Nike has been an offshore manufacturing leader in that they published the locations of their factories. Their online maps proved uninformative at any scale since factory locations were tightly clumped often within blocks of each other in particular cities.
Uploaded image
My first suggestion was to offer instead regionally aggregated numbers using a facetted query mechanism. This they published online as their 2011 interactive Manufacturing Map 
.
I remained interested in the clustered data issue since I had faced it in my own rendering of radio relay station locations which I solved with the abstract Compass Rose rendering.
I had already chosen to feature the d3.js library for visualization within federated wiki. I set out to combine its map projection logic with the force-relaxation layout logic my making factory location just an attractor that datums could approach.
This approach preserves one-to-one mark-datum mapping while producing roughly circular clusters when actual map locations are in close proximity.
This has several advantages over clustering nearby datum into a single composite dot.
• color and area remain available for coding additional parameters
• non-circular clusters result when two clusters are of medium proximity
• individual marks can be added or removed as datasets evolve over time.
See github for plugin source.