Shopify PDP research: YouTube corpus to ProSpikeFlow dev subdomain

Shopify PDP research: YouTube corpus to ProSpikeFlow dev subdomain

By Idalia Ward

For ProSpikeFlow (volleyball DTC on a DSC client dev subdomain), we did not start with a mood board. We started with a 56-video YouTube corpus on Shopify PDP and landing-page conversion, transcribed with yt-dlp + AssemblyAI and synthesized into a canonical report in Sanctum Tasks. This post covers the methodology and the first thing we built on the dev subdomain — not the full purchasable report.

Research → spec, not “watch a few videos”

The pipeline is deliberate:

  • Query design — practitioner-oriented searches (PDP teardowns, sticky ATC, bundle-first offers), excluding generic “start a Shopify store” content.
  • Transcription + timestamps — sidecar *.words.json so findings cite seconds in-source.
  • Consensus tagging — patterns tagged consensus (≥65% of 56 videos), mixed, or edge-case so we do not treat one loud creator as law.
  • PRD + Tasks — locked decisions (e.g. bundle-first offer card, Tasks #99 / #102) before anyone writes theme CSS.
  • Dev subdomain — implementation on prospikeflow.client.decisionsciencecorp.com (Ada DNS + multihost provisioning), not production client DNS until Mark signs visual milestone gates.

What the corpus converged on

Across 56 videos, a small set of patterns show up in roughly 65–98% of the corpus and became non-negotiable baseline for phase 1:

  • One dominant promise above the fold + one primary CTA
  • Mobile sticky add-to-cart on long PDPs (41/56 — consensus)
  • Social proof between offer and first decision point
  • FAQ / guarantee / shipping immediately before checkout
  • Block grammar: promise → mechanism → proof → offer → objections → repeated CTA

Phase 1 on the dev subdomain: sticky add-to-cart

On prospikeflow.client.decisionsciencecorp.com, the first implementation task after hero layout was not “more polish.” It was mobile sticky add-to-cart on long PDPs — because the corpus tags it as consensus, not an A/B afterthought.

Practitioners treating long PDPs as scroll-heavy funnels keep a persistent ATC control so interest captured mid-page does not require a scroll hunt. For ProSpikeFlow phase 1 (PRD §10 Phase A), we built sticky ATC before below-the-fold proof blocks. Those sections only convert when the user can act from anywhere on the page.

We did not invent conversion-rate percentages for this post. Telemetry on the client production domain stays behind PRD §6.1 gates. Dev-subdomain instrumentation is intentionally narrow and revocable: hero_view, primary_cta_click, sticky_cta_click, offer_card_view, addtocart_click, faq_open, checkout_start.

Sticky ATC sits inside the larger block grammar from the same report:

  • Hero: one promise + one CTA
  • Fast proof strip → product mechanism
  • Bundle-first offer card (Tasks #99 / #102) with single-unit path visible but muted
  • Richer social proof → FAQ + guarantee + shipping → CTA repeat

That sequence is what we are shipping on the dev subdomain — volleyball PDP patterns from YouTube corpus → Sanctum spec → Ada-provisioned host → engineering, not generic ecommerce fiction.

Why this matters for client work

YouTube is not the deliverable — the spec is. When we say “Shopify PDP research at scale,” we mean reproducible ingestion, consensus labeling, and a dev subdomain that implements the report — not three blog posts worth of opinions.

The full ProSpikeFlow funnel report (figures, footnotes, anti-patterns) is available as a purchasable research deliverable from Decision Science Corp. For build or CRO on your own PDP program, contact us here.