CrunchBase on Neo4j

NeoTechnology was featured on TechCrunch after raising a Series B round, and it has an entry on CrunchBase. If you look at CrunchBase closely you’ll notice it’s a graph. Who invested in what, who co-invested, what are the common investment themes between investors, how are companies connected by board members, etc. These are questions we can ask of the graph and are well suited for graph databases.

So let’s get to it. This project and dataset are available on Github, but if you want to play it on your own you’ll have to register on and get an API key. Once you have one you can take a look at the JSON returned by heading over to the IO Docs. Here is part of the “neo-technology” company record.

"name": "Neo Technology",
"permalink": "neo-technology",
"crunchbase_url": "",
"homepage_url": "",
"blog_url": "",
"blog_feed_url": "",
"twitter_username": "",
"category_code": "software",
"number_of_employees": null,
"founded_year": 2007,
"founded_month": null,
"founded_day": null,
"deadpooled_year": null,
"deadpooled_month": null,
"deadpooled_day": null,
"deadpooled_url": null,
"tag_list": "graphs, open-soure-graphs, graph-systems, commercial-graphs, graph-database",
"alias_list": "",
"email_address": "",
"phone_number": "",
"description": "Graph Database",

I’m not going to map all of CrunchBase, but I’ll grab a few things of interest. I’ll import companies, people, financial organizations, and tags as Nodes and employee, investor, tagged, and competitor connections as Relationships into Neo4j.

I am using the wonderful CrunchBase gem by Tyler Cunnion. I start by getting all the entities I am interested in (for example here is People):

 all_people = Crunchbase::Person.all
  all_people.each do |ap|
      next unless ap.permalink
      file = "crunchbase/people/#{ap.permalink}"
      if File.exist?(file)
        person = Marshal::load(, 'r'))
        person = ap.entity, 'wb') { |fp| fp.write(Marshal::dump(person)) }      

      people << {:name => "#{person.first_name || ""} #{person.last_name || ""}",
                 :permalink      => person.permalink,
                 :crunchbase_url => person.crunchbase_url || "",
                 :homepage_url   => person.homepage_url || "",
                 :overview       => person.overview || ""

    rescue Exception => e   
      puts e.message


I am saving the JSON returned from the CrunchBase API to a file in case something goes wrong and I need to restart my import. Neo4j doesn’t like null values for properties, so just in case I use an empty string if no value is found.

I will be using a fulltext lucene index to allow people to find their favorite company, person, tag, or financial organization by permalink, so I create that first.

neo =
neo.create_node_index("node_index", "fulltext", "lucene") 

Then I will be creating the nodes for these entities (while saving the node id returned in a hash):

  people_nodes = {}
  people.each_slice(100) do |slice|
    commands = []
    slice.each_with_index do |person, index|
      commands << [:create_unique_node, 
    batch_results = neo.batch *commands

    batch_results.each do |result|
      people_nodes[result["body"]["data"]["permalink"]] = result["body"]["self"].split('/').last

I am using the Neo4j Rest Batch method sending 100 commands at a time.

I then create the relationships to each other. For example employees of companies:

  employees.each_slice(100) do |slice|
    commands = []
    slice.each do |employee|
      commands << [:create_relationship, 
    batch_results = neo.batch *commands  

For the front end, I need two basic methods. One will return all the nodes and relationships types connected to one node:

cypher = "START me=node(#{params[:id]}) 
          MATCH me -[r?]- related
          RETURN me, r, related"

connections = neo.execute_query(cypher)["data"]

A second query will handle full text search and format the output to a JSON hash that JQuery Autocomplete can work with:

  get '/search' do 
    content_type :json
    neo =    

    cypher = "START me=node:node_index({query}) 
              RETURN ID(me),
              ORDER BY
              LIMIT 15"
    query = "permalink:*#{params[:term]}* OR name:*#{params[:term]}*"
                      {:query => query })["data"].
                             { label: x[1], 
                               value: x[0] }

Notice how I am passing in the whole lucene query as a parameter? This is the right way to do it. I indexed people, companies, and financial organizations by permalink, but indexed tags by name. I use an “OR” clause to grab them both.

You can skip the graph building step if you just want to use the data as it existed the day of this blog post since it is available on the github repository.

You can clone the repository, create a heroku app, and get this up and running in no time.

git clone
cd neo_crunch
heroku apps:create
heroku addons:add neo4j:try
git push heroku master

Then go to your Heroku apps

Find your app, and go to your Neo4j add-on.

In the Back-up and Restore section click “choose file”, find the file and click submit. It will take a little while to upload the data, but once you will be able to navigate to your application and see this:

Click the image above to see a live version of it.

Tagged , , , , , , ,

6 thoughts on “CrunchBase on Neo4j

  1. Are the neo4j databases really hosted directly on the internet using HTTP and basic authentication?

    • maxdemarzi says:

      Hi Willem,

      The Heroku Neo4j:Try plans are not secure. The production plans on Heroku will have SSL and other security measures in place.


  2. Luc Bartkowski says:

    Hi Max,

    Interesting story, thanks. I am using since 2009 as my personal “graph database” to support my business. It contains almost 10K objects like companies, news items, etc. But unfortunately TheBrain doesn’t support normalized data like telephone numbers and email addresses as attributes that I can easily export to other, CRM-like, applications. And it doesn’t support an API to mashup with for example social media. So I am considering Neo4j as an replacement of my PersonalBrain.

    I have to convert my TheBrain to Neo4j so I have to develop import scripts using Cypher. But at this stage I would like to have a personal Neo4j Graph Visualizer in which I can CRUD objects and their relationships and attributes within a GUI preferably without programming in an on-premise stand-alone environment. Which Neo4j Visualizer do you recommend for this purpose?

    Best regards,

  3. SimonKwok says:

    Thank you Max. Your great job help me out of my performance problem.
    After I change my cypher FROM “START n = node(*) WHERE has(n.ID) AND n.ID=~ ‘.*000000.*’ return n” TO “START n = node:node_auto_index(‘ID:*000000*’) return n”, the time is reduced to 874ms from 41234ms in the case of 1000M nodes.
    BTW , is there any more space for performance tuning in cypher syntax ? Or any article about it? Thanks again.

  4. […] those who are curious to understand the business graph of Crunchbase, I recommend this tutorial by Max de Marzi. It show how to use Crunchbase API to populate a Neo4j database and then visualize […]

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: