Venture Capital (Data)

From edegan.com
Jump to navigation Jump to search


McNair Project
Venture Capital (Data)
Project logo 02.png
Project Information
Project Title Venture Capital (Data)
Owner Adrian Smart, Jake Silberman
Start Date
Deadline
Keywords Data
Primary Billing
Notes
Has project status Active
Copyright © 2016 edegan.com. All Rights Reserved.


Project Objective

To create a master venture capital data set on which all other projects requiring venture capital data will be built.

Specifically, we will be building data to support the following:

  • Lists and counts by year as well as state, MSA, and city, zip code, etc.
  • Identify seed, early and later stage deals (first investment), dollars invested and rounds
  • Identify transactional VC dollars invested and rounds
  • Acquisitions and IPOs of VC backed firms
  • Identify when VC backed firms are "alive"
  • Geocode all portfolio companies
  • Geocode VC headquarters and branch offices
  • Compute distances between investors and their portfolio companies
  • Identify lead investors for portfolio companies
  • Identify top 100 VCs

Retrieving data

Old notes are Old Venture Capital Data Work

Retrieving US VC Data From SDC requires SDC platinum, perl scripts and SQL scripts, as well as some manual processing.

Scripts and other info

SDC and Perl Scripts are in:

E:\McNair\Projects\VC Database

This includes:

  • NormalizeFixedWidth.pl
  • RoundOnOneLine.pl - Needed for round on one line

SQL Scripts and finished data are in:

Z:\VentureCapitalData\SDCVCData

Notes:

  • Portfolio companies attributes to be extracted from roundlevel information
  • Round-on-one-line processed using RoundOnOneLine.pl and NormalizeFixedWidth.pl
  • Firms includes branch office so attributes must be extracted
  • Portfolio company descriptions - just the portco name, state, date of first inv, and the long description - has to be custom processed.

The next steps are detailed in VC Database Rebuild.


Additional datafiles (in E:\McNair\Projects\VC Database):

  • GeocodedVCData.txt 43,724 records, tab-delimited with companynames but with "none" for some geocoords.

Processing the base tables

The SQL scripts need to do the following:

  • Extract company infomation from roundbase then process it
    • Add variable for growth (1), transactional (2), or non-VC (3) (see below)
    • Create clean 5-digit ZIP field
  • Extract firm information from firmbase, then:
    • Add indicator variable for US
    • Create clean 5-digit ZIP field
  • Build out BranchOffice (BO) table
  • Build lookups for:
    • State Code
    • raisestatus

SQL file to process data is:

Z:/VentureCapitalData/SDCVC/ProcessData.sql

Lookup Tables

Stage of Investment

We are generating 6 binary indicators as follows:

Growth VC (4)
--------------
Seed (1)
Early Stage (2)
Later Stage (3)
Tranactional VC (5)
--------------
Acq. for Expansion
Acquisition
Bridge Loan
Expansion
Pending Acq
Recap or Turnaround
Exclude (non-VC) (6)
--------------
LBO
MBO
Open Market Purchase
PIPE
Secondary Buyout
VC Partnership
Other

Building companybase and round

Both are built from roundbase which contains portfolio companies and round details. The scripts are below.

CREATE TABLE companybase AS
SELECT DISTINCT 
coname,updateddate,foundingdate,datelastinv,datefirstinv,investedk,city,description,msa,msacode,nationcode,
statecode,addr1,addr2,indclass,indsubgroup3,indminor,url,zip 
FROM roundbase
ORDER BY coname;
CREATE TABLE round AS
SELECT DISTINCT coname,statecode,datefirstinv,rounddate,stage1,stage3,rndamtdisck,rndamtestk,roundnum,numinvestors
FROM roundbase
ORDER BY coname;

Building stageflags

Stageflags are built from the stage3 column of round. Using the tables above, I made 6 flags for seed, early, later, growth, transaction and exclude. The script is below. Verify that no row entries are missing flags in excel by copying the round table to a txt file. The last 3 columns should always sum to 1 for every row entry. There should be no zero sums and no sums greater than 1.

