18
tamr

Tamr Launch with Andy Palmer

  • Upload
    tamrinc

  • View
    150

  • Download
    1

Embed Size (px)

DESCRIPTION

Tamr launched at DataBeat in May 2014 via a presentation by co-founder Andy Palmer. Tamr connects and enriches the vast reserves of underutilized internal and external data, allowing enterprises to use all their data for analytics and decision making. Tamr combines machine learning and advanced algorithms with collective human insight to identify data sources, understand relationships and curate siloed data at scale. Leverage All Data Organizations have vast, untapped reserves of structured and semi-structured data from internal and external sources. Instead of using just a portion of available data, Tamr lets customers analyze any and all of their sources. As Tamr connects more sources, it gets better, improving accuracy and speed. And Tamr works smoothly with the data tools and experts customers have already invested in. A Growing Data Challenge The cost and complexity of preparing the massive variety of internal and external data to power analytics and applications are unacceptably high. As a result, most organizations use less than 10% of the relevant data available. Tamr dramatically reduces the time and effort required to connect and enrich data sources, allowing data scientists, analysts and managers to focus on data-driven innovation. The Tamr Solution Tamr starts by analyzing data sources—tens, hundreds, thousands—applying advanced algorithms and machine learning to connect and curate 90% or more of attributes and records. To improve precision, it taps experts with the best knowledge or insight on particular sources. This unique collaborative approach allows Tamr to speed up data connection and preparation and to provide enterprise scalability. Big Benefits in Little Time Tamr unleashes the power of 100% of your data. One Tamr customer has connected more than 15,000 data sources in a single view. Another that struggled with slow, manual curation saw Tamr automatically curate 80% of attributes. And a third saw Tamr finish a connecting and enrichment project in 2 weeks that typically takes 6 months.

Citation preview

Page 1: Tamr Launch with Andy Palmer

tamr

Page 2: Tamr Launch with Andy Palmer

New tech is great, but the quality and connectedness of enterprise data often sucks

the dirty data secret

Page 3: Tamr Launch with Andy Palmer

scientific freedom

Good for research creativity, bad for data connectivity

Page 4: Tamr Launch with Andy Palmer

scientific freedom

Good for research creativity, bad for data connectivity

the integrated view Collaborative R&D through open data sharing

Page 5: Tamr Launch with Andy Palmer

Good for research creativity, bad for data connectivity

the integrated view Collaborative R&D through open data sharing

the source challengeFifteen thousand strong…and in need of a new approach

scientific freedom

Page 6: Tamr Launch with Andy Palmer

top down integrationNeat, clean…

Page 7: Tamr Launch with Andy Palmer

Neat, clean…and relatively inflexible

top down integration

Page 8: Tamr Launch with Andy Palmer

Neat, clean…and relatively inflexible

The Choice:Ignore itOr start all over!

The Consequences: Missed opportunity Ballooning costs

top down integration

Page 9: Tamr Launch with Andy Palmer

An exponential challenge

the missing capability

Connecting and curating in an automated way

semi-structured data: JSON sources

Page 10: Tamr Launch with Andy Palmer

Embrace the reality of data variety across the entire enterprise

bottom-up curation

Probabilistic approach as primary design pattern — some semantic web mojo

the time has come

Page 11: Tamr Launch with Andy Palmer

1990’s web:probabilistic search and website connection!

2020’s enterprise:probabilistic data source connection & curation

back to the future

Page 12: Tamr Launch with Andy Palmer

Can we remove the ceiling on the number of data sources that can be dynamically integrated?

hypothesis

Page 13: Tamr Launch with Andy Palmer

NEA

®

Page 14: Tamr Launch with Andy Palmer

early production results

15K sources integrated into one view

Tamr unified view

Page 15: Tamr Launch with Andy Palmer

early production results

Over 90% reduction in manual reviews

Records

90%reduction

Unique

Manual ReviewMatched

3% to manually review

Proprietary

Tamr

Page 16: Tamr Launch with Andy Palmer

!key design point!

• Continuous bottom-up/ probabilistic approach Combination of Machine Learning and Expert SourcingIntegrated data and metadata through APIs

Page 17: Tamr Launch with Andy Palmer

NEA

®

Page 18: Tamr Launch with Andy Palmer