Merchandiser blog

Rolling Back a Bad Sort: Version History for Merchandising Teams

2026-08-21T08:00:00Z

It is Monday morning and the main category looks wrong. The campaign products are gone from the top, and something from last season is in first place. Nobody knows what happened. Somebody starts dragging products back into place from memory.

This is the situation version history exists for. It is rarely a fault of the sorting logic. Almost always, a person changed something.

How a sort goes wrong

The usual causes are ordinary:

A rule was switched. Someone tried a different sort type on a collection and saved it.
Pins were cleared. "Unpin all" was meant for another collection.
The wrong pin list was uploaded. A file for last week's campaign, or for a different category.
A setting changed. The auto-unpin period was shortened and released a campaign's pins early.
Many collections were updated at once, and one of them should not have been.

None of these is exotic, and each takes seconds to do. What takes long is the repair, when nobody knows what the state was before.

What a version has to contain

An undo that restores only the order is not enough. If the rule was changed, the next scheduled sort applies the new rule again and undoes the repair. A useful version holds the whole state of the collection:

the order of its products,
the pins, with their positions,
the rules: the sort type, the sort frequency and the auto-unpin settings.

In Merchandiser a version with exactly these is saved whenever someone changes a collection in the panel: saving an order on the sort page, uploading a pin list, changing the settings, updating many collections at once, or rolling back.

Scheduled re-sorts are not stored one by one. They follow the rules, and the rules are what the versions keep.

Finding the change

With versions, Monday morning starts differently. The version list of the collection shows each change with its author, its time and a summary of what it altered: how many products moved, how many pins changed, which rule changed.

The version's own page has the detail: products added and removed, and for each changed rule the old and the new value side by side. Usually the cause is visible within a minute. "Sort type changed on Friday at 18:40" answers the question.

Rolling back

Choose the last version that was right and roll back to it. The order, the pins and the rules return to what they were in that version.

A few things are worth knowing:

New products are kept. Products that joined the collection after that version are placed at the end, unpinned. Nothing is removed.
The rollback is a version too. If it turns out to be the wrong call, it can be undone the same way.
The store follows. The restored order is sent to the store like any other change.
Automatic sorting continues. From then on the collection follows the restored rules. If they sort automatically, the next scheduled sort applies them to today's data, with the restored pins in place.

Rolling back is reserved for managers. Everyone on the team can open the version list and see what changed.

A second pair of eyes

Rollback fixes a mistake after it has been seen. A review helps it get seen.

In Merchandiser a manager can approve or reject a saved version and leave a comment. The change is already live at that point; the review does not hold it back. That is a deliberate trade: the team keeps working at full speed, and important categories still get a second look, with a rollback one click away when the reviewer disagrees.

Approved versions have one more use. Older versions are removed over time, but approved ones are kept, so a state that someone signed off can always be restored.

Habits that make it work

Preview before saving. The sort page shows the order a different rule would give, and nothing changes until you save.
Review the big categories. Have a manager look at changes to the collections that carry most of the traffic.
Approve the states you would want back. The version before a campaign starts is a good one.
Check the change log when something looks odd. Next to the versions, each collection lists the changes of its sort type, frequency, pins and auto-unpin setting, with the user and the time.
Roll back first, investigate second. Restore the page, then find out what happened.

For how pins get out of hand in the first place, see Pinned Products: When to Pin and When to Let Go.

Brand Diversity on the Shelf: Why One Brand Shouldn't Own the Page

2026-08-14T08:00:00Z

A category called "Running Shoes" promises a choice. If the first three rows are one brand, the page keeps a different promise: it has become that brand's page under another name. Visitors who came to compare do not find a comparison, and the other brands you stock might as well not be there.

How it happens

Nobody decides to give a category to one brand. It follows from sorting by performance.

A strong brand sells well. Its products rank high. High positions bring views, views bring sales, and sales confirm the positions. Each product earns its place fairly, and the sum is a page with one logo on it. The better your sorting follows sales, the more it tends this way.

The assortment can add to it. If one brand supplies half the products in a category, even a random order would show it everywhere.

Measure it

The simple measure is the share of the collection that belongs to its largest brand. Count the products of each brand, take the biggest count, divide by the number of products.

Two cases should be left out:

Brand collections. A brand page has one brand by design.
Small collections. With ten products there is nothing meaningful to say about variety.

This is the Product Diversity signal of Merchandiser's Collection Health Score. It names the largest brand and its share of the assortment, skips brand pages and small collections, and lowers the collection's score as that share grows. The recommendation that comes with it is to cap repetition from a single brand to keep the shelf varied.

Merchandiser reports this; it does not enforce it. It will not hold back a product because the brand already has three in the row. What to do with the finding is your decision, and the options are below.

