Ladders Nobody Had to Build: Smart Upsell Ladders at Scale

Introduction

One upgrade campaign is easy to build by hand. Six hundred products' worth is not.

When a catalog is small, the merchant can name the premium sibling of every product and type it into a campaign. When the catalog is wide and changes weekly, that map is out of date before it is finished.

The instinct in most planning meetings is to start a spreadsheet:

"Let's list every product and its upgrade, then build it in."

A solutions architect asks what the spreadsheet is really encoding:

"Is each row a judgment call, or is it the same rule applied 640 times?"

If it is the same rule, the rule should do the work. This case study designs an upgrade program for a cookware brand whose catalog was too wide to hand-pick an upgrade for every pan.

Merchant Scenario

Consider Atlas Kitchenware, a fictional Shopify brand selling pots, pans, knives and small appliances.

Almost every product had a premium sibling: same type, higher price, better margin. Nobody had time to map 640 of them.

The Spreadsheet That Never Finished

The team started a spreadsheet with three columns: product, upgrade, difference. It got to 90 rows and stalled. Products launched and retired faster than the sheet could follow, and every new launch reopened rows that were already "done."

Business Objectives

Evaluating the Options

Option 1: Hand-Built Manual Picks

Option 2: Smart Ladders, Left Fully Automatic

Option 3: Smart Ladders With Pins and Exclusions

Comparison Matrix

The Discount Prime Configuration

A wide trigger list. Smart ladders as the offer source. A small incentive to make the first upgrade easy.

The 5% was small, but enough to make the banner say "only $X more" with a slightly smaller X. Like every Upsell incentive, it applied only to lines swapped by the widget, so a premium pan bought directly kept its full price.

Atlas tested 0% and 5%. Five percent converted a third better and cost less than the margin gain.

Why Smart Ladders Fit Upsell at Scale

Smart ladders is the Upsell offer source where rules find the upgrade. For each trigger, candidates must be the same product type, priced higher, and carry a better margin, using variant costs where they exist. The engine ranks them, and the live preview shows the top four per trigger with the price difference and whether cost is known.

It is an Ultimate feature. On Prime the tile opens with a lock, the preset still applies, and validation explains the gate before activation.

The team pinned a few picks where they disagreed with the ranking and excluded two discontinued lines. Everything else, the rules handled.

Downsells and Weak Margins, Filtered by Rule

A hand-built list of 640 rows would have had mistakes, including an upgrade priced below its trigger somewhere in the middle. Smart ladders only ever proposes candidates priced above the trigger. The downsell check that appears as a warning in The Derby and the Loafer is built into how Atlas' candidates were chosen.

"Better margin" was a filter, not a hope. A premium pan whose cost had crept up, leaving it with worse margin than the entry pan, did not appear as an upgrade.

Items with unknown cost were flagged in the preview so the team could decide, rather than being silently included or excluded.

Ladders That Updated Themselves

A new premium sauté pan launched in month two. Nobody edited the campaign. The next time the engine ran, the pan became a candidate for the entry sauté pan, ranked by price and margin, and appeared in the banner. An old pan was retired, and its ladder closed on its own.

Customer Journey

1. A shopper opens an entry-level sauté pan from the trigger list.
2. Below Add to cart, a banner offers the top-ranked premium pan, with {diff} showing the real gap after the 5% incentive.
3. They add the entry pan anyway and go to the cart.
4. The cart swap card repeats the offer. They tap Switch, the entry line is replaced by the premium pan, and the 5% applies to the tagged line.

What Smart Ladders Cannot Do

The rules know type, price and margin. They do not know that the entry pan is non-stick and the premium one is carbon steel, which some customers would consider a different product, not an upgrade. The team pinned and excluded around those cases. The engine proposed. Humans adjusted.

The trigger list is still specific products, because rule triggers are not in Upsell yet. 180 triggers was a one-time selection that needs a refresh when new entry-level products launch.

Metrics That Measure Success

• Upgrade rate: swapped orders as a share of eligible cookware orders
• Total 5% incentive paid against the margin gained on upgraded lines
• Count of triggers with no candidate, or only unknown-cost candidates, in the preview
• Undo rate after a swap, by product type, to spot "different product" pairings
• Estimated Profit per cookware order before and after launch

Common Mistakes

• ❌ Hand-building upgrade maps for a catalog that changes weekly, and shipping a stale one.
• ❌ Treating Smart ladders as fully automatic and never reviewing the preview for upgrades that are technically higher-priced but practically different products.
• ❌ Ignoring the "cost unknown" flag and letting a low-margin item be offered as an upgrade.
• ❌ Setting a large incentive on a smart campaign across hundreds of products without modeling the aggregate cost.
• ❌ Forgetting the trigger list is specific products, and not refreshing it when new entry-level items launch.

Key Lessons

Scale changes the question. For one product, the question is "what is the upgrade." For 640, it is "what is the rule for finding one."

Smart ladders answered the second. Pinning and exclusions kept human judgment where it mattered.

Conclusion

A spreadsheet is a snapshot. A rule is a process.

Upsell with Smart ladders turns "same type, higher price, better margin" into a living upgrade map that follows the catalog. In this scenario, Atlas' upgrade rate across cookware went from effectively zero to 17% of eligible orders. Nobody finished the spreadsheet. Nobody needed to.