CREATE TABLE stageflags AS
SELECT coname, stage3,
CASE
 WHEN stage3 = 'Seed' THEN 1
 ELSE 0
 END AS seedflag,
CASE
 WHEN stage3 = 'Early Stage' THEN 1
 ELSE 0
 END AS earlyflag,
CASE
 WHEN stage3 = 'Later Stage' THEN 1
 ELSE 0
 END AS laterflag,
CASE
 WHEN stage3 = 'Seed' OR stage3 = 'Later Stage' OR stage3 = 'Early Stage' THEN 1
 ELSE 0
 END AS growthflag,
CASE
 WHEN stage3 = 'Acq. for Expansion' OR stage3 = 'Acquisition' OR stage3 = 'Bridge Loan' OR stage3 = 'Expansion' OR stage3 = 'Pending Acq' OR stage3 = 'Recap or Turnaround' OR stage3 = 'Mezzanine' THEN 1
 ELSE 0
 END AS transactionflag,
CASE
 WHEN stage3 = 'LBO' OR stage3 = 'MBO' OR stage3 = 'Open Market Purchase' OR stage3 = 'PIPE' OR stage3 = 'Secondary Buyout' 
OR stage3 = 'Other' OR stage3 = 'VC Partnership' OR stage3 = 'Secondary Purchase' THEN 1
 ELSE 0
 END AS excludeflag
FROM round;

Sample output

The psql file was copied to an output text file using:

\COPY stageflags TO 'flag.txt' WITH DELIMITER AS E'\t' HEADER NULL AS  CSV; 

A sum column was added in excel which is the sum of the last 3 columns. Notice how all entries are 1. This is good.

Checking companybase table

The following psql code creates a new table called companybase1 that checks for Undisclosed companies and nonUS companies in the companybase table and also inserts a primary key column which is a concatenation of coname, statecode and datefirstinv.

CREATE TABLE companybase1 AS 
SELECT *,
concat(coname::text, statecode::text, datefirstinv::text) AS pkey,
CASE
 WHEN nationcode = 'US' THEN 1::int
 ELSE 0::int
 END AS alwaysusflag,
CASE
 WHEN coname = 'Undisclosed Company' THEN 1::int
 ELSE 0::int
 END AS undisclosedflag
FROM companybase;
\COPY companybase1 TO 'companybase1.txt' WITH DELIMITER AS E'\t' HEADER NULL AS  CSV;

Checking if pkey is valid

The following psql code creates a new table called companybasepkeyvalid that counts the number of times the pkey created in companybase1 occurs in companybase1. In this case there are 3 pkeys that are not valid for companies: Undisclosed Company, New York Digital Health LLC and Undisclosed Company.

CREATE TABLE companybasepkeyvalid AS
SELECT coname, pkey, COUNT(pkey)
FROM companybase1
GROUP BY coname, pkey;
--44771
\COPY companybasepkeyvalid TO 'companybasepkeyvalid.txt' WITH DELIMITER AS E'\t' HEADER NULL AS  CSV;

The Undisclosed Companies will be updated manually using the description to extract the company name. To remove the duplicate records of the company New York Digital Health LLC from the roundbase table run the following command:

DELETE FROM roundbase WHERE url = 'www.digitalhealthaccelerator.c';
--1

After the Undisclosed company names are updated, you will need to recreate the companybase and roundbase tables. Copy all the undisclosed companies to a text file and pull out the company names from the description.

DROP TABLE undisclosedcompanies
CREATE TABLE undisclosedcompanies AS
SELECT coname, description, url  FROM roundbase
WHERE coname = 'Undisclosed Company';
--37
\COPY undisclosedcompanies TO 'undisclosedcompanies.txt' WITH DELIMITER AS E'\t' HEADER NULL AS  CSV;

After manually pulling out the company names from the description from all 37 entries, I created the following commands to update the original roundbase table below. After the updates there are still 18 Undisclosed companies. Recheck to make sure there are no primary key issues again.

