/lng coordinates -> that becomes the midpoint.

3.I store that as center_lat and center_lng in the post table.

4.The Nearby feature uses this coordinate to calculate distance from the user's current location using the Haversine formula.

﻿

Why It Works

This small change made a big impact:

•Nearby queries became lightweight

•I only needed to compare one point per post

•It's scalable even if I have thousands of posts

•It still feels accurate enough for the user experience

It's not GPS precision but it's UX-accurate

﻿

Final Thoughts

Even in a side project, decisions like these matter.

Sometimes, solving a user-friendly feature requires thinking deeper about data structure and query design.

If you're building a "Nearby" feature and dealing with multi-location records, ask yourself:

•Should I use a midpoint?

•Or a bounding box?

•Maybe even geohashing?

Every option has trade-offs. In my case, midpoint + Haversine worked great.

﻿

Built with:

Next.js, Express.js, and curiosity.

﻿

Let me know how you'd approach this, I'd love to hear other tricks!