Training timeline: day 0 to day 90
What Luria is doing at each stage of training, what you'll see in the dashboard, and how traffic volume changes the clock.
Training moves through five stages. The day ranges below assume roughly 3,000 or more sessions per month on the pages Luria is testing. Lower-traffic stores stretch proportionally: a store with 1,000 sessions a month should expect each stage to take about three times as long. Luria moves on conversions, not on the calendar.
| Stage | When | What Luria is doing | What you see |
|---|---|---|---|
| Baseline | Day 0 to 3 | Measuring your current page, learning your visitors, building the brand profile | Live traffic, heatmaps, session recordings, no changes to your site |
| First tests | Day 3 to 10 | Serving its first variants against your original | First measurable movement, usually a headline or offer framing win |
| Learning | Day 10 to 30 | Testing bigger swings, narrowing to what works on your traffic | Lift stabilizes, a clear winner emerges |
| Converged | Day 30 to 60 | Locks in winners, splits by traffic source | Per-channel results, different pages for different ad sets |
| Compounding | Day 60+ | Personalization, seasonality, offer and price testing | Continuous small gains on top of the locked-in lift |
Baseline: day 0 to 3
baselineDay 0 to 3Luria watches and changes nothing. It measures how your current page performs: conversion rate, where visitors scroll, where they hesitate, where they leave. It also finishes building your brand profile from what you entered in the Train tab.
You see live traffic, heatmaps, and session recordings in the dashboard. You don't see variants yet, and that's deliberate.
Don't pause Luria, redesign the page, or launch a promo during the baseline. Any of those resets the measurement, and Luria starts counting again from zero.
Why day 3 and not day 1
Luria needs a real baseline before it can prove anything. If it started serving variants on day 1, it could show you a number, but that number would have nothing to compare against. A number without a baseline is a guess, and we won't show you guesses. Three days (at typical traffic) is the minimum to capture a weekday and a weekend pattern and enough conversions to mean something.
First tests: day 3 to 10
first testsDay 3 to 10Luria serves its first variants alongside your original page. These are usually the safest, highest-probability changes: headline framing, the way the offer is stated, the order of benefits. Some visitors see the original, some see a variant, and Luria compares.
Most stores see their first measurable movement here. Usually it's a headline or offer framing win. It will be modest and the number will wobble day to day. That's normal; small samples are noisy. See What to expect for how to read early results.
Learning: day 10 to 30
learningDay 10 to 30With the easy wins banked, Luria tests bigger swings: different angles, different proof, different page structure. It also starts narrowing. Variants that clearly lose on your traffic get retired; variants that clearly win get more visitors.
By the end of this stage, lift stabilizes and a clear winner usually emerges. This is the first point where the number in your dashboard is one Luria will stand behind. See Lift and significance for what "stand behind" means.
Converged: day 30 to 60
convergedDay 30 to 60Luria locks in the winners and starts splitting by traffic source. Visitors from a Meta ad, a Google search, and a Klaviyo email arrive with different intent, and the same page doesn't serve all three best. Luria begins testing different versions for different channels, then for different ad sets.
You see per-channel results in the dashboard, and you can see which page a given ad set is landing on. This is where connecting your ad accounts pays off; see Traffic sources and Ad-level pages.
Compounding: day 60+
compoundingDay 60+The locked-in lift is your new floor. On top of it, Luria keeps testing in smaller, more targeted ways: personalization by visitor type, seasonal framing as your calendar shifts, and (if you've set the rules in How to train Luria) offer and price testing.
Gains here are continuous and small rather than dramatic. That's the point. A store that has been with Luria for six months has a page, and a Brand Brain, that no two-week test could produce.
If you're below 3,000 sessions a month
Everything above still happens, in the same order. It just takes longer, because each stage needs a certain number of conversions to clear, not a certain number of days. A rough rule: divide 3,000 by your monthly sessions and multiply each stage's length by that.
Ways to shorten it: concentrate traffic on fewer pages, make sure every conversion is tracked, and avoid pausing. Getting better results faster covers all three.