Look at the first rows too

The share across the whole collection is the stable number. What the visitor experiences is the first screen. A brand with a fifth of the assortment can still fill the first row if its products are the top five performers.

So when the signal flags a collection, open the sort page and look at the top. In Merchandiser the sort page can be set to the same number of products per row as your storefront, which shows the page as customers see it.

What to do about it

Start with the assortment. If one brand is most of the category, the page is telling the truth about what you stock. Sorting cannot create variety that is not there. Either widen the range or accept that this category is effectively a brand page.

Then look at the rule. A sorting rule can use brand as an attribute. A rule can give the other brands of a category a modest lift, or the dominant brand a modest reduction, so that the order still follows performance and the first rows mix. Keep the adjustment small: the goal is a page that offers a choice, not one that hides what sells.

Use pins for the first row, sparingly. Pinning one strong product of a second and a third brand into the first row guarantees variety where it matters most. This is curation, and it needs the upkeep pins always need; see Pinned Products: When to Pin and When to Let Go.

Split the category. If customers think in brands in this category, give them brand pages and let the mixed page be the overview.

When dominance is fine

Variety is not a goal in itself. Some categories are one brand's territory, and customers come for that brand. If the dominant brand is what people search for and buy, forcing others into the first row costs sales.

The signal is a prompt to look, and the test is the customer's intention. Did they come to choose between brands? Then show them a choice. Did they come for the brand? Then the page is doing its job.

Where it fits

Brand diversity is one of seven signals of a category page's condition. The others, including availability at the top of the page and the freshness of the sort, are covered in What a Healthy Category Page Looks Like.

High Traffic, Low Conversion: Seven Warning Signs in Your Catalog

2026-08-07T08:00:00Z

Automatic sorting takes care of the order. It cannot reorder stock, change a price or rewrite a product description. The data that drives the sort can, however, tell you where one of those is needed.

Below are seven patterns worth looking for every week. Each is a pair of numbers that disagree. They are the patterns Merchandiser reports as actionable insights, and you can look for them in any catalog that has views, add-to-carts, sales and stock per product.

1. High traffic, low conversion

The pattern: an in-stock product gets a lot of views and converts very few of them.

Customers are interested enough to click and not convinced enough to buy. The product is using a lot of attention and returning little.

What to check: price first, against your own range and against competitors. Then the images and the description. Then availability of sizes and variants, because a product whose popular sizes are gone converts badly for exactly that reason.

2. A best seller that is nearly out of stock

The pattern: strong weekly sales, and only a few units left.

At the current rate the product is gone within days, and it is probably in campaigns and in top positions right now.

What to do: replenish it. If that is not possible in time, reduce its promotion, so that the positions go to products that can still be sold next week.

3. Overstock with low demand

The pattern: a lot of stock, no sales and little traffic.

This product is not failing in front of customers. It is not in front of them at all. Capital is sitting in the warehouse while the page shows other things.

What to do: give it visibility before judging it. Move it up, include it in a campaign, review the price. If it still does not sell with traffic, you have learned something about the product and not about its position.

4. Cart interest that does not convert

The pattern: many add-to-carts, no purchases.

Customers got as far as the cart and stopped. Something happened between the intention and the payment.

What to check: the price including shipping, a competitor's offer, whether the stock in the cart is really sellable, and anything in the checkout that is specific to this product.

5. Demand for a product with few sizes left

The pattern: customers still view or buy the product, and most of its sizes are out of stock.

There is demand and no way to serve most of it.

What to do: restock the popular sizes if you can. Until then, lower the product's position, so that it does not hold a place most visitors cannot use. More on this in Size-Broken Products.

6. An active collection with no products

The pattern: a collection is live and contains nothing.

It can happen quietly, when the last products of a seasonal category sell out or a filter stops matching. The page is still linked from the menu and from search engines.

What to do: add products, or take the collection offline until it has some.

7. A sharp drop in collection traffic

The pattern: a collection gets far fewer visits this week than last week.

What to check: whether the page lost a link in the menu or on the home page, whether a campaign ended, and whether a recent change of the assortment or the order explains it. This is a prompt to look, not a verdict. Some drops are seasonal and expected.

Why pairs of numbers

Each of these looks harmless as a single number. High views are good. Low stock is normal. Many add-to-carts are welcome. The warning is in the combination, which is why these cases are easy to miss in a report sorted by one column.

It also means the same product can show up for different reasons over time. A product with high traffic and low conversion this week may be the one with too few sizes next week, and the action is different each time.

Make it a routine

A workable weekly routine:

Start with the critical findings: products with heavy traffic that do not convert, and best sellers about to run out. These cost money every day.
Send the stock-related findings to whoever orders stock.
Pick a few overstocked products for the next campaign.
Fix or close empty collections.
Read the traffic drops last. Most have an ordinary explanation.