UPDATE roundbase SET coname = 'Standard Bariatrics' WHERE coname = 'Undisclosed Company' AND description = 'Standard 
Bariatrics is taking bariatric procedure and creating instruments';
--1
UPDATE roundbase SET coname = 'Jersey Watch' WHERE coname = 'Undisclosed Company' AND description = 'Jersey Watch provides 
organizers to professional.';
--1
UPDATE roundbase SET coname = 'CareSkore' WHERE coname = 'Undisclosed Company' AND description = 'CareSkore provides an 
evidence-based predictive systems.';
--1
UPDATE roundbase SET coname = 'Navistone' WHERE coname = 'Undisclosed Company' AND description = 'Navistone provides digital 
marketers a way to target the right web browsers';
--2
UPDATE roundbase SET coname = 'PolySync' WHERE coname = 'Undisclosed Company' AND description = 'Offers PolySync platform 
for driverless vehicle applications.';
--2
UPDATE roundbase SET coname = 'Platform9' WHERE coname = 'Undisclosed Company' AND description = 'Platform9 Systems Inc is a 
United States-based company.';
--2
UPDATE roundbase SET coname = 'Salamanca' WHERE coname = 'Undisclosed Company' AND description = 'Salamanca provides 
products and services.';
--1
UPDATE roundbase SET coname = 'Sourcegraph Inc.' WHERE coname = 'Undisclosed Company' AND description = 'Sourcegraph Inc. 
provides a code search engine for open source.';
--1
UPDATE roundbase SET coname = 'Sona' WHERE coname = 'Undisclosed Company' AND description = 'Sona provides products and 
services.';
--1
UPDATE roundbase SET coname = 'AlphaSource' WHERE coname = 'Undisclosed Company' AND description = 'AlphaSource in provides 
healthcare technology solutions.';
--2
UPDATE roundbase SET coname = 'Reflecktive' WHERE coname = 'Undisclosed Company' AND description = 'Reflecktive provides 
products and services.';
--1
UPDATE roundbase SET coname = 'SuperH' WHERE coname = 'Undisclosed Company' AND description = 'SuperH provides an SaaS 
platform for open source data';
--1
UPDATE roundbase SET coname = 'Islands Media Inc' WHERE coname = 'Undisclosed Company' AND description = 'Islands Media Inc 
is an IT company.';
--1
UPDATE roundbase SET coname = 'Envoy Commerce Corp.' WHERE coname = 'Undisclosed Company' AND description = 'Envoy Commerce 
Corp. provides online grocery delivery service.';
--1
UPDATE roundbase SET coname = 'Quilt' WHERE coname = 'Undisclosed Company' AND description = 'Quilt is a consumer-centric 
insurance company focused on renters insurances';
--1

Removing duplicates from ipos table

If you create the following two tables you'll notice that there are duplicate entries in the ipos table.

DROP TABLE ipospkey;
CREATE TABLE ipospkey AS
SELECT *,
concat(issuer::text, issuedate::text, statecode::text, principalamt::text) AS pkey
FROM ipos;
--10440
DROP TABLE iposcount;
CREATE TABLE iposcount AS
SELECT issuer, issuedate, statecode, COUNT(pkey)
FROM ipospkey
GROUP BY issuer, issuedate, statecode;
--9491

To clean these up I built the following tables and selected the max principalamt values if there were duplicates. There is probably a simpler way to do this but I'll leave that for the next person to discover.

