Case study · Anaplan

Creating data movement and processing pipelines

Giving IT and business specialists a native way to bring data into Anaplan, process it and send it back out.

Role
Lead product designer
Team
Designers, researcher, product managers, engineering teams
Users
IT specialists and business specialists
Impact
Several large customers moved from third-party ETL tools
Anaplan Data Orchestrator's pipelines map view in edit mode, showing a pipeline on a canvas that runs from a Snowflake source node through a Validation node to a Remove duplicates node.
Product

What Anaplan does

Anaplan is a planning tool for large businesses. Businesses use it to build models that help them forecast what would happen in different scenarios.

Team

Who I worked with

I was the lead designer, working with other designers, a researcher, several product managers and several engineering teams. I focused on the core part of the product, and made sure the surrounding parts, which other designers owned, were consistent with each other and with the core.

The problem

Customers’ data had to leave Anaplan to be moved

Anaplan is a data modelling product, so customers needed to bring in data from many separate systems and, once it was processed, push it out to other systems. To do this, it relied on third-party ETL (extract, transform, load) tools to handle the data movement.

This worked to an extent, but the tools weren’t native to the Anaplan ecosystem, which made for a poor experience for users. As a result, a project was launched to build Anaplan’s own ETL tool.

  1. Source systems
  2. Third-party ETL tool
  3. Anaplan
  4. Third-party ETL tool
  5. Other systems
Before the project, data moved in and out of Anaplan through third-party ETL tools.
Consideration

Designing for IT and business users

Research showed that Anaplan customers had no single way of sourcing and transforming data. Some needed extensive IT solutions, handled by IT specialists. Others relied on ad-hoc processes carried out by business specialists, such as HR managers or data analysts.

So the product had to work for both. I made sure it was easy to use for people who weren’t technically minded, while technical users could still apply their existing knowledge to build custom solutions.

Four states of the pipelines map view: the default view, history mode with a run history panel listing a completed and a failed run, edit mode with add buttons between nodes, and a test run on sample data with a console open below the canvas.
One visual map for viewing, editing, testing and reviewing a pipeline, with a console for those who want the detail.
Consideration

Building on industry standards

Our customers already used ETL tools, and had picked up certain behaviours and ways of working from them. For them to pick up Anaplan’s solution easily, I couldn’t design something from scratch. I had to use the standards that already existed and the language our customers already used.

To do this, I researched existing tools in depth and worked with a researcher to interview customers about their current processes. I also reviewed Anaplan’s content guidelines to write copy that followed them while still being familiar to our customers.

Four screens of the orchestration designer: a pipeline between Start and End nodes, a menu offering Add pipeline and Add parallel block, a parallel block with two branches, and an Add pipeline dialog listing published pipelines.
Orchestrations use patterns customers already knew: start and end points, parallel blocks and branches.
Consideration

Leading across multiple teams

This wasn’t a single scrum team. The project had several. So as lead designer, my job wasn’t only to focus on what I was creating, but also to review what other designers were working on so that we stayed consistent.

This meant regular catch-ups to keep each other in check and ad-hoc workshops when they were needed. It also meant making sure the components we created were easy to share, and that each problem was solved once, so no one duplicated work.

A component sheet showing pipeline node variants in default, selected and warning states, connector line styles, sink nodes for Snowflake, Amazon S3, Microsoft SQL Server, PostgreSQL, Amazon Redshift and MySQL, and the nodes in use in a pipeline structure view.
Shared components meant each problem was solved once.
Outcome

Customers moved to a native tool

Several large customers

As a direct result of this work, several large customers switched from general ETL tools to Anaplan’s own.

More efficient modelling

This made the core part of Anaplan, modelling, much more efficient in how it handles data, which reduced operating costs.

The Pipeline runs history page in Data Orchestrator, listing a single pipeline and an orchestration with three grouped pipelines, all with a status of Completed.
Run history for pipelines and orchestrations in Anaplan’s own tool.