Merchandiser grades each insight as critical, warning or info, lets you filter by type and severity, and exports the list as a spreadsheet for the teams that act on it. The whole-collection view of the same data is the Collection Health Score, described in What a Healthy Category Page Looks Like.

Using Store Sales to Rank Products Online

2026-07-31T08:00:00Z

A retailer with physical stores has two records of what customers buy. The website usually sorts by one of them. The other, often the larger one, stays in the ERP. This post is about putting it to work.

Why online sales alone can mislead

A category page is sorted by evidence, and for many products the online evidence is thin. A product that sold four units on the website last week might be a quiet seller, or it might be one of the chain's best products that happens to sit on page three, where nobody sees it.

That is the circular part of sorting by online sales: products sell where they are shown, and they are shown where they sold. A product that starts low stays low, because it never gets the views that would produce the sales that would move it up.

Store sales break the circle. In a store, the product is on a rail next to everything else. What customers pick there is evidence that does not depend on your website's current order.

Three numbers, not one

The useful structure is to keep three numbers per product and period:

online sales, from the website,
offline sales, from the stores,
the total of the two.

Keeping them apart means the choice is made in the sorting rule, per collection, and not once for the whole catalog when the data is prepared.

This is how Merchandiser's product metrics API is built. Your system sends, for each SKU, the sales of the last day, week and month as online sales, offline sales or totals. When you send an online or an offline figure, the total of that period is recalculated from the two. When you send a total, it is taken as it is.

When totals are the right signal

Rank by totals when store and web customers are broadly the same people buying the same things. Core ranges, basics and anything whose appeal does not depend on the channel are good candidates.

Totals help most with:

New products. A product that has been in stores for two weeks arrives online with evidence instead of with nothing.
Deep catalogs. Where online sales per product are low, totals give the ranking more to work with.
Seasonal turns. Stores often show the shift to a new season before the website does.

When to stay with online sales

Totals mislead when the channels differ.

Store-only favorites. Some products sell in stores because they are by the till, or because customers want to try them on. Online they may not move at all.
Different assortments. If the stores carry lines the website barely stocks, their sales say little about what to show online.
Different customers. A chain whose stores serve one region and whose website serves the whole country has two audiences.

In these categories, rank by online sales and treat store sales as a second opinion to look at, not as part of the score. Because the numbers are kept apart, both choices can live in the same store.

How the data gets there

The sales already exist, in the ERP or in the data warehouse. What is needed is a small scheduled job that sends them:

Collect units sold per SKU for the last day, week and month, split by channel.
Send them to the product metrics API in batches.
Check the answer. It reports how many products were updated and lists SKUs it did not recognize, which is usually a sign that the two systems name a product differently.

SKUs are the join. If the ERP and the website use different codes for the same product, settle that first; no ranking logic makes up for a mismatch there.

Using the numbers

Once the totals are in, they are ranking signals like any other. A rule can use them directly, for example total weekly sales scaled by size availability. Merchandiser AI learns from the totals when they are present, so the model sees the demand of both channels.

A sensible first step is to change one large category to a rule that uses totals, preview the order and compare it with the current one. The products that jump are the ones your stores knew about and your website did not. Look at a few of them. If they are products you would be glad to see higher, widen the change. If they are store-only favorites, that category is better left on online sales.

For choosing between a rule and the model for a category, see Sorting Rules vs. AI Ranking.

Sorting a Shopify Collection Automatically: What the Admin API Allows

2026-07-24T08:00:00Z

Shopify gives every collection a sort order, and for many stores the built-in options are enough for a while. This post is about what comes after: how a collection gets an order that Shopify's options cannot express, and what that means for your store and your theme.

The built-in options, and where they stop

In the Shopify admin, a collection is sorted by one of a fixed list: best selling, product title, price, date, or manually.

Each automatic option uses one criterion. That is the limit. "Best selling" does not know that the best seller has two sizes left. "Newest" does not know that last week's arrivals are not selling. There is no option for "what sells, as long as it is available, with new products given a chance".

The remaining option is manual. A manually sorted collection shows its products in exactly the order you arrange. It can express anything, and somebody has to do the arranging, in every collection, again and again.

What the Admin API adds

The manual order of a collection does not have to be set by dragging products in the admin. Shopify's Admin API has an operation that moves the products of a collection to given positions. An app with the right permission can call it.

That turns "manual" into "set from outside". The order is calculated elsewhere, from whatever data and logic you like, and written to the collection through the API. To Shopify, and to your theme, it is an ordinary manually sorted collection.

Two details matter:

Only manually sorted collections accept an order. A collection that Shopify sorts by best selling or by price has to be switched to manual first.
It is the Admin API, not the Storefront API. The Storefront API reads data for a storefront. It cannot change the order of a collection.