DROP TABLE iposcounthigh;
CREATE TABLE iposcounthigh AS
SELECT *
FROM iposcount
WHERE count > 1; 
DROP TABLE duplicateipodata;
CREATE TABLE duplicateipodata AS
SELECT ipos.*
FROM ipos INNER JOIN iposcounthigh ON (ipos.issuer = iposcounthigh.issuer AND ipos.issuedate = iposcounthigh.issuedate AND 
ipos.statecode = iposcounthigh.statecode)
ORDER BY principalamt DESC;
DROP TABLE ipoduplicatemin;
CREATE TABLE ipoduplicatemin AS
SELECT issuedate, issuer, statecode, MIN(principalamt) AS principalamt 
FROM duplicateipodata
GROUP BY issuedate, issuer, statecode;
--939
\COPY ipoduplicatemin TO 'ipoduplicatemin.txt' WITH DELIMITER AS E'\t' HEADER NULL AS  CSV
DROP TABLE ipoduplicatekeys;
CREATE TABLE ipoduplicatekeys AS
SELECT *,
concat(issuer::text, issuedate::text, statecode::text, principalamt::text) AS pkey
FROM ipoduplicatemin;
DROP TABLE iponoduplicates;
CREATE TABLE iponoduplicates AS
SELECT ipospkey.*
FROM ipospkey
WHERE NOT EXISTS(SELECT * FROM ipoduplicatekeys WHERE (ipospkey.pkey = ipoduplicatekeys.pkey)); 
--9490

To verify that there are no more duplicates create another table.

DROP TABLE iposcountdupes1;
CREATE TABLE iposcountdupes1 AS
SELECT issuer, issuedate, statecode, COUNT(pkey)
FROM iponoduplicates
GROUP BY issuer, issuedate, statecode;
\COPY iposcountdupes1 TO 'iposcountdupes1.txt' WITH DELIMITER AS E'\t' HEADER NULL AS  CSV

You will notice that there are still 9 duplicate records that slipped thru somehow. Remove these manually.

DELETE FROM iponoduplicates WHERE pkey = 'PacTel Corp1993-12-02CA$299.00';
--1
DELETE FROM iponoduplicates WHERE pkey = 'Templeton Dragon Fund Inc1994-09-21FL$169.50';
--1
DELETE FROM iponoduplicates WHERE pkey = 'Sterling Commerce1996-03-08TX$240.00' AND proceedsamt = '288';
--1
DELETE FROM iponoduplicates WHERE pkey = 'Sothebys Holdings Inc1988-05-13NY$27.00';
--1
DELETE FROM iponoduplicates WHERE pkey = 'TD Waterhouse Group Inc1999-06-23NY$655.20';
--1
DELETE FROM iponoduplicates WHERE pkey = 'Berlitz International Inc1989-12-13NJ$34.65';
--1
DELETE FROM iponoduplicates WHERE pkey = 'Spain Fund Inc1988-06-21NY$39.00';
--1
DELETE FROM iponoduplicates WHERE pkey = 'Ultramar Corp1992-06-26CT$85.39';
--1
DELETE FROM iponoduplicates WHERE pkey = 'Goldman Sachs Group Inc1999-05-03NY$424.00';
--1

iponoduplicates is now the master ipo table that contains unique keys which are concatenated from issuer, issuedate and prinicpalamt. It should contain 9481 lines.

Cleaning mas table

There were duplicate records in this file because the original text file from SDC contained addresses that were on two rows for some companies. Therefore the easiest way to get rid of these records is to remove lines in textpad and reimport into the mas table.

Remove lines in textpad using regex: ^#.*\n and replacing with nothing.

After reimporting there will still be duplicate rows. Slap an id number on them and take the min as shown below.

