migration, unified
TRANSCRIPT
Migration, Unified
WH
ITE
PA
PE
R
Table of ContentsIntroduction............................................................................................................................3
Moving Day........................................................................................................................................4
Case Study: Auto Sales and Service.................................................................................................5
How it Works......................................................................................................................................6
Integration and Cleaning...................................................................................................................6
Mastering with Unify.........................................................................................................................7
Move-in Ready...................................................................................................................................8
IntroductionThe process of migrating your data to new software is a lot like moving into a new house:
you want to consolidate and clean. This white paper discusses the role of data unification
and cleaning during a migration project, and it shows an example of how a major
automaker leveraged Tamr’s solution to effectively and efficiently consolidate data
in Salesforce.
Tamr White Paper – Migration, Unified 4
Moving DayMigrating your data to new software is like moving to a new house. You don’t just throw everything into the truck. You take
stock of what you have. You decide what to keep and what to donate. And you clean. You want to start fresh and get the
most out of the new place.
That’s the role played by Tamr’s data unification and cleaning services during a migration. First, you want to start with a
solid understanding of how many customers, leads, suppliers, or parts you have, which will establish accurate baselines
for metrics. More abstractly, you want to know how much data you really have, so you can request the optimal storage and
processing resources.
Second, you want to impress your users, especially those who will touch the new system all day long. It’s frustrating enough
to learn new CRM or ERP software. Clean and reliable data will excite your users and keep them productive during the
transition. After all, their old tricks for handling bad data won’t work, and the new dashboards are sure to reveal data issues
they didn’t expect. Cleaning and mastering the data with Tamr Unify helps minimize this friction. Plus, the system’s everyday
users can contribute their expertise to Unify’s machine learning algorithms, so the users feel invested in the result.
Unifying and cleaning your data during a migration sounds like making a hard project harder, but the moving analogy holds:
you’ve already set aside the time to touch everything in your house and the money to ship it. Likewise, for your software
migration, the database designers are already mapping the old schema to the new one. The data engineers are already
assembled for export and import. Unify works with these teams to find new relationships in your data, endow the new
system with reliable results, and develop a repeatable process to keep it that way. What is more, Unify learns ground-up
from your data and experts, achieving higher accuracy than rules based systems in less time.
Tamr White Paper – Migration, Unified 5
Case Study: Auto Sales and ServiceFor these reasons, when a major automaker migrated their customer, vehicle, and company contract data to Salesforce,
they turned to Tamr. Working with an integration partner, Tamr ensured the new Salesforce instances were trustworthy and
clean. Tamr consolidated customer data volume by 55% by clustering together records that represent the same person.
Vehicle and company records were consolidated 44%.
Reviewing high impact matches in Unify to train the model
Tamr Unify builds these clusters of records by first identifying pairs of records that represent the same entity. For example,
the same person may change addresses or phone numbers between service contacts. Because Unify is a machine learning
solution, it presents your CRM experts with pairs of customer records where it needs the most guidance to determine
similarity, and it asks the experts a single, simple question: “match” or “no match”?
As the experts label more matches, Unify suggests “high impact” pairs that are most likely to improve its clustering model.
When accuracy reaches a target level, the experts are done. They have built a robust clustering model ground-up by
answering simple questions about representative samples of the data, not writing mountains of rules about patterns and
exceptions.
Tamr Dedup Results
Raw Mastered
Mill
ions
of R
ecor
ds
Entity
Customers Vehicles Companies
4.5
2.50.45 0.25
10
8
6
4
2
0
10
4.8
Tamr White Paper – Migration, Unified 6
How it WorksIntegration and CleaningTamr’s Professional Services team works with your data engineers to integrate Unify into your migration pipelines. For the
automaker’s migration, Tamr consultants provided the tools to convert raw data into JSON to upload to Unify via RESTful
APIs, which of course can be automated.
Once data is in Unify, a schema mapping interface is available that makes it easy to map old columns to new columns and
transform their contents. These mappings can simultaneously prepare data for migration targets, enrich data for Unify’s
machine learning, and preserve the original data for provenance.
Tamr’s expertise with CRM data enables customers to understand the key cleansing and validation transformations required
for the incoming data: standardizing special characters, filtering out placeholder values, and validating and parsing email
and street addresses. These transformations improve data quality generally and maximize Tamr’s accuracy. Tamr wrote
these transformations for the automaker, but data engineers can write them as well in the Unify interface, in a syntax
comfortable to SQL users. Behind the scenes, Unify packages transformations into fast Spark jobs.
Sample architecture for migrating and mastering CRM data with Unify
Tamr White Paper – Migration, Unified 7
Mastering with UnifyUnify learns to cluster records together by first learning to identify pairs of matching records. Of course, most of these
“matches” are not exact matches. For example, the same vehicle may appear with multiple owners, or the same company
may appear with multiple shipping addresses.
At the automaker, CRM experts provided examples of such fuzzy matches, based on their experience. Unify then learned
from these examples how to balance similarities in all of the record attributes. It might lend higher weight to the similarity of
customer names over the similarity of their addresses. Teaching Unify is easy. If Unify is making incorrect matches because
the records’ country values match, then experts can search for pairs of records with matching countries and label more
“non-matches.” Likewise, if Unify is missing correct matches with the same email addresses, then experts can search for
records with matching email addresses and label more matches.
During this process, users never need to write a rule or any code. They just need to be able to provide examples of what
they consider “matches” and “non-matches”. This process dramatically reduces the time and technical resources required
to get to business-ready data, and it improves the quality of results by tightly integrating data consumers and experts into
the process.
Unify clusters together records referencing the same customer before creating a Golden Record
Unify provides Spark-powered transformations via a familiar, SQL-like syntax
Move-in ReadyA system like Salesforce that is so vital to so many users has to look good on day one.
In the case of the automaker, Tamr’s continual touchpoints with data engineers and
CRM users kept the project moving and ensured everyone felt like they contributed to
its success. In the moving analogy, it’s as if everyone loves the new layout and decor,
because everyone played a role in the transition.
With Tamr’s help, the automaker unified and mastered seven customer sources, three
vehicle sources, and three company sources into the Salesforce instance, in six months.
Schema-mapping in Unify aligned the original data to the new system’s schema.
Transformations cleaned invalid values and placeholders and parsed more complex
attributes. Finally, three machine learning models trained by CRM experts mastered the
customer, vehicle, and company entities, cutting total data volumes by 44-55%.
The success of the migration encouraged the automaker to use Tamr to clean and
consolidate customer data for other marketing and sales systems, in multiple countries.
Tamr’s flexibility, scalability, and history of collaboration empowered the automaker to
incorporate more data sources, maintain privacy, and build a personalized experience
for each customer. In this way, unifying your data with Tamr during a migration amplifies
the productivity and possibilities in its new home.
About TamrTamr is the enterprise-scale data unification company trusted by industry leaders like GE, Toyota, Thomson
Reuters, and GSK. The company’s patented software platform uses machine learning supplemented with customers’ knowledge to unify and prepare data across myriad silos to deliver previously unavailable
business-changing insights. With a co-founding team led by Andy Palmer (founding CEO of Vertica) and Mike Stonebraker (Turing Award winner) and backed by founding investors NEA and GV, Tamr is transforming
how companies get value from their data.
To find out more or register for a demo visit tamr.com
© Copyright 2019 Tamr, Inc. All rights reserved. | 02.19