What this means for your theme

Nothing. The order lives on the collection, so the theme does what it always does: it asks Shopify for the collection's products and shows them.

There is no script to add, no app block and no extra request when the page loads. This is also why this kind of sorting does not affect the speed of the storefront. All the work happens before the visitor arrives.

One thing to check: a collection page shows the saved order when its sorting is left at the collection's default. If your theme offers visitors a "sort by" menu, their choice still overrides it for their visit, as it should.

What an app needs from you

To read your catalog and write an order, an app needs Admin API access with these kinds of permission:

Read products, to know your products, variants, stock and collections.
Write products. This is the permission Shopify requires for reordering a collection's products, even though no product is edited.
Read orders, if sales are counted from your orders.

It is worth asking any sorting app what it writes. The honest answer for a pure sorting tool is short: the order of products in collections, and the sort mode of a collection when it has to be switched to manual.

How Merchandiser uses it

Merchandiser's Shopify integration works exactly this way.

It reads your active, published products with their variants, the collections they belong to and their metafields, and counts sales from your orders. Views, add-to-carts, revenue and conversion rate come from Google Analytics 4.
It calculates the order of each collection with a sorting rule or with Merchandiser AI, keeping your pins in place.
It saves that order on the collection, and switches a collection to manual sorting when it first applies an order to it.
It does not change your products, prices, stock, orders or theme.

Each collection is re-sorted on its own schedule; see How Often Should a Collection Be Re-Sorted?

Before you automate

List the collections that matter. Sorting every collection is possible; start with the ones that carry traffic.
Note which are sorted automatically by Shopify today. They will become manually sorted collections.
Check the size data. If sizes are variants with their own stock, size availability can be calculated and used in the sort. See Size-Broken Products.
Decide the rule per collection. New arrivals, clearance and core categories usually need different ones; Sorting Rules vs. AI Ranking covers the choice.
Preview before applying. Look at the order a rule produces before it reaches the store.

How Often Should a Collection Be Re-Sorted?

2026-07-17T08:00:00Z

Once sorting is automatic, the next question is how often it should run. "As often as possible" sounds safe and is rarely right. The useful answer depends on the category, and working it out takes three questions.

What changes between two sorts?

An order goes stale because the things it was based on change. There are three of them, and they move at different speeds.

Stock changes fastest. A product can sell out, or lose its popular sizes, in an afternoon. If the sort uses availability, as it should, a new sort moves that product down as soon as the data shows it.

Demand changes more slowly. Signals like weekly sales and weekly views are sums over days, so one more hour of data moves them only a little. A weekend, a campaign or a change in the weather shifts them over days.

The catalog changes in steps. New products arrive when a delivery is put online, and prices change when someone changes them.

So the question "how often?" is mostly the question "how fast does stock move in this category?"

Which signals does the rule use?

A rule built on monthly sales is not going to give a different order an hour later. A rule built on daily sales and current stock might.

Look at the periods in the rule. Performance signals come for the last day, the last week and the last month. The shorter the period the rule leans on, the more a frequent sort can show. If the rule is all weekly and monthly numbers, daily sorting is plenty.

How fresh is the data?

A sort can only be as current as its inputs. If product data arrives once a day, sorting every hour repeats the same calculation on the same numbers twenty-three times.

Find out how often your product feed and your performance data are refreshed, and treat that as the ceiling. This is also why the freshness of the feed is worth watching: a sort that runs on time over data that stopped arriving is confidently wrong.

A way to decide

Category	What moves	Reasonable frequency
Campaign and sale pages during the campaign	Stock, within hours	Every hour to every few hours
Large core categories with steady traffic	Stock daily, demand over days	Several times a day to daily
New arrivals	The catalog, when deliveries go online	Daily
Brand pages and the long tail of small categories	Little	Daily to weekly
Evergreen categories with stable stock	Very little	Weekly or less often

These are starting points. The check is simple: if two sorts in a row give nearly the same order, the interval can be longer. If products that sold out are still near the top when you look, it should be shorter.

Do not treat every category alike

A single frequency for the whole store is either too slow for the busy categories or wasted on the rest. Most stores have a few dozen collections that carry most of the traffic and hundreds that carry little.

Give the busy ones a short interval and the rest a long one, and review the split now and then, because which categories are busy changes with the season.

In Merchandiser each collection has its own sort frequency, from every hour to once a month, and it can be changed for many collections in one step. The shortest interval depends on your plan; the plans list it.

What faster sorting cannot fix

Frequency is the last thing to tune. It does not help when:

The rule ignores availability. Then a sold-out product stays on top no matter how often you sort. Fix the rule first; see Size-Broken Products.
Pins hold the top positions. A pinned product does not move on any schedule. See Pinned Products: When to Pin and When to Let Go.
The data is stale. Then every sort repeats yesterday.

