Understanding an understudied cohort

We knew Paid Social was underperforming. We didn’t know why.

Paid Social generated 13% of our total traffic, becoming our second largest acquisition channel. However, the conversion rate was only at 0.3%, comparing to 3% sitewide average.

Because we were actively spending money on Facebook and Instagram ads to bring these users in, every dropped user represented both wasted marketing spend and lost revenue.

Paid Social generated 13% of our total traffic, becoming our second largest acquisition channel. However, the conversion rate was only at 0.3%, comparing to 3% sitewide average.

Company is also sending money on funding paid ads, so if we can increase traffic conversion, we could save roughly ~$x on spends.

Unlike users who moved through Thumbtack’s native discovery flow, Paid Social users landed directly on a Service Provider page (SP page) - without marketplace context and pro comparison.

Most existing research done on SP page also reflected users who were already closer to hiring, leaving major knowledge gap around Paid Social users.

My challenge was to turn those unknowns into hypotheses the team could actually test.

Turing fragmented signals into a behavioral model

Rather than jumping directly into another redesign, I triangulated evidence from the existing journey, historical research and experiments, behavioral data, and the Paid Social team's domain knowledge.

Each investigation helped answer a different question:

1.Journey diagnostic: Where might the journey be breaking?

• Typical Thumbtack journey

Search → Define project → Compare pros → Evaluate pro → Request

By the time these users reached a Service Provider page, they had already established project intent and comparison context.

• Paid Social journey

Social ad → Evaluate pro → Request

Paid Social users bypassed those context-building steps and arrived directly at evaluation. The same SP experience was therefore serving users with very different levels of intent and context.

2.Cross-functional hypothesis workshop

Which assumptions were actually worth testing?

Because much of our understanding was still inferred, I partnered with Product and Paid Social to bring together what we knew from the product with what Marketing understood about acquisition behavior.

Rather than treating those assumptions as user truths, I organized them into four behavioral hypotheses:

• Match expectation
  Preserve the context that earned the ad click

• Build confidence
  Surface credible pro evidence earlier

• Create clarity
  Explain why you’re here and what happens next

• Earn commitment
  Make the next action appropriate to the user’s level of intent

This gave the team a shared framework for discussing Paid Social friction, and turned a collection of ideas into explicit hypotheses we could design and test against.

From behavioral hypotheses to a testing strategy

Once we had a clearer model of the problem, I used it to structure our experimentation at two different altitudes.

Track 1: Optimize the existing journey

Hypothesis: Confidence

Could stronger social proof help users build trust without changing the underlying journey?

I designed a low-lift experiment that elevated raw reviews earlier on SP, balancing trust-building against the risk of disrupting conversion momentum.

Track 2 — Challenge the journey itself

Hypotheses: Expectation + Clarity + Commitment

On the other hand, we wanted to take a bigger bet and see if having an interstitial page sits between ad → SP would orient Paid Social users and increase their confidence in hiring.

I asked myself: How might we bridge the gap between ad curiosity and provider evaluation?

Working closely with User Research, I translated each hypothesis into a high-fidelity prototype for moderated concept testing.

We wanted to understand: Which approach makes first-time visitors feel ready to continue? And if a net new surface strengthen their confidence.

Testing the concepts side by side let us evaluate the underlying confidence-building strategy.

From assumptions to evidence

The study tested three concepts against the existing experience. The baseline ultimately performed best overall, while the research surfaced several key insights of cohort behavior:

• Validated high media impact

• Proved single-pro focus > multi-pro directory