How Narrator Works

Model business concepts, not fact and dimension tables.

Companies typically have hundreds of data sources and thousands of source tables. Despite this complexity, all data questions can actually be answered from a small set of core business concepts. Narrator drastically simplifies data modeling by defining those concepts and combining them to make datasets.

All insights begin with a “simple” data question

When data questions are asked, they come in a natural, human format, and seem straightforward:

DATA CONSUMER

ASKS A QUESTION...

How many people who called us signed a contract?

ANALYST

...RETURNS AN ANSWER.

275 people who called us between Jan 1, 2019, and Jan 15, 2019, signed a contract.

But for analysts, the process is slow, tedious, and error-prone

DATA QUESTIONS AND ANSWERS

1. Data consumer asks data question

2. Analyst translates the question into the proper format

3. Analyst researches relevant tables and their dependencies

4. Analyst writes a data query

5. Analyst validates the data output

How many people who called us signed a contract?

How many new unique customers who called us (via Zendesk) for the first time ever during Jan 1 and Jan 15 signed a contract?

2. Analyst translates the question into the proper format

1. Data consumer asks data question

4. Analyst writes a data query

3. Analyst researches relevant tables and their dependencies

5. Analyst validates the data output

275 new unique customers who called us via Zendesk during Jan 1 and Jan 15 for the first time signed a contract.

Narrator solves this problem by reframing data modeling to be from the point of view of the data consumer:

NARRATOR PROCESS: DATA QUESTIONS AND ANSWERS

How many customers called us between Jan 1 and Jan 15, and then signed a contract at any time?

1. Data consumer asks data question

2. Consumer assembles the question using relevant business concepts

275 new unique customers who called us via Zendesk during Jan 1 and Jan 15 for the first time signed a contract.

3. An answer is returned

1. Data consumer asks data question

2. Consumer assembles the question using relevant business concepts

3. An answer is returned

Our Approach

Overview

DATA ENGINEERS

Identify and define core business concepts in SQL once

Analysts and data engineers no longer need to manage the relationships and dependencies of thousands of tables.

DATA CONSUMERS

Provide tools for anyone to combine these concepts to build any dataset, over and over again

Data consumers can now ask data questions in a way that mirrors how they already understand the business.

people

who

Called Us

and

Signed Contract

Our Data Modeling Process

DATA ENGINEERS

Define Concepts – How we build our core data models:

Write simple SQL on top of your warehouse, based on what your customers do.

SELECT 
  u.email                  as customer, 
  ‘subscription_upgraded’  as activity,
  s.id                     as activity_id,
  s.upgraded_at            as  timestamp
FROM subscriptions AS s
JOIN users AS u
  ON s.user_id = u.id
WHERE s.status = ‘upgraded’

via Internal db

Upgraded Subscription

Submitted a Ticket

Page View

Call Support

Opened Marketing Email

Signed a Contract

Logged In

1 week to define all activities

DATA CONSUMERS

Assemble Datasets

1. Start with a data question:

How many

people

who

visited the website converted to a lead?

2. Rephrase it in terms of customer and business concept:

Count of

people

who

visited site

then

Submitted Lead

Within 7 Days

3. Assemble concepts to build a dataset:

Benefits

Simpler
Definitions

Easy to write SQL (average of 14 lines vs. 300+).

Less mess,
Fewer Arguments

No redundancy—each concept is clearly defined only once, and new activities don’t need to be created often.

No more ad-hoc
requests

Everyone can answer their own questions and assemble datasets in 5 minutes.

Try it with your own data

Your team could be using Narrator in as little as 24 hours. Setup is easy.