Keep an eye on it

A schedule can fail without anyone noticing: the page still shows an order, only an old one. The Collection Health Score includes Sort Freshness, which compares the time since a collection was last sorted with its own frequency. A collection that is overdue shows up there before a customer notices.

Sorting Rules vs. AI Ranking: Which to Use Where

2026-07-10T08:00:00Z

"Should we sort with AI or with rules?" is usually asked as if one had to win. In practice they answer different questions, and a store of any size has collections for both.

What a sorting rule is good at

A sorting rule is a formula. It takes signals, such as weekly sales, conversion rate, stock and days on market, and turns them into a score. You, or someone on your behalf, decide which signals count and how much.

Its strengths follow from that:

It is predictable. You can read the rule and know why a product is where it is.
It carries policy. "Newest first in the new arrivals category." "In clearance, the deepest discount first." These are decisions, not discoveries, and a rule states them exactly.
It works from day one. A rule needs today's data, not months of history.
It is easy to defend. When a brand manager asks why a product moved, the answer is one sentence.

Its weakness is the other side of the same coin. A rule only knows what you put into it. If season matters more for coats than for socks, the rule does not find that out. Someone has to notice and write a second rule.

What a model is good at

A machine-learning model does not get the weights from you. It learns them from your data: how price, discount, stock, brand, category, season, color, freshness, views, add-to-carts and sales have related to demand in your store. Then it predicts a score for each product.

It weighs many signals at once, including combinations nobody would write into a formula.
It adapts. Retraining on new data picks up shifts in demand without anyone editing anything.
It is specific to the store. In Merchandiser every store has its own model, trained only on that store's products and performance. What sells in a homeware shop does not leak into a fashion store's ranking.

The costs are real too. A model needs enough history to learn from, and "why is this product third?" has a less tidy answer than with a rule.

Where each one fits

Collection	Better fit	Why
New arrivals	Rule	The policy is the point: newest first.
Clearance and outlet	Rule	You decide what to clear: discount, stock or age.
Campaign pages	Pins plus a rule	The campaign products are decided; the rest follows a simple order.
Large core categories	Model	Many products, many signals, plenty of data.
A new store, or a category with little traffic	Rule	Too little history for a model to learn from.
Brand pages	Either	A rule is often enough; a model helps when the brand is large.

The pattern: use a rule where you know what you want, and a model where you want to find out what works.

Whichever you choose, a few things should hold.

Availability belongs in both. A product that cannot be bought should not be at the top, however good its numbers. A rule should multiply demand by size availability. Merchandiser AI does this by design: each prediction is weighted by the share of the product's sizes that are in stock. See Size-Broken Products for why this matters.

Pins sit above both. A pinned product keeps its position under a rule and under the model. Your decisions are not something the automatic order gets to argue with.

Your priorities can be added to a model. A model predicts demand; it does not know your margin targets or the season you want to push. In Merchandiser, products can carry a boost that is added to the model's estimate, which is how business priorities enter an AI-ranked collection.

Both should be previewed. Before switching a collection from one to the other, look at the resulting order. In Merchandiser the sort page shows the preview, and nothing changes on the store until you save.

How to start

Put every collection on a sensible rule first. That alone removes manual sorting and makes sure availability counts.
Let performance data build up.
Move the large, busy categories to the model, one at a time, and compare the previews.
Keep rules where the policy is explicit.
Review the choice when the store changes: a new category starts on a rule, a grown one may be ready for the model.

How Merchandiser's model works, what it learns from and what it predicts is described in Merchandiser AI.

What a Healthy Category Page Looks Like: Seven Signals

2026-07-10T08:00:00Z

When a category underperforms, the first suspects are the products and the prices. Often the cause is simpler and easier to fix: the page itself is in poor condition. The first row is half sold out, the order is two weeks old, or one brand fills the screen.

Here are seven signals that describe the condition of a category page. They are the signals behind Merchandiser's Collection Health Score, and they are worth checking whatever tool you use.

1. Availability at the top of the page

Most visitors see the first rows and little else. So the first question is: can they buy what they see?

Count how many of the first products are fully available, meaning in stock and not missing most of their sizes. A first row with two sold-out products and a third that only comes in XS is a wasted first row, however well those products sold last month.

This signal carries the most weight, together with the next one. See Size-Broken Products for how to keep such products out of the top positions.

2. Stock health

Next, the whole collection: what share of its products is in stock?

A category where a third of the products are unavailable feels picked over, even if the first row is fine. And an empty collection that is still live is the worst case. It is a dead end that visitors and search engines both reach.

3. Traffic trend

Compare this week's visits to the collection with last week's. A sharp drop is not a verdict, but it is a question. Did the page lose its place in the menu? Did a campaign end? Did a change of the sort move the products people were coming for?

