How EliteAI™ training works
When EliteAI™ learns from your orders, how long changes take to reach buyers, what each status means, and what to do if training fails.
Last updated
EliteAI™ learns which products to offer in a training run: it reads your orders and products from Shopify and works out the best picks for each product. This article explains when training runs, how long your changes take to reach buyers, and what to do if a run fails.
When EliteAI™ trains
EliteAI™ only trains while something uses it: a published flow with an EliteAI™ offer screen, a flow whose A/B test is running with EliteAI™ in Version B, or Place another order turned on, on either page. Until then, the app says Trains once a published flow uses EliteAI™.
Once something uses it, a training run starts:
- when you publish the flow, if EliteAI™ needs a run, for example the first time or for a new engine
- a few minutes after you save a change to an engine, to a rule set's rules, or to Exclude products by tag
- once a day, from 5 a.m. Central European Time, to pick up new orders and product changes
Each run reads up to the last 365 days of orders. The first run usually finishes within 15 minutes; a large store can take longer.
How long changes take
| Change | When buyers see it |
|---|---|
| A screen's flat discount, or a rule set's Default discount | As soon as you save |
| A rule set's discount rules | After the next training run, up to one hour |
| An engine's setup, or the excluded tags | After the next training run |
| New products, and changes to your products | Up to 24 hours |
While a run is on its way, buyers keep getting the picks and rules from the last successful run, so offers keep showing. The one exception is a rule set with rules that no run has trained yet, such as a new one: until a run trains it, its offers get no discount at all, not even its default discount. See Set discounts with rule sets.
Before the first training run
An EliteAI™ screen has nothing to offer until EliteAI™ has trained, so checkout skips it. You can still publish the flow; publishing is what starts the first run. Meanwhile, the flow's EliteAI™ card says: EliteAI™ has no picks yet, so checkout skips this offer until its first training run.
The same applies to a new engine: an offer using it is skipped until a run has trained it. Next to the engine select, the card names engines that are Not trained yet, and shows Training in progress while the chosen engine is in its first run.
What the status badges mean
The Upsell engines and Discount rules tabs show each engine's and rule set's Status:
| Status | Meaning |
|---|---|
| Ready | Trained as saved. A rule set without rules is always ready. |
| Not trained yet | New, and nothing uses EliteAI™ yet: no published flow, and Place another order is off. |
| Queued | New, waiting for the next training run. |
| Training | New, and the current run is training it. |
| Changes pending | Saved since its last training. Buyers get the last training's picks until the next run. |
| Training failed | The last run couldn't train it. |
While training runs normally, the EliteAI™ & discounts page shows no banner. It shows one only when something needs your attention.
If training fails
When a run fails, buyers keep seeing the picks from the previous training run. The EliteAI™ & discounts page shows a red banner starting The last training run failed, with the reason when there is one you can fix. Unless it's something you need to fix, it adds EliteAI™ will try again automatically.
Common reasons:
- No product can be offered. Every product is excluded, often by Exclude products by tag. Check that card on the Upsell engines tab.
- An engine couldn't be trained. The banner names it, and the engine shows Training failed. Open it: its page says why, for example that the last run failed for it, and asks you to check its rules and save it again. An engine with failed training can't be chosen on an offer until a run succeeds.
- The run ran out of time for an engine or rule set. It tries again in the next run; you don't need to do anything.
A rule set with Training failed works on: if it has rules, its offers get no discount until a training run succeeds, so pick another rule set or fix its rules. If it has no rules, its offers still get its default discount.
Still stuck?
Chat with us and we'll help you get your offer live. Or email contact@elitecart.app.