Training: overview

What training means in Luria, the three inputs it learns from, and what to do first.

Luria ships with a pixel. The pixel doesn't just count, it learns. Every session teaches Luria who your buyers are, what language moves them, and which parts of your page they ignore. Training is the process of turning that raw traffic into a brand-specific model of what converts for you, not for stores in general.

That last part matters. Generic "best practices" are averages across thousands of stores that don't sell what you sell, to the people you sell to, in your voice. Training is how Luria stops using averages and starts using you.

Three inputs to training

Luria learns from three sources. One is automatic. Two need a little help from you.

1. Your traffic

Sessions, scroll depth, where visitors hesitate, where they leave, and which ones convert. This starts the moment your install is live. You don't have to configure anything; every visitor adds to what Luria knows about your store.

Traffic is the input that teaches Luria what actually works. Everything else is context. A headline can sound perfect and still lose to a plainer one on your visitors, and only your traffic can settle that.

2. Your brand

What you sell, who you sell to, how you talk, and what you refuse to say. You provide this once during onboarding, then refine it as things change. It takes about 15 minutes to do well the first time.

Without this, Luria can still test, but it tests in a generic voice. With it, every variant sounds like you wrote it, stays inside the claims you can legally make, and respects the offers you can and can't run. The How to train Luria page walks through exactly what to give it.

3. Your market

Luria studies the highest-performing public pages of the biggest players in your niche, plus adjacent niches, and imports the patterns that transfer. If the leaders in your category all lead with a guarantee above the fold, that's a pattern worth testing on your store. If a neighboring category converts on a specific kind of social proof, that's worth testing too.

Two things to be clear about:

This is public-page analysis only. Luria looks at what anyone can see by visiting a site. It never accesses anyone's private data, analytics, or accounts.
Luria never copies anyone's text. It learns structures, angles, and patterns, then writes original copy in your voice, inside your rules. Nothing a competitor wrote ends up on your page.

How the three fit together

Think of it as: your market suggests what to try, your brand decides how it can be said, and your traffic decides what stays. Luria proposes variants drawn from market patterns and your own copy, filters them through your brand rules, then lets your visitors vote with their behavior. Winners stay. Losers get retired. Over weeks, what's left is a page shaped specifically by your buyers.

What training is not

It is not a one-time setup. The first version of your brand profile gets Luria started; updates as you launch products or shift positioning keep it sharp.
It is not instant. Luria needs a real baseline before it tests anything, and real conversions before it calls a winner. The training timeline shows what to expect and when.
It is not a guarantee of a specific number. Results depend on traffic volume, goal clarity, and how much context you give it. What to expect is honest about this.

What to do next

Make sure your install is verified and conversions are tracking. Training runs on conversions; if those aren't counted, nothing else matters. See Verify your install.
Pick your goal. Revenue, margin, leads, booked calls: tell Luria what winning means so it optimizes toward the right thing. See Choosing your goal.
Spend 15 minutes on your brand profile. Paste real copy, real reviews, and your do-not-say list. See How to train Luria.
Then leave it alone through the baseline. Day 0 to 3 is measuring, not changing. Pausing or editing the page mid-baseline resets the clock.