The trend needs history to mean anything. A page with a handful of visits a week has no trend, only noise, and should not be judged on this signal.

4. Sort freshness

An order is a snapshot of the data at the moment of sorting. Every hour after that, it drifts from the truth: products sell out, new ones arrive, demand moves.

Compare the time since the last sort with how often the page is supposed to be sorted. A collection that should be sorted daily and was last sorted four days ago has a problem, either with the schedule or with whatever runs it. How a schedule is set per collection is described in Sorting Rules and Ranking Signals.

5. Feed freshness

Sorting is only as good as the data behind it. If the product feed stopped arriving on Monday, every sort since then has used Monday's stock and Monday's prices, and the page has been confidently wrong ever since.

Check when the last successful product feed came in. This is an account-wide signal. When it fails, every collection is affected at once, which makes it the first thing to rule out.

6. Product diversity

Open the page and count the brands in the first rows. When one brand makes up most of a mixed category, the page stops being a category and becomes that brand's page, and visitors who wanted to compare leave.

This signal only makes sense for mixed collections of a reasonable size. A brand page is one brand by design, and a collection of eight products cannot be judged for variety.

7. Pin discipline

Finally, how much of the page is pinned? A few pins are a sign of a team that curates. A page where half the products are pinned is manually sorted, with all the cost that comes with it: positions that no longer react to stock or demand.

The share of pinned products is the signal. When it is large, review the pins. Pinned Products: When to Pin and When to Let Go covers how to keep them in check.

From seven signals to one list

Checking seven things on one page is easy. Checking them on three hundred pages every week is not, which is why the signals are worth combining into a score.

Merchandiser scores each signal from 0 to 100 and combines them into one score per collection, with availability at the top of the page and stock health weighing the most. Signals that cannot be judged fairly, such as a traffic trend without history or brand variety on a brand page, are left out for that collection.

The result is a list of your collections with the weakest first, each with the evidence behind its score and a recommended action. The work changes from inspecting pages to working down a list.

What the signals do not tell you

None of these measure sales. A page can be in perfect condition and sell little, because the products or the prices are wrong for the audience. Health tells you that the shelf is in order. It removes the avoidable reasons for poor performance, so that what remains is a real question about the assortment.

Pinned Products: When to Pin and When to Let Go

2026-07-10T08:00:00Z

Automatic sorting decides the order from data. A pin is where you overrule it: this product, in this position, whatever the numbers say. Both are needed. The trouble starts when pins are used for the wrong job, or when nobody removes them.

What a pin is for

A sorting rule knows what happened: views, sales, stock. It does not know what you have planned. That gap is what pins are for.

Campaigns. The products of this week's campaign belong at the top of the campaign's category from the first hour, before they have any sales to show.
Launches. A new collection piece has no history. Pinning it gives it the exposure it needs to earn a position of its own.
Commitments. A placement agreed with a brand is a fact about your business that no signal carries.
The look of the page. Sometimes the first row has to tell a story: the hero product of the season, next to what goes with it.

In each case the reason comes from outside the data. That is the test.

What a pin is not for

The tempting misuse is pinning to repair the sort. A product looks too low, so it gets pinned higher. Another looks too high, so something is pinned above it. After a few weeks the first two rows are all pins, and the category is sorted by hand again, only less visibly.

If the automatic order is wrong, the rule is wrong, and the rule is what should change. A product that is "obviously" too low usually points at a missing signal: new arrivals that need a boost, a season that should count for more, availability that is not part of the score. Fixing that corrects every collection that uses the rule. A pin corrects one position on one page.

For how rules are built, see Sorting Rules and Ranking Signals.

Pins and the automatic sort work together

A pinned product keeps its position on every re-sort. Everything else is ordered around it. So a collection with two pins is still almost entirely data-driven: positions one and three are yours, the rest follows the rule or the AI model.

This is what makes pins cheap to use for the right reasons. You do not have to choose between a curated page and an automatic one.

Every pin needs an end

Setting a pin takes a second. Remembering to remove it takes a process, and most teams do not have one. The campaign ends, the pin stays, and a product that stopped selling holds position one for a month.

The reliable answer is to give pins an end when they are set:

Auto unpin after a set time. Choose the period per collection: a week for campaign categories, longer for others. Pins older than that are released.
Unpin products that are missing a size. A pinned product that has lost a size gives its position back to the automatic sort. A pin should not keep a product on top that part of your visitors cannot buy.
Unpin all. After a big campaign, clear a collection in one step and let the sort take over.

Pinning many products at once

For a campaign with dozens of products, pin from a list. In Merchandiser you export the collection's product list, fill in the Pin column with the position of each product and upload the file. The file is the complete pin list: products without a pin in it are unpinned.

