The Revlance Blog

Static merchandising rules work when your catalog fits on a single page. Here's what goes wrong when you have 10,000 SKUs and five rules that haven't been updated since Q3 last year.

Purchase history tells you what someone bought last time. Scroll depth and hover duration tell you what they're considering right now. That's the gap between reactive and real-time personalization.

Shopify's built-in recommendation engine is good for starter stores. Here's where it stops being enough — and what to look for in a third-party personalization layer.

Tile position, image thumbnail, price prominence, availability signal, and relevance to prior session. These five variables explain most of the variance in product grid CTR.

How to structure product recommendation blocks in Klaviyo flows so they show items the subscriber is likely to buy next — not just what they already bought.

Scroll depth alone is noise. Scroll depth combined with tile hover duration and session timing is signal. Here's how to read it.

When a shopper hovers over a product tile for more than 800ms, something is happening. Here's what our data says about how to use that signal without overweighting it.

Rules are fast to ship and easy to reason about. ML learns patterns you'd never write as rules. A practical framework for knowing which you need.

Homepage module tests take longer than product page tests and produce noisier results. Here's how to structure your hypothesis, set sample sizes, and avoid the common attribution traps.

Cross-sell works when it's contextual. It backfires when it's just 'frequently bought together' from the category bestseller list. The difference is individual behavioral context.

New visitor arrives. You know nothing. What do you show them? How quickly can you build a useful preference signal from the first 60 seconds of a session?

Step-by-step walkthrough of how Revlance behavioral profiles sync with Klaviyo custom properties, and how to use them to trigger personalized recommendation blocks in automated flows.

The conversion tactics that worked in 2021 are table stakes now. Here's what mid-market DTC brands are doing differently in 2025 — and what the data says about which changes move the needle.

Batch personalization runs nightly and serves stale signals. Real-time personalization updates within the session. For DTC with high inventory turnover, the difference can be 15+ percentage points in relevance score.

Black-box recommendations are a liability. When a VP of Growth asks 'why did the homepage show hiking boots to a shopper who browses kitchen?' — you should have an answer.

What you actually lose when third-party cookies deprecate — and why on-site behavioral personalization built on session-local signals is largely unaffected by the change.

How to measure incremental revenue from a personalization layer using holdout groups and incrementality testing — and why "influenced revenue" is the wrong number to bring to a finance review.

Fewer above-fold tiles, faster bounce decisions, and thumb-zone constraints mean mobile grid ranking needs different logic than desktop. Here's how tile position priority shifts.

When 40% of your SKUs change at a seasonal collection launch, behavioral signals trained on last season's catalog start pointing at products that no longer exist. Here's how to manage the transition.