On-Site Recommendation Engine
Generic "you may also like" widgets convert at 1–2% because they show the same products to every visitor. ShopPulse builds an individual model per shopper, reading browse path, scroll depth, time-on-PDP, past purchases, and repeat frequency — then serves recommendations at under 80ms p95 on every surface where shoppers make decisions.
- Surfaces: PDP, category list, homepage hero, cart upsell, post-purchase
- Updates with every click — no cached results served
- Cold-start handling for first-time visitors
- Target lift: +12–22% on-site conversion
Triggered Abandoned-Cart Email
Most abandoned-cart emails go out 24 hours after drop-off with a static product image and a 10% discount code. ShopPulse fires a three-touchpoint sequence at 30 minutes, 4 hours, and 24 hours — each pulling the shopper's actual browse history and live inventory at send time. Connects to Klaviyo, Mailchimp, Sendgrid, and Iterable; no new ESP required.
- Three-touchpoint cadence: 30 min / 4 hr / 24 hr
- Per-recipient product selection from live catalog
- ESP connectors: Klaviyo, Mailchimp, Sendgrid, Iterable
- Target lift: +30–50% vs. batch abandoned-cart campaigns
Price-Sensitivity Triggering
Not every shopper who hesitates needs a discount — and over-discounting trains buyers to wait for deals. ShopPulse scores each shopper's price elasticity in real time using scroll depth, repeat visit count, time-on-PDP, and purchase frequency. Discount offers fire only when the signal says the margin trade is worth it.
- Signals: scroll depth, repeat visits, time-on-PDP, purchase history
- Threshold rules you control — set minimum margin floor before any offer fires
- Surfaces on PDP and cart, not homepage (avoids training all visitors to expect deals)
- Full revenue attribution per triggered discount
Built-in A/B Testing
Every recommendation surface and triggered email cadence can be tested against a control group. ShopPulse runs the split, tracks conversion and revenue attribution in real time, and flags statistical significance — no data team or separate testing tool required.
- Test recommendation placement vs. no recommendation
- Test email cadence timing and product selection logic
- Real-time revenue attribution per variant
- Auto-ship winning variant when significance threshold is met
