1. Square meme graphic showing an ad performance dashboard with $1,074 spent, 8 installs, $134.25 cost per install, and a laughing clown saying “Optimization is going great. HONK HONK.” PHOTO Popsmart Clerk Ad Performance Clown Meme Square meme graphic showing an ad performance dashboard with $1,074 spent, 8 installs, $134.25 cost per install, and a laughing clown saying “Optimization is going great. HONK HONK.”

The uncomfortable signal

For the last month I have been pushing two Shopify apps in parallel: PopSmart Studio and PopSmart Clerk. Studio is the product for AI product visuals. Clerk is the AI store-clerk concept for product pages.

The Clerk signal is the one I cannot ignore: Shopify App Store installs are still 0.

That usually does not mean nobody clicked. It means the people who clicked did not see their current job-to-be-done strongly enough to install. Paid traffic can create attention. It cannot manufacture need.

What I tried before calling it a pivot

Before changing the product direction, I tried the obvious listing-side fixes:

  1. a public demo store
  2. listing keyword and copy changes
  3. integration angles around Sidekick, Shopify Flow, and better-known apps

Those were worth doing because they remove excuses. If the listing is unclear, fix the listing. If trust is missing, show a demo. If workflow fit is unclear, make the integrations visible.

But if all of that still does not move install intent, the issue is probably not polish. It is product demand.

The read

The current read is simple: the “AI clerk” framing may be interesting, but not urgent enough for the merchants I am reaching.

Clicks without installs are a useful negative signal. They point to curiosity, not purchase intent.

So I am going to pivot Clerk toward the functional demand that shows up more clearly: less character, less assistant theater, more concrete merchant workflow value.

What happens next

Two tracks now:

  1. finish the listing cleanup — keywords, screenshots, images, and video — so the storefront does not leak demand I actually have
  2. segment the 40,000 Shopify stores I collected and send targeted cold email to the stores where the use case should be relevant

This should not be a blind blast. The list needs filtering by category, storefront maturity, product-page complexity, and visible merchandising pain.

The goal is to learn whether the rewritten problem statement creates replies, not just traffic.

Why I am posting this

The useful part of building in public is not showing a perfect graph. It is showing the moment where the graph tells you to stop defending the idea.

This is that point for Clerk.

The next version has to earn intent before it earns more engineering time.