> ## Documentation Index
> Fetch the complete documentation index at: https://docs.nextintent.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Held back to compare

> Why NextIntent leaves a random share of stuck shoppers alone, how the share is chosen, the 20% ramp on new stores, and how the comparison is shown on Money.

In active mode, some stuck shoppers are never spoken to, on purpose and at random. They are the comparison group. Without them, you would know how many spoken-to shoppers bought but not how many would have bought anyway. With them, Money can show the difference.

## How it works

1. A shopper stalls and NextIntent decides to speak.
2. Before the sentence goes out, the shopper is assigned to a group. The assignment is random, sticks to the shopper for the visit, and has nothing to do with what is in the cart or what they did.
3. Shoppers in the "held back" group are left alone. Their moment is recorded on **Moments** as "Held back to compare", with the sentence they would have seen.
4. Ten and thirty minutes later, and again when an order arrives, the app records what each shopper did next. **Money** compares the two groups.

## The share

**Where:** **Settings → What NextIntent may do → Held back**. Choose 5%, 10% or 20%. The default is 10%.

**New stores start at 20%** until 30 shoppers have been held back, whatever the setting. The comparison needs a minimum number in each group before it means anything (Money waits for 30 in both), and at 5% that would take a long time on a smaller store. Once 30 have been held, the setting takes over.

**Which to choose:**

* **5%.** Once the comparison is established and you want the most shoppers spoken to. On 1,000 stuck shoppers a month, 50 are held back.
* **10%.** The default. A steady comparison without giving up much.
* **20%.** When you want the comparison to firm up fast, for example after a big change to your facts or your store.

## What you see on Money

Money shows two rates side by side once both groups have 30 shoppers:

* **Spoke to:** of the stuck shoppers who saw a sentence, the share that bought within seven days.
* **Left alone:** of the stuck shoppers held back, the share that bought on their own.

And the difference between them, which is the part of the credited revenue the comparison says you would not have had without the sentence. Until both groups reach 30, Money shows "Until then: who bought on their own" with the numbers so far.

<Info>
  The comparison is proof, not the bill. You are charged on every order credited to a moment (a sentence was shown, the shopper ordered within seven days), in full. The held-back share is never subtracted from what you are charged, and never added to it. See [Pricing](/shopify/billing/pricing).
</Info>

## Why hold anyone back

Every cart-recovery app claims the sales it touched. Merchants in the category's reviews do not believe the claims, and they are right not to: a shopper who saw a message and bought might have bought anyway. The only honest way to know is to leave some shoppers alone and count. NextIntent builds that in and shows you the count. It costs you a few sales from the held-back group; it gives you a number you can trust about all the rest.

## Held back is not the same as watch mode

In watch mode everyone is left alone and nothing is compared, because there is no spoken-to group. In active mode the held-back share is the control group for the shoppers who were spoken to. Both are recorded on Moments, with different labels: "Left alone, nothing shown" for watch mode, "Held back to compare" for the control group.

## Related

* [Money](/shopify/reports/money)
* [Held back share](/shopify/settings/held-back-share)
* [Watch mode](/shopify/how-it-works/watch-mode)
* [What "spoke to" means](/shopify/how-it-works/spoke-to)