CREATE TABLE mas1 AS
SELECT *,
concat(targetname::text, targetstate::text, announceddate::text) AS pkey
FROM mas;
ALTER TABLE mas1 ADD COLUMN id SERIAL PRIMARY KEY;
--114890
DROP TABLE masminid;
CREATE TABLE masminid AS
SELECT targetname, targetstate, announceddate, MIN(id)
FROM mas1
GROUP BY targetname, targetstate, announceddate;
--114825
DROP TABLE masnodupes;
CREATE TABLE masnodupes AS
SELECT mas1.*
FROM mas1 JOIN masminid ON mas1.id = masminid.min;
\COPY masnodupes TO 'masnodupes.txt' WITH DELIMITER AS E'\t' HEADER NULL AS  CSV

masnodupes should have no duplicates based on targetname, targetstate, announceddate. After running it thru the matcher you will get about 800 matching errors. We decided to exclude any rows that had equivalent targetname, targetstate where the announced dates fell within the same week. In this case we took the minimum date value and excluded the other row. Below are the queries to make those tables.

DROP TABLE masdistinctkeys;
CREATE TABLE masdistinctkeys AS
SELECT DISTINCT targetname, targetstate, announceddate
FROM masnodupes;
--114825
\COPY masdistinctkeys TO 'masdistinctkeys.txt' WITH DELIMITER AS E'\t' HEADER NULL AS  CSV
DROP TABLE masdistinctkeysmindates;
CREATE TABLE masdistinctkeysmindates AS
SELECT targetname, targetstate, MIN(announceddate) AS announceddate
FROM masdistinctkeys
GROUP BY targetname, targetstate;
--113236
DROP TABLE masdistinctkeysdatewindow;
CREATE TABLE masdistinctkeysdatewindow AS
SELECT masdistinctkeys.*, masdistinctkeysmindates.announceddate as minanndate,
CASE WHEN masdistinctkeys.announceddate - INTERVAL '7 day' > masdistinctkeysmindates.announceddate THEN 1::int
ELSE 0::int
END AS dateflag
FROM masdistinctkeys JOIN masdistinctkeysmindates ON (masdistinctkeys.targetname = masdistinctkeysmindates.targetname AND 
masdistinctkeys.targetstate = masdistinctkeysmindates.targetstate);
--114825
\COPY masdistinctkeysdatewindow TO 'masdistinctkeysdatewindow.txt' WITH DELIMITER AS E'\t' HEADER NULL AS  CSV 
CREATE TABLE masdistinctkeys1 AS
SELECT targetname, targetstate, announceddate
FROM masdistinctkeysdatewindow
WHERE dateflag = 0;
--113267
\COPY masdistinctkeys1 TO 'masdistinctkeys1.txt' WITH DELIMITER AS E'\t' HEADER NULL AS  CSV

Now when you re-run the matcher on portcokeys and masdistinctkeys1 you will get 428 errors and excluding Undisclosed fields gets you down to 356.

In Excel create a column to exclude any entries where the announced date occurs before the firstinvdate. Afterwards, manually flag the other duplicate mismatches picking the earlier announced date.

Import Flagged Matcher Data Into PSQL

After everything in the matcher output files is manually flagged create two new tables in Excel for iposmatcherouput and masmatcheroutput and copy those files into the tables.

CREATE TABLE masmatcheroutputwithexcludeflag(
 file1coname varchar(255),
 file2targetname varchar(255),
 warningmsg varchar(255),
 file1coname1 varchar(255),
 file1statecode varchar(2),
 file1datefirstinv date, --mm-dd-yyyy
 file2targetname2 varchar(255),
 file2statecode varchar(2),
 file2announceddate date,
 exclude boolean,
 excludemaster boolean,
 excludemanual boolean,
 excludemanualmaster boolean,
 excludefinal boolean
);
\COPY masmatcheroutputwithexcludeflag FROM 'masmatcheroutput-nohead.txt' WITH DELIMITER AS E'\t' HEADER NULL AS  CSV
--9400
CREATE TABLE ipomatcheroutputwithexcludeflag(
 file1coname varchar(255),
 file2issuer varchar(255),
 warningmsg varchar(255),
 file1coname1 varchar(255),
 file1statecode varchar(2),
 file1datefirstinv date, --mm-dd-yyyy
 file2issuer1 varchar(255),
 file2issuedate date,
 file2statecode varchar(2),
 excludedate boolean,
 excludemaster boolean,
 excludemanual boolean,
 excludemanualmaster boolean,
 excludefinal boolean
);
\COPY ipomatcheroutputwithexcludeflag FROM 'iposmatcheroutput-nohead.txt' WITH DELIMITER AS E'\t' HEADER NULL AS  CSV
--2591

