"Run a sale" is one intent, and not a very precise one. The same strategy editor, group list and schedule can be pointed at three different jobs. Each one is a different set of settings, and knowing which job you are doing tells you what to look at afterwards.

1. Find out what sells, and at what price

You have a catalog and guesses. A run of personal catalogs is a cheap way to replace the guesses, because every customer sees a different sample of products at different prices, and the Products page afterwards shows what sold and the discount depth each product sold best at.

Settings for this job:

  • Strategy: wide Chance pools covering the collections you are curious about, with the Distribution setting on Even so stock levels do not bias the draw. A moderate range with Higher discounts on Even, so depths are spread rather than bunched at the floor.
  • Catalog size: toward the top of the range, so each customer samples more products.
  • Audience: a group of 50 to 200 engaged customers. Large enough to read, small enough that one order does not swing the result.
  • Run: one manual run; no automatic rule yet.

Read afterwards: the Products tab in Analytics (top products, best-selling options and tags, discount depth by product) and the Discount dependency chart. Then build the next strategy from what sold, not from what you expected to.

2. Clear inventory without a clearance sale

A clearance sale tells the whole store that these products are worth less. A Limited pool tells a set number of people that they can have one at a price, and nobody else hears.

Settings for this job:

  • Strategy: a Limited pool built from a collection of the stock you need gone, at a deeper fixed percentage, capped at the number of customers it would take to clear it. Keep the rest of the catalog on your normal pools so the page still looks like your store.
  • Distribution: Inventory, so the engine prefers products with more stock when it fills Chance slots.
  • Inventory rules: set the stock minimum in Settings → Inventory so a product leaves catalogs before it oversells, and choose Replace so the slot is refilled.
  • Campaign type: Storefront in Guest mode reaches everyone who visits without an email; an Outreach run to a group that bought from that collection before reaches the people most likely to want it.

Read afterwards: the collection's products on the Products page, and stock in Shopify.

3. An exclusive drop

A small number of people get something others cannot, for a short time. This is the one job a store-wide sale can never do, and it is where Randio's limits stop being guardrails and become the offer.

Settings for this job:

  • Strategy: one Guaranteed pool with the drop product, or a Limited pool capped at the number you are releasing, at whatever discount makes it a drop. Nothing else in the catalog, or very little, so the page is about one thing.
  • Audience: a group at the top of your group order, so it claims these customers before any other campaign can. Define it by what they have done: lifetime revenue, order count, recent engagement.
  • Run: a short manual run with a fixed window and the reminder email on, so the end date does the work. The countdown on the catalog page counts to the same end.
  • Limits: one unit per product per catalog, which is the default, so the drop is a drop.

Read afterwards: the campaign's own page. Who it reached, who opened, who bought, and whether the Limited pool ran out before the run ended. If it did, the next drop can be smaller or the group larger; if it did not, the group was wrong or the offer was.

Pick one per campaign

A campaign can only be read against one intent. A run that is half experiment and half clearance tells you nothing about either. Give each job its own strategy, its own group and its own run, and let the next run borrow what the last one learned. Choosing the groups that make the comparison fair is the subject of Segment before you randomize; the pool mechanics are in Designing a discount strategy.