SC

Data network effects

What are Data network effects?

Usage generates observations that feed back into product quality for other participants. The effect depends on the usefulness of additional data, diminishing returns and the availability of substitute data. Data exchange, private-account personalization and data volume alone do not establish this mechanism.

How Data network effects work

Strength depends on whether marginal data still improves the product, whether the data is proprietary or reproducible, and whether the improvement is visible to users. Most claimed 'data moats' fail at least one of these tests.

How companies won with Data network effects

From the win chains in Strategy Canon case studies: what the winner did, and what its rival did at the same step.

vs. Practice Fusion

athenahealth

Every denial becomes a rule for all clients

What it did
50+ staff researching payer rules, 100+ added monthly (2007)
Practice Fusion
No claims flow of its own; data sold to pharma and labs instead
Read the athenahealth case study →
vs. Rdio

Spotify

Listener scale feeds personalized discovery

What it did
Bought Tunigo and The Echo Nest (2014) to feed editors song and skip data
Rdio
Discovery through following other listeners; dropped Spotify-owned Echo Nest recommendations for Gracenote (2014)
Read the Spotify case study →
vs. Akamai · 2010–26

Cloudflare

Counter-positioning: free traffic lowers cost and trains defenses

What it did
CDN free since 2010 because it serves the paid security products
Akamai
Traffic only from paying customers, billed by use
Read the Cloudflare case study →
vs. Afterpay · 2014–19

Affirm

Underwriting each transaction in real time

What it did
Levchin's PayPal fraud-model work carried into per-purchase risk models using 70,000 variables (2014)
Afterpay
No credit-score check; ASIC flagged limited inquiry into borrowers' finances (2018)
Read the Affirm case study →

Also tagged: ZoomInfo

The question to ask of a company

Does the ten-millionth user's data still improve the product for the first user? Could a competitor buy or synthesize equivalent data?

Who names it — 2 of the canon

AuthorWhat they call it / where it appears
NFXData network effect (type #9)
HelmerA route to Network Economies or Process Power, depending on whether the loop is cross-user or internal