Joining portco with ipo and mas

In order to match the company with the ipo or acquisition date we need to join the three tables together. Do this in a two step process. First join with ipo and then take the resulting table and join with mas. This will give you companies with an ipo column and acquisition column. There will be a lot of blanks because some companies didn't have ipos or acquisitions.

DROP TABLE portcoipojoin;
CREATE TABLE portcoipojoin AS
SELECT coname, statecode, datefirstinv, matcher.file2issuer, matcher.file2issuedate
FROM companybase LEFT JOIN ipomatcheroutputwithexcludeflag AS matcher ON companybase.coname = matcher.file1coname AND 
companybase.statecode = matcher.file1statecode AND companybase.datefirstinv = matcher.file1datefirstinv AND 
matcher.excludefinal = FALSE; 
--44773
\COPY portcoipojoin TO 'portcoipojoin.txt' WITH DELIMITER AS E'\t' HEADER NULL AS  CSV
DROP TABLE portcoipojoinmajoin;
CREATE TABLE portcoipojoinmajoin AS
SELECT portco.*, matcher.file2targetname, matcher.file2announceddate
FROM portcoipojoin AS portco LEFT JOIN masmatcheroutputwithexcludeflag AS matcher ON portco.coname = matcher.file1coname 
AND portco.statecode = matcher.file1statecode AND portco.datefirstinv = matcher.file1datefirstinv AND matcher.excludefinal 
= FALSE;
--44774
\COPY portcoipojoinmajoin TO 'portcoipojoinmajoin.txt' WITH DELIMITER AS E'\t' HEADER NULL AS  CSV

Then join with the portco table again and extract out the city and address.

DROP TABLE portcoipomalocation;
CREATE TABLE portcoipomalocation AS
SELECT portcojoin.*, city, addr1, addr2, indclass, indsubgroup3, indminor
FROM portcoipojoinmajoin AS portcojoin LEFT JOIN companybase ON portcojoin.coname = companybase.coname AND 
portcojoin.statecode = companybase.statecode;
--44878
\COPY portcoipomalocation TO 'portcoipomalocation.txt' WITH DELIMITER AS E'\t' HEADER NULL AS  CSV

Joining geo with portcoipomas

Check the portcoipomas table first. Undisclosed Companies may have slipped through so you might have to rerun the joins against the companybase1 table selecting the appropriate flags. You will match portcoipomas with the geo data using coname, city, year so you will need to strip the year from the datefirstinv. Use the EXTRACT(YEAR FROM datefirstinv) command. I broke these into separate tables:

CREATE TABLE portcoipomalocationexclude AS
SELECT *, EXTRACT(YEAR FROM datefirstinv) AS datefirstyear
FROM portcoipomalocation AS portco
WHERE NOT EXISTS(SELECT * FROM excludeipomadupes as exclude WHERE (portco.coname = exclude.coname AND portco.statecode = 
exclude.statecode AND portco.datefirstinv = exclude.datefirstinv AND exclude.excludeflag = 1));
CREATE TABLE companybasegeo AS
SELECT portco.*, geo.lattitude, geo.longitude, geo.noaddress
FROM portcoipomalocationexclude as portco LEFT JOIN geo ON (portco.coname = geo.coname AND portco.city = geo.city AND 
portco.datefirstyear = geo.startyear);

You will notice that the output has blank entries for the noaddress flag. This indicates that the tables aren't joining correctly. Indeed you will see that the company Capella Systems, Inc. in the geo table is named Capella Systems Inc. in the portcoipomalocationexclude table. Therefore you will need to use the Matcher tool to match the keys in the two tables.