This also makes pins reviewable. A spreadsheet can be checked by the campaign owner before it is applied, which is hard to do with positions set one by one.

How to tell that pins are getting out of hand

Three signs:

The pinned share keeps growing. Merchandiser shows the number of pinned products for every collection, and the Collection Health Score includes Pin Discipline, which falls when a large share of a collection is pinned.
Pinned products are out of stock or size broken. A pin that holds an unavailable product is costing you the best position on the page.
Nobody knows why a pin is there. The version history shows who changed pins and when. If the answer is "last spring", unpin it.

A rule of thumb

Pin what you have decided. Sort what you have measured. And when you find yourself pinning to fix the order, stop and look at the rule instead.

All the pin features mentioned here are described in Product Pinning and Manual Control.

Size-Broken Products: How to Keep Them Off Your Top Row

2026-07-10T08:00:00Z

A best seller rarely leaves the first row because it stops selling. It leaves because it sold so well that the popular sizes are gone. What stays behind is a product with strong sales history, a "best seller" position and two sizes that fit almost nobody. This post is about that product, and about how to move it without watching every category by hand.

What "size broken" means

Stock is usually tracked per product: in stock or out of stock. For anything that comes in sizes, that is too coarse. A jacket with XS and XXL left is in stock. For most of the people who see it, it is not for sale.

The measure that captures this is size availability: the share of a product's sizes that are in stock. Six sizes with all six available is full availability. Six sizes with two available is one third.

A product is size broken when that share has fallen so far that most visitors will not find their size. Merchandiser puts the line at 30%: a product with 30% or fewer of its sizes in stock gets the Size Broken badge in the panel.

Why sales history makes it worse

Most sorting, manual or automatic, looks at what sold. Sorting by weekly sales puts last week's winners first, and last week's winners are exactly the products that are running out this week.

The result is a first row that looks good in a report and performs badly on the page. Visitors click the product, open the size selector, find their size greyed out and leave. The product keeps collecting views, which a naive rule reads as more interest, and it holds its position.

The problem is not the sales signal. It is using the sales signal alone.

Sort by demand and availability together

The fix is to let availability scale demand. In a sorting rule, that means multiplying the demand part of the score by size availability:

A product with strong demand and all sizes in stock keeps its full score.
A product with the same demand and a third of its sizes keeps a third of it.
A product that is out of stock falls to the bottom.

This is gentler and more accurate than a hard cut-off. A product with one missing size loses a little. A product with one size left loses most of its position and makes room for something customers can buy.

In Merchandiser, size availability is one of the ranking signals a rule can use, together with stock units and in-stock status. Merchandiser AI applies the same idea without a rule: each prediction is weighted by size availability, so a promising product that has sold out of most sizes ranks below an equally promising one that is fully in stock.

Check the first row, not the average

A collection can have healthy stock overall and a broken first row. What matters is what the visitor sees before scrolling.

That is why the Collection Health Score has a signal for it. Top-of-Page Availability counts how many of the first products of a collection are fully available, meaning in stock and not short of sizes. When the count drops, the collection's score drops with it, and the recommendation is plain: replenish or demote the out-of-stock and size-broken products near the top.

The health page lists every active collection with the lowest scores first, so the categories with a broken first row surface on their own.

Do not let pins undo it

Sorting rules respect pins. That is the point of a pin, and it is also how a size-broken product stays in position one for weeks: someone pinned it for a campaign when it had every size, and nobody came back to unpin it.

Two settings help:

Auto unpin after a set time. The pin is released once it is older than the period you choose.
Unpin products that are missing a size. Once any size of a pinned product is out of stock, its pin is released and the product returns to the automatic sort.

The second is strict on purpose. A pin overrides the data, so it should only hold while the product is fully available. More on this in Product Pinning.

When the product is worth restocking

Demoting a size-broken product treats the symptom. Sometimes the right answer is to get the sizes back.

Merchandiser's insights separate the two cases. "Demanded product has limited variant availability" marks a product that customers still look at or buy while most of its sizes are missing. That is a restocking signal for the buying team, not only a sorting problem. A size-broken product that nobody looks at needs no action: the sort has already moved it down.

A short checklist

Make sure your product feed carries stock per size, not only per product.
Use size availability in the sorting rule of every collection that sells sized products.
Review the collections with a low top-of-page availability first.
Turn on auto-unpin for campaign pins.
Send the list of demanded, size-broken products to whoever orders stock.

What is Visual Merchandising? The Ultimate Digital Guide

2025-12-01T08:00:00Z

From manual shelf-stacking to AI-powered algorithms. Discover how optimizing your digital shelves can unlock hidden revenue.

The Evolution of the Storefront

In the physical world, Visual Merchandising is the art of displaying products in a way that stimulates interest and entices customers to make a purchase. Think of the mannequins in a window display, the strategic lighting in a luxury boutique, or the "impulse buy" candy racks at the checkout counter.

