Measurement
Did the new copy work? Compare it with the people who never saw it.
Every Lens site can keep a control group. They see your original page while everyone else from the campaign sees the campaign version, over the same days, from the same ads.
Why a holdback beats a before-and-after
Before-and-after
The usual way to judge a copy change is to compare this month with last month. That comparison mixes your change with everything else that moved: budgets, seasons, a competitor going quiet, a sale ending.
Holdback
A holdback removes most of that. Lens assigns each campaign visitor to one of two groups when they first arrive. The control group keeps seeing the original page. Both groups arrive from the same ads at the same time, so the copy is the main thing that differs between them.
You set the holdback rate for the site, anywhere from 0 to 20% of campaign visitors. A bigger control group gives you a steadier baseline sooner. A smaller one shows more visitors the campaign copy.
Original page
Campaign version
Holdback rate 0 to 20% of campaign visitors
Smaller holdbackMore visitors see the campaign copy.
Bigger holdbackA steadier baseline sooner.
Assigned once, kept in the same group
Assignment is decided from a random visitor id stored in the visitor's browser, so a person who comes back sees the same version they saw the first time. Visitors who arrive without campaign UTMs are not part of the test at all: they see your original page.
First visit
A campaign visitor arrives from your ad.
Random visitor id
Stored in their browser. It picks their group.
Same version on return
They see what they saw the first time.
No UTMs
Original page, not in the test.
Conversions, counted the same way for both groups
A comparison only works if both groups are measured with the same ruler, so Lens counts conversions on your page rather than trusting each ad platform's own numbers. It recognizes common conversions without setup: form submissions, clicks on calls to action and clicks through to tools such as Calendly, HubSpot or Intercom. For anything else, one call does it:
window.__lensInstance.trackConversion('purchase')Conversion types are grouped by the kind of site you run, because not every site sells something:
Lead capture
Contact forms, demo and quote requests, callback requests.
Scheduling
Appointments and consultations booked, including clicks through to tools like Calendly.
SaaS
Signups, trial starts and upgrades.
E-commerce
Add to cart, checkout started, purchases.
Content
Newsletter signups, webinar registrations, social shares.
Engagement
CTA clicks, downloads, scroll depth and time on page, for when the main conversion is too rare to read yet.
How the report reads
Control on the left bar of each pair, the campaign version on the right, and the overall lift below.
Campaign: Healthcare Compliance
| Metric | Control (Original) | Personalized |
|---|---|---|
| Sign-ups | 2.1% | 3.4% |
| Demo Requests | 1.8% | 2.9% |
| CTA Clicks | 4.5% | 6.8% |
| Avg. Time on Page | 12s | 28s |
What the numbers can and cannot tell you
Can
Lift is the difference between the two groups' conversion rates, as a share of the control group's rate. The dashboard shows the rates, the visitor counts behind them and the lift for each campaign once its control group has visitors.
Can't
Lens shows those numbers as they are. It does not run a significance test for you, so read a small sample with care: with a few dozen visitors in the control group, one or two conversions either way can swing the lift a long way. Give a campaign enough traffic, and enough days, before you act on its number.
The holdback also tells you when the campaign copy is not helping.
If the control group converts as well or better, keep the original and try a different angle.
Put a control group on your next campaign.
Set the holdback rate, publish one campaign and let the two groups run side by side.