Live Demand

The new-user day-7 survey was the starting point. Users reported not knowing where to look and wanting personalized recommendations. I used those findings to focus on how people found content relevant to them.

Cost was also a major concern. Live Demand focused on one part of that feedback: helping people find and choose relevant content.

I also reviewed purchase records. Active content lines had grown to 42, while the share of sales from new releases had fallen to 7%. Top-selling products showed customers repeatedly paying to check a changing situation.

Questions such as what someone thinks now or how today might unfold gave people a reason to return. We focused on helping customers find a relevant product when those questions arose.

active content lines

42

sales share from new releases

7%

weekly repeat purchases per buyer for a top product

2.13×

weekly repeat purchases per buyer for a real-time product

3.39×

I initiated the shift from on-demand content to Live Demand by combining sales patterns with new-user survey insights. Using Claude, I authored the high-level proposal and decided product structure, pricing, benefits and UX direction.

I considered new users' discovery difficulties alongside existing customers' repeat purchases. Instead of asking people to browse a premium catalog, we would ask about their current concern and guide them to relevant products.

Discovery converted the first Today post into a conversational entry point. Purchase connected each response to a concern category and product shelf, then tracked purchases made within five minutes of a product-detail visit. Repurchase activation turned visits to 12 event pages into a CRM re-entry point through a one-time in-app message and a one-day benefit.

The conversational post and CRM re-entry structure shipped as Phases 1 and 2. Phase 3 was designed as a follow-up expansion that would turn preset daily questions into a separate monetization layer.

shipped scope

Phases 1 and 2 launched

weekly concern categories

4

post-attributed purchase window

Within 5 min

event pages · message · benefit

12 · 1 · 1 day

Initially, concerns were shown as text for users to choose from. From Week 9, those same concerns were presented as image cards. Selecting a card led to follow-up questions and relevant products.

After choosing a card, users answered a more specific question and moved to a product list matched to that context. Content, questions, and product detail worked as one resolution flow rather than separate screens.

We defined a post-attributed purchase as one made within five minutes of a product-detail visit from the post. This definition let us track weekly how content discovery moved into purchase behavior.

I opposed excluding products already visible on other shelves. Operational separation mattered less than showing every product relevant to the four concerns a user could choose.

After expanding the eligible pool in Week 8, post-attributed purchases rose from 93 to 199 and attributed force points from 70,410 to 180,600. This established contextual fit—not shelf exclusivity—as the operating principle for future assortments.

Planned placement was missing in Weeks 3–7, so exposure differed across weeks. These figures do not isolate the effect of the assortment change.

post-attributed purchases

93 → 199

attributed force points

70,410 → 180,600

In Week 9, we changed the text-based concern choices to image cards while keeping the conversational flow into relevant products.

Following the update, user touch rate rose from 5.16% to 23.69%, while post-attributed purchases increased from 199 to 660. The share of recommended-product sales attributed to the post grew from 5.0% to 33.8%. Product assortment and placement also changed, so these comparisons do not isolate the effect of the cards.

user touch rate

5.16% → 23.69%

users who touched the post

3,130 → 15,879

post-attributed purchases

199 → 660

share of sales attributed to the post

5.0% → 33.8%

To extend the intent signal created in Phase 1, customers who entered any of 12 event pages during a defined period received a one-time in-app message and benefit.

The benefit expired one day after issue, bringing customers back while interest was still active. The outcome of this stage was the operating structure itself: a CRM re-entry point designed to activate another purchase, rather than a measured increase in repurchase rate.

We recorded placement, product assortment, and post format together, then compared touchpoint metrics before and after each weekly change.

Live Demand established an operating loop connecting a current customer concern to a product entry point, contextual shelf, and CRM. Discovery and purchase touchpoints were measured weekly; message response and incremental repurchase remained the next measurement layer. Each week's result informed the next format, assortment, and placement decision.