In the digital world, the concept remains the same, but the execution has shifted entirely. Your "storefront" is your homepage, and your "shelves" are your category pages. Online Visual Merchandising is the strategic arrangement of products on these digital shelves to maximize visibility, click-through rates, and ultimately, revenue.

The Hidden Cost of Manual Sorting

For years, e-commerce managers have relied on "gut feeling" to sort products. This manual process is not only tedious but costly:

Time Drain: Merchandising teams spend hours every week simply dragging and dropping products to change their order.
Human Bias: We tend to favor products we "like" personally, rather than what the data shows customers actually want.
Static Shelves: A manual list sorted on Monday is obsolete by Friday. Trends change hourly.

Optimizing Your Digital Shelves

Space on a mobile screen is the most expensive real estate in retail. Products located "above the fold" (visible without scrolling) get most of the attention. If your best-selling item is buried in row 5, you are actively losing money.

Effective digital visual merchandising requires a mix of strategies:

Predefined Sorting Rules: Basic logic like "Sort by Price: High to Low" or "Sort by Stock Level". Good for basic organization but lacks intelligence.
Smart Collections: Grouping items dynamically, such as "Best Sellers under $50" or "New Arrivals in Summer Wear".

The Game Changer: Merchandiser AI

The future of visual merchandising is automation. Merchandiser AI moves beyond simple rules and uses machine learning to rank products based on real-time performance metrics. It acts as a 24/7 digital store manager that never sleeps.

How Automated Ranking Works

Instead of guessing, the AI analyzes thousands of data points to determine the perfect score for each product. Commonly used ranking signals include:

Revenue & Sales Velocity: Products that generate the most cash are prioritized. The system detects "rising stars"—products that are selling fast despite having low historical data.
Traffic & Views: High traffic but low sales? The AI might de-rank these "window shopper" items or signal you to check the price, while boosting items with high conversion rates.
Stock Availability: Nothing kills conversion like a "Sold Out" badge on the first row. Merchandiser AI automatically buries out-of-stock items and pushes inventory-rich products to the top.
New Arrivals (Freshness): New products need visibility to prove their worth. The algorithm gives "freshness" a temporary boost to test performance before settling into a long-term rank.

Ready to Automate Your Visual Merchandising?

Stop wasting hours on manual sorting. Let Merchandiser AI analyze your sales, traffic, and inventory to create the perfect collection pages instantly. Request a demo or explore the features.

Merchandiser blog

Rolling Back a Bad Sort: Version History for Merchandising Teams

How a sort goes wrong

What a version has to contain

Finding the change

Rolling back

A second pair of eyes

Habits that make it work

Brand Diversity on the Shelf: Why One Brand Shouldn't Own the Page

How it happens

Measure it

Look at the first rows too

What to do about it

When dominance is fine

Where it fits

High Traffic, Low Conversion: Seven Warning Signs in Your Catalog

1. High traffic, low conversion

2. A best seller that is nearly out of stock

3. Overstock with low demand

4. Cart interest that does not convert

5. Demand for a product with few sizes left

6. An active collection with no products

7. A sharp drop in collection traffic

Why pairs of numbers

Make it a routine

Using Store Sales to Rank Products Online

Why online sales alone can mislead

Three numbers, not one

When totals are the right signal

When to stay with online sales

How the data gets there

Using the numbers

Sorting a Shopify Collection Automatically: What the Admin API Allows

The built-in options, and where they stop

What the Admin API adds

What this means for your theme

What an app needs from you

How Merchandiser uses it

Before you automate

How Often Should a Collection Be Re-Sorted?

What changes between two sorts?

Which signals does the rule use?

How fresh is the data?

A way to decide

Do not treat every category alike

What faster sorting cannot fix

Keep an eye on it

Sorting Rules vs. AI Ranking: Which to Use Where

What a sorting rule is good at

What a model is good at

Where each one fits

What they share

How to start

What a Healthy Category Page Looks Like: Seven Signals

1. Availability at the top of the page

2. Stock health

3. Traffic trend

4. Sort freshness

5. Feed freshness

6. Product diversity

7. Pin discipline

From seven signals to one list

What the signals do not tell you

Pinned Products: When to Pin and When to Let Go

What a pin is for

What a pin is not for

Pins and the automatic sort work together

Every pin needs an end

Pinning many products at once

How to tell that pins are getting out of hand

A rule of thumb

Size-Broken Products: How to Keep Them Off Your Top Row

What "size broken" means

Why sales history makes it worse

Sort by demand and availability together

Check the first row, not the average

Do not let pins undo it

When the product is worth restocking

A short checklist

What is Visual Merchandising? The Ultimate Digital Guide