Streamline 'Create new number' wizard

Problem

Too many steps leading to drop-off and inaccurate setup

The tracking number creation wizard had a ~30% drop-off rate in the questionnaire section alone. Many numbers that were created were also misconfigured, which undermined attribution accuracy. I made the case to prioritize this as the highest-leverage opportunity.

The flow depended on a rigid sequence of binary questions that overwhelmed new users and slowed experienced users, especially when creating in bulk. It also failed to show the range of platforms where tracking numbers could be used. As a result, many customers created a single number for everything instead of separate numbers per campaign, limiting both attribution accuracy and the number of tracking numbers per account.

Research and FullStory analysis showed friction at both ends of the spectrum:

New users experienced decision fatigue, leading to hesitation and drop-off.
‍Experienced users found the flow repetitive and click-heavy, especially when creating multiple numbers.

The series of binary questions, "this or that" cause decision fatigue and confusion for new users.

The drop-off rate is consistently around 30% for the questionnaire section. (FullStory)

Design thinking

Streamlining the flow has the highest impact

I made an early call to keep UI and content changes minimal. The issue was the flow: too many decisions up front and no guidance toward the right choice. A focused scope also meant faster shipping, less engineering effort, and a cleaner before/after signal for measuring what actually changed.

Another constraint I designed around carefully: this wizard is used constantly, including by power users who had memorized the click sequence. Any change had to work for both ends of the user spectrums without disrupting experienced workflows.

Reduced the questionnaire to two paths

I restructured the questionnaire into two clear paths: “On my website” and “Somewhere else.” This replaced a branching set of binary questions with one decision, reducing drop-off for new users and cutting clicks cutting clicks for experienced users. In complex setups, clicks dropped 65%.

The tradeoff: Removing progressive questioning meant that the consolidated step had to carry more context, which led to the visual platform library in Question 2 (Somewhere else).

Using behavioral data to advocate for design change

My instinct was to remove the dedicated destination-number step entirely. In the original flow, users had to enter a destination number before they had even selected a number, which felt backwards. In practice, you pick the number first, then decide where it should route. I proposed consolidating destination number, call recording, and whisper message setup into a single Number Routing step placed after number selection.

The VP of UX pushed back, and the concern was valid: the destination number was used to pre-populate the area code in the number selection step. Moving it later would break that logic.

I pulled Looker data showing that about 85% of accounts had only one destination number. With that pattern, we could pre-populate the area code from the existing destination number on the account and let users change it when needed. That data addressed the risk, and we moved forward with the combined Number Routing step.

I applied the same principle to post-creation suggestions. Data showed users often routed tracking numbers to call flows immediately after creation, so I surfaced those routing options directly in the new step instead of leaving them for a disconnected follow-up experience.

Visual platform library to drive discovery

Research revealed that most users didn't know how many platforms tracking numbers could be applied to, many were creating a single number for everything. This directly limited source tracking numbers per account, the metric the business was focused on growing. I introduced a visual platform library organized by popularity.

A "Most Popular" section at the top, followed by categorized sections for search ads, directory sites, and social media. Familiar logos and names made it fast to choose the right platform and reinforced “one platform, one number” for better attribution.

Search was used only 3%, suggesting browsing worked, and new-customer numbers per account increased from 12.2 to 13.5 after launch.

Contextual suggestion is better than generic ones

Tracking numbers are most effective when integrated with external platforms, yet many users were unaware of these integrations .The issue was visibility: integration setup was buried after wizard completion, and most users never reached it.

I introduced contextual integration suggestions on the confirmation screen, tailored to the type of tracking number just created. This replaced a generic "Suggested Next Steps" module with guidance tied directly to the user’s action, driving 24% higher engagement than the previous module.

Outcomes

13% boost in wizard completion rate

The new updated wizard achieved a 61% completion rate compared to 54% for the old wizard, representing an 13% increase.

11% Increase in numbers per account for new customers since release, growing from a steady 12.2 to 13.5 tracking numbers per account in November 2024.

32% Decrease in the drop-off rate for the questionnaire section. Now, 77% of users successfully navigate through it, up from 66% previously.

11% Boost in numbers per account for new customers grew from 12.2 to 13.5.

Reflection

This project reinforced something I carry into every project: the most impactful design decision is often about what not to change. I could have redesigned the entire wizard. Instead, I kept asking what the smallest change was that would create the biggest impact. That focus is what made it possible to ship quickly, measure cleanly, and see real behavior change.

It also reinforced the value of working from behavioral data. The 85% destination number reuse pattern came from looking at what people actually did, and it changed the design in a meaningful way.

Streamlining tracking number creation: more numbers & fewer mistakes