5.1 Measure
Numbers you can take into a meeting.
Page, variant, language, and site cut from the same filtered traffic - before anyone questions the number.
Report real visitors first
Bots, datacentres, and instrumentation noise are excluded before KPIs are calculated.
Same definitions, four cuts
Page, variant, language, and site views share one metric model so numbers reconcile.
1.0
What each metric actually counts
Unique visitors, impressions, conversions, conversion rate, and time on page - defined the same way in every view.
Rate from people, not hits
Conversion rate is conversions divided by unique visitors, not by impressions.
Invalid traffic never enters the number
Bots and junk are filtered before reporting, so meetings start from a defendable baseline.
Distinct people in the period, deduplicated across sessions and devices where identified.
Page renders served, including repeat views by the same visitor.
Completions of the goal event you define per page, counted once per visitor per period.
Conversions divided by unique visitors, not by impressions.
Median-trimmed active time, excluding backgrounded tabs.
2.0
Real traffic, not raw traffic
Roughly a fifth of landing page traffic in paid channels is automated or invalid. Reporting it inflates impressions, deflates conversion rate, and quietly corrupts every test on the page.
BeforeRaw visitors
Unfiltered231,447Same day, same page
After sieveReal visitors
LiftLander184,092Invalid traffic removed before KPIs
Filter rulesDropped before KPIs
- 01Declared crawlersDrop
They never convert, so they only dilute rate.
- 02Datacentre and proxy rangesDrop
Not consumer traffic.
- 03Headless browsersDrop
Automation, scraping, and click fraud.
- 04Sub-second bounce with no paintDrop
Pre-fetch, not a visit.
- 05Duplicate impressions within one secondDrop
Instrumentation noise.
3.0
Traffic and conversion over time
Every chart carries its comparison against the previous period. Test results feed straight into the same reporting layer.
Period comparison built in
Trend charts show the previous period alongside the current one so lift is obvious without a spreadsheet.
Tests and analytics share one layer
Variant readouts and page KPIs use the same filtered traffic definitions.
4.0
Cut the same data four ways
Per page, per variant, per language, per site - plus export when the number needs to leave the product.
| Page | Real visitors | Conversions | Rate |
|---|---|---|---|
spring-offer | 84,120 | 5,402 | 6.42% |
pricing | 42,880 | 1,914 | 4.46% |
welcome-bonus | 31,204 | 1,608 | 5.15% |
CSV on every view, scheduled exports, and a read API on Scale and above. Filtering rules are documented so your warehouse can reproduce the same numbers.
See data handlingFAQ
Analytics questions
For page and variant performance, yes. For whole-journey attribution across your site and apps, keep GA4 and send LiftLander events into it.
A visitor that passes bot signature, datacentre IP, headless browser and behavioural checks. Everything else is retained as raw but excluded from reported metrics.
Yes. Every view can be toggled between real and raw, and the difference is shown explicitly rather than hidden.
Twenty-five months on all plans, which covers year-over-year comparison. Enterprise can extend it.
On Scale and Enterprise. Read endpoints cover pages, variants, languages and sites with the same filtering rules as the interface.
No. Assignment and counting work server-side with a first-party identifier, so consent-declining traffic is still measured in aggregate without personal data.
Report numbers you can defend
Free for 14 days, no card required. Larger teams can book a walkthrough instead.