# Getting better results faster
Source: https://docs.luriart.com/train/improve-training
Plain text: https://docs.luriart.com/train/improve-training.txt

The practical levers that speed up training and raise the ceiling on results, and the habits that slow it down.

Luria trains on evidence. Everything on this page either gives it more evidence, cleaner evidence, or better material to test. None of it requires technical work; most of it is ten minutes in the dashboard.

## Add customer reviews and support language
If you do one thing from this page, do this. Paste real reviews, support tickets, and DMs into the Train tab.  Your customers' exact words are the raw material for the variants most likely to win, and Luria can't invent them. Twenty reviews is a good start; a hundred is better. Include the critical ones.

See [How to train Luria](/train/how-to-train) for what else belongs there.

## Tighten the do-not-say list
Every variant Luria would have served and then retired for being off-brand is a wasted test. A complete do-not-say list (banned claims, competitor names, phrasing you hate) means Luria only spends traffic on variants that could actually win. Spend five minutes on it. Add to it whenever a variant makes you wince.

## Pick one primary goal and keep it
Changing the goal re-baselines. Splitting attention across several goals slows every decision. Pick the one closest to money, confirm it's tracked, and leave it. [Choosing your goal](/train/goals) has the decision list.

## Concentrate traffic, don't spread it
Luria needs conversions per page, not conversions per store. A store with 3,000 sessions spread across twenty landing pages trains slower than a store with 3,000 sessions on two.

If you run ads, point them at the pages Luria is testing. If you have several near-identical landing pages, consolidate. You can always widen later; early on, depth beats breadth.

## Separate your traffic sources cleanly
Luria splits results by channel and, later, by ad set. It can only do that if it can tell the channels apart. Make sure your ad links carry UTM parameters (utm\_source, utm\_medium, utm\_campaign at minimum) and that you haven't got one campaign tagged three different ways. Meta, Google, and TikTok can all add these automatically.

Clean tagging is what unlocks per-channel pages and ad-level pages later. See [Traffic sources](/optimize/traffic-sources) and [Ad-level pages](/optimize/ad-level-pages). Connecting your ad accounts directly makes it cleaner still; see [Meta Ads](/connect/meta-ads) and [Google Ads](/connect/google-ads).

## Tell Luria before promos and launches
A sale changes what converts. If Luria doesn't know one is coming, it sees a sudden shift in behavior and has to figure out why. If you tell it the day before (a one-line note in the Train tab is enough ), it adjusts on purpose instead of by surprise, and it doesn't misread the post-sale dip as a losing variant.

Same for launches, price changes, and new audiences. Before, not after.

## Don't pause during the baseline
The first three days (at typical traffic) are measurement. Pausing, editing the page, or swapping the theme during that window resets the measurement, and Luria starts from zero. If you need to make a site change, make it before you switch Luria on, or wait until the baseline is done. See [Training timeline](/train/timeline).

## Give it the whole funnel
Luria optimizes toward the conversion you track. If you only track purchases but half your value comes from email signups that convert later, Luria is optimizing blind to half your funnel. If your thank-you page doesn't fire a conversion at all, Luria is optimizing toward nothing.

Check that every step that matters is counted: purchase, lead, booking, signup, whatever your goal is. On Shopify, purchases are automatic. Everywhere else, confirm the conversion fires on the confirmation page. [Verify your install](/install/verify-install) walks through it.

## Keep the brand profile current
A 60-second update when something changes keeps Luria working from reality. New product, new positioning, new audience: add a line. Stale context doesn't break training, but it does mean Luria spends a week testing around a gap you could have closed in a minute.

## What slows training down
* **Thin or scattered traffic.** Fewer conversions per page means slower decisions. Concentrate.
* **Untracked conversions.** Luria can't learn from a sale it can't see.
* **Changing the goal.** Each change re-baselines.
* **Pausing mid-baseline or mid-test.** Evidence resets.
* **Editing the page while Luria is testing it.** Luria is comparing variants against your original; if the original moves, the comparison breaks.
* **Surprise promos.** Behavior shifts Luria has to decode instead of anticipate.
* **Empty customer-language section.** Luria tests from generic material instead of your buyers' own words.
* **Messy UTMs.** Channels blur together and per-channel results can't separate.
* **Several goals at once.** No clear definition of winning.

## The short version
Upload reviews. Pick one goal. Concentrate traffic. Tag your links. Tell Luria before anything changes. Don't pause. Track every conversion. Do those and you're in the fast lane.

## Next steps
  * [How to train Luria](/train/how-to-train)
  * [Traffic sources](/optimize/traffic-sources)
  * [Ad-level pages](/optimize/ad-level-pages)
  * [Verify your install](/install/verify-install)