Skip to content
Yujia Zhao

Product-led growth for a no-code data prep enterprise tool

Improved trial-to-paid conversion rate from under 0.1% to over 2% through journey & friction analysis, A/B testing experiments, and streamlining user experience.

Trifacta, a no-code cloud data cleansing & preparation tool, established a Growth Team, with the mission of acquiring, engaging, and retaining small to medium customers through trials and product-led growth. I was the Lead Designer on the Growth Team, responsible for:

  • Define growth strategy with Product and Engineering Leaders
  • Plan design projects
  • Execute growth projects hands-on while overseeing a junior designer's work in a fast-paced environment.
The Trifacta Dataprep home screen, headed "Welcome, Alice!" A getting-started panel explains that a flow is needed to transform data and offers a Create a flow button, above three template cards for moving data into BigQuery, filtering rows, and removing duplicates. A right-hand rail shows "15 trial days left" with a Subscribe button.
My role
Lead Designer
Time
2022-2023
Worked with
1 designer, 2 PMs, 1 Tech Lead, 1 Data Analyst

Journey & frictions workshop

Surface and align on the end-to-end user journey

Growth practice is based on the cycle of hypothesis, A/B experiment testing, and learning. However, as a designer, I want to ground hypotheses in user insights, and that starts from the end-to-end user journey and our shared understanding of the frictions distributed in cross-functional teams.

Three overlapping workshop boards. "What is a new user like?": a journey from prospects through acquisition to new user and activated user, above columns of sticky notes for things new users understand, have done, want to know and want to do. "User progression analysis result from workshop 1": for data pipeline work, what users need, know and don't know at each stage, from connectivity and transformation to automation and performance. "User journey & hurdles" for Trifacta Cloud and Dataprep: the path from prospect through sign-up, new user, connecting to data and running transformations to paying user, with the hurdles at each step listed in orange.

First focus: Top of the funnel

We are losing half the people at the very beginning of the journey

Equipped with a holistic understanding, we also dove into the data to see which part of the journey to focus on first. We learned that we have a near 50% drop-off for Flow creation. This is very concerning, because Flow is the only object/vehicle in Trifacta for users to transform data. If they are not even creating Flows, they are definitely not experiencing the aha moment, let alone subscribing.

A funnel chart of trial users. Of 763 users with any event, only 412 reach CreateFlowEvent — the drop-off highlighted in orange — then 233 add a dataset, 230 add a recipe, 177 add a script line and 99 run a transforming job.

Hypothesis

Users don’t know what a Flow is, while the UI ushers them down the most error-prone path to create one.

If we can: 1) establish the mental connection between Flow and user’s data task, 2) usher users to create a Flow directly after such understanding, we should be able to increase the Flow creation rate.

Left, the old Trifacta homepage with Import Data as the primary button above Create Flow, outlined in orange. Right, a diagram of the routes from landing in a trial to creating a flow: the Import Data path, highlighted in orange, runs through the import data page and library page, far longer than clicking Create Flow directly.

Test 1: Clear starting point

Introduce a Get Started section & flip the order of CTAs.

The Get Started section helps users understand that a Flow is what they need to transform data, and guides them to create one with very simple language and a direct CTA. The primary CTAs follow the same action priorities.

The redesigned Trifacta homepage. A new Get Started panel, outlined in yellow, reads "Welcome, John! To transform your data in Trifacta, you need a flow" with a Create a flow button. Top right, also outlined, Create a flow is now the primary button above Import data.

Test result & learning

  • +21%Flow Creation Rate (experiment success metric)
  • +17%First Job Run Rate (activation metric)
  • +11%Second Job Run Rate (conversion indicator)

We saw positive impact not only on the direct experiment success metric, but also on downstream key levers of growth. We learned that introducing what a Flow is helps users’ onboarding, leading them to the product’s key value surface — a Flow.

Second focus: Making it dead simple to get data in

A significant drop-off at “add a dataset” step indicates users are having trouble getting data in.

The whole point users are in Trifacta is to transform their data. If we can’t get users to connect to data, an empty Flow, even though understood and created, is useless. As shown in the image, the “Add datasets to flow” modal doesn’t have a clear way to add new data.

The old "Add datasets to flow" modal over an untitled flow: a searchable list of already-imported files with checkboxes and All, Imported and Reference tabs. The only way to bring in new data is a small "Import datasets" link in the bottom corner.

Analysis & hypothesis

For new users, this modal is empty and has no clear CTA to import data.

If we can: 1) make this a comprehensive UI for all data importing options, 2) make it very simple, users should be able to add data to their Flow.

A flow of a user's thinking when adding data: "I need to add data to this flow," then "Where is this data?" — branching into whether it is already imported in the library or sitting on S3 or Snowflake. Beside it, two design options with open questions, such as whether new users would understand "Library" and "Import".

Test 2: Streamline adding data

A right panel outlining all options for where and how to add data

This panel works for new and existing users when they want to add data to the current Flow. It outlines all options right there instead of directing users to the “Import Data” page.

The redesigned "Add data from Library" modal. The dataset list stays on the left, and a new right panel lists every way to import data: Trifacta storage, Amazon S3, Shopify and Google Sheet connections, plus Add connection, Upload data file and Use demo dataset.

Test result & learning

  • +6%Data Added Rate

We learned that sometimes a usability optimization for a page without changing any feature or functionality can go a long way.

Other experiments

We conducted numerous experiments and made many improvements mostly without changing functionality.

We continued to analyze, hypothesize, and experiment with new ideas along the user journey, addressing many friction points. Some noteworthy ones are: improving in-context help, cleaning up terminology, adding dataset-level recommendations, etc.

Three experiments side by side. In-context help in the transformation grid: Tutorials and Tour this page buttons in the header, and a "How to filter data" 2-minute video and article link in the filter menu, both outlined. A terminology cleanup on a flow's output details panel: "Scheduled Output" struck through and renamed "Output", with "Manual settings" and "Scheduled settings" renamed "Manual output" and "Scheduled output". Dataset-level recommendations in the recipe panel: before any steps are added, a Suggestions box offers "Fix missing values in 2 columns", "Fix empty columns" and "Fix mismatched values".

Outcome & impact

  • <0.1%→2%+Trial-to-paid Conversion Rate
  • +17%Trial Signup Rate
  • 132k→1.6MMSelf-service ARR
  • 2→20New logos per month through Trial

Over 16 months, we achieved significant business outcomes by optimizing the user experience through thoughtful analysis of user needs, mental model, journeys, and frictions, and rolling out changes with the utmost simplicity.