A Proposal and Request for Comments

My primary interests in sending out this email ...

- Is anyone interested in working together to
  build a set of open source linux/bash/python scripts
  to auto-download and crunch census sf1/sf3/tiger
  data into ready-to-map digested postgis tables along
  with appropriate indexes?

- (begin advertisement)
  I offered to setup Georgia's 2000 SF1/SF3/Tiger
  data (complete, full granularity) in postgis/mapserver
  for $25,000 based on a javascript/mapscript front end
  that I would write. Or 1990 and 2000 data and providing
  both simple linear growth rate projections and age cohort
  component projections with pop pyrmaids, etc. for $75,000.
  Georgia is not buying because they're at the tail end of
  a $1M+ ESRI system.
  Is anyone interested in such a system or a subset thereof?
  (end of advertisement)

==========================================================

What follows is focused on the United States,
though may be applicable in other countries:

I believe that due to the economies of scale a commercial
interest will eventually provide a google/mapquest style gis site
for the commodity U.S. Census SF1, SF3, and Tiger data.
This site will:

- be very fast (think google) for interactive data mining

- population/race/ethnicity/age data at finest grained level,
  i.e. down to census block, along with all other granularities
  and geo boundaries supported

- all economic and housing data, i.e. _ALL_ SF1/SF3 data

- drill down to population pyramid charts or numeric data tables

- support logon/save/restore session state which allows
  someone to develop a presentation over several
  days using a consistent image size and registration

- support private data access via the same logon mechanism

- support all map image sizes up to 4' by 4'

- include older decennial data, i.e. 1980, 1990

- generate growth rates for map or drill down display

- provide simple linear total pop projections (actually
  statistically one of the more accurate techniques
  for total pop projections, though insufficient
  for planning purposes)

- provide full age cohort-component pop projections at
  sub-county level for individual ethnicity/race groups based on
  '90/'00 SF1 data and mortality/natality tables; and
  present via maps, line charts, pop pyramids or drill
  down to numeric tables

This site will be funded through a combination of:

- Discrete advertising since every government agency planner
  will use this site along with residential and commercial
  developers, banks, and corporations.

- Service fees for mounting private corporate data and granting
  access to this private data via login mechanism.  Using this
  mechanism a corporation can get their data mounted for a
  small fraction of the cost of setting up the equivalent
  capability in house.

Thoughts?

==============================================================

Filling in some blanks ...

I left the CDC last February and am casting around for
either employment or a business model for a product. While
at the CDC's Information Technology Branch I worked on
several projects and am familiar with the problem of complex
data mining user interfaces, sql optimization, etc.

Mid summer, while working with Mapserver/Postgis and some GA
Census block data I had this vision  :)  ... and approached
the State of Georgia with the offer to mount all of 2000
SF1, SF3, and Tiger data using Mapserver/Postgis on one of
their servers for $25,000.  I planned on spending roughly
$7,000 out of that $25,000 on my own server and on making my
investment back by selling the same setup to 4 or 5 states.
By which time it should be possible to pick up even more
states and turn a profit.  The fantasy went further, that
I would return later and pick up another $20K or so for
1990 data, some more for linking the two together with the
Census relationship files and providing the growth rates, etc.

The spoiler is the State of Georgia has already invested
5 years and over $1M on an ESRI solution.  While what I
offered was more value for less money than they will spend
to bring their current project to a close, having made that
major time and money commitment they are not in a position
to change their direction.  I may yet crunch a SF1 1990/2000
subset into SQL and write the age cohort component pop
projection for sub-county areas backend as a web service for
them.  That's still on the table for negotiation.

Let's review the economies of scale that make this work.
Once the user interface and backend structure is setup,
the incremental cost to add a state is, roughly:
- one week to purchase and setup additional hardware (hardware time)
- one week to download/crunch/tune/QA data (db time)
- one week to add or upgrade a user interface feature (code time)
- one week client management time
I estimate a good small two person team could crank it out in one
man month with $7,000 going into hardware and network fees.



Heitzso
GrowthModels
MetaMedia, Inc.