Proof

Tracked through a permanent holdout.

Every result below is measured against a randomized holdout group that never sees personalization - not a running count of clicks and orders. The holdout is where these numbers come from, calculated once at the end of each client's contract term using the conservative, lower-bound end of the measured lift.

Start Your Free Pilot

Case studies

Three verticals, same method.

Supplements

National vitamin and supplement retailer

+29%: Repeat-buyer rate vs. control group
+13%: Average order value vs. control group
−9%: Bounce rate vs. control group

Beauty

Professional haircare brand

+27%: Conversion rate vs. control group
13x: ROI vs. control group
40: Developer hours saved First quarter

Apparel

B2B apparel wholesaler

+28-31%: Average order value vs. control group

The measurement

One comparison behind every number.

Every figure Fulcrum reports comes from the same randomized comparison.

A fixed 10% randomized holdout never sees personalization for the full length of your contract. Both groups shop the same site over the same window, so whether a shopper got Fulcrum's personalization is the only variable between them.

The difference in revenue per user between the two groups is calculated once, at the end of the 12-month term, at a 95% confidence level. Your guaranteed figure is the conservative lower bound of that range - not the raw average - and it qualifies when that lower bound is above zero.

Nothing Fulcrum reports claims it caused an order. The measurement says how personalized traffic performed against untreated traffic under identical conditions, not why any individual order happened.

Methodology

How the measurement works.

A permanent control group

A fixed 10% of your traffic never sees personalization, for the full length of your contract. We don't check in early or stop once a result looks favorable - the comparison runs for the entire 12-month term.

Significance thresholds, agreed up front

We calculate the lift once, at the end of the 12-month term, at a 95% confidence level, and use the conservative lower bound of that range rather than the raw number. A result counts as statistically significant once that lower bound is above zero. There's no early cutoff or interim checkpoint - it's the full term's data that gets measured.

Impression to conversion, joined

Both groups - holdout and personalized - see the same site, over the same time window, so the only variable between them is whether they got Fulcrum's personalization. That's what makes the comparison valid. See how Fulcrum measures impact .