Graph Analytics and Clustering for marketing

Google “social graph“,  go to the image tab and you’ll see amazing visualization of social graphs, text graphs…

yet there are very few articles covering the practical use of graphs for marketing and sales.

At eCairn, we use two different types of graphs/maps for specific sales and marketing objectives. We also do text maps for custom projects.

  • Network map – to understand connections between people
  • Affinity maps – to understand affinities between people (or between groups of people, or between messages)

1.Network Map

The network map is the most obvious map. It represents relations in the social network i.e people following people, people RT people, blogger linking back to another blogger …

Here below is the map of the most influential people in Bend (Oregon):

 

Bend Influencer Map

We use Touchgraph for the visualization and analysis of our maps.

The main interest for the Network map is for:

  • Sales/ Referral:  Seeing that one of your friends is a connection to your prospect and can introduce you.
  • Influencer Marketing /Thought Leadership: As you map your target audience, you identify clusters and key leaders in that cluster (people who are followed the most, RT the most) and you can focus your programs toward these influencers.

 

Zooming in a far end corner of the Bend map, we can see a group of people who are passionate about biking along with a few influencers.

  • Competitive Analysis: By mapping your target market/audience and positioning your competitors (brands, sales rep) in the graph, you can understand how aggressive competitors are engaging your audience or your client base and how effective they are.
  • Segmentation: Works best for B2B or very targeted audiences, analyzing the network, you can spot clusters of people who talk to each other.

Here is an (old)  segmentation of the B2B tech community.

2.Affinity Map

Affinity maps are maps where people are grouped from the things they have in common. These are the co-occurrences map.

For example, if you have a large group of people and the list of items they purchased, you can build an affinity map to identify groups of people buying similar items.

At eCairn, we build affinity maps based on who people follow.

Here is an affinity map for the people who follow SlackAPI.

We clearly have four clusters (of different sizes). Looking at the green cluster, we see that these are people are Founders & Execs, not that much interested in Tech but more in the VC/Startup ecosystem. Here below is a detail of what they put in their bio, who they follow the most (comparing to the other clusters i.e everybody follow slackapi, donaldtrump and obama but this is not really useful to profile the cluster), and who they don’t follow who are followed the most by the other clusters.

Interpreting cluster is tricky and so far can’t really be automated.

The main use case for the Affinity map is for market understanding/discovery and persona definition. The client loads a list of people (clients, prospects, followers) and our algorithm will map and create clusters of people following the same “celebrities”.

  • Audience understanding / Segmentation: Works best for “random groups” of people of groups where the diversity in the group is high.

 

3.More applications for mapping

At eCairn, we also use mapping on “custom” projects, like in the example below:

  • Mapping “segments of customers” based on the topics/keywords they are using the most. Interesting when you need to align message to market segments.

In this example, each cluster (color) is a segment and each gray box is a Shopping Keyword they are using. Cluster 1 (i.e segment #1)  in this example is “Price Conscious Moms”.

  • Looking for affinity in a text corpus. As an example, we collect articles on different themes and see the sub-topic that they have in common. This is very interesting if you are trying to find the right keywords to use in a marketing message.

 

 

 

Leave a Reply

Your email address will not be published. Required fields are marked *