Assortment Planning for Online Stores: What to Add and Keep

Plan your assortment as a monthly cycle. Add new products in small batches, mostly to categories that already sell. Give each a fair test in your ads, keep what sells and move budget away from the rest. Growing stores in our data renewed their assortment rather than enlarging it. In stores that doubled revenue from ads, products added during the period brought 47.3% of revenue in the period’s last three months. In stores that declined, the share was 26.1% (GetProfit data, 100 stores, June 2025 – June 2026).

This guide is the map of our series on assortment strategy. Each section answers one question briefly and points to the article that covers it in full. Every figure names its sample and period.

What is assortment planning for a store that sells through Google Ads?

Assortment planning is deciding which products a store carries, in how many variants, and when to add or drop them. Marketing describes a product assortment by its width and depth, among other measures. According to Wikipedia’s article on product mix, width is the number of product lines a company offers, and depth is the number of variations within a line.

Classic retail also treats each category as a small business. Wikipedia’s article on category management describes managing a product category “as a strategic business unit”. Each category gets its own targets and a cycle of steps that ends in a review.

A store that sells mostly through Google Ads adds one more layer. A product on your shelf sells only if the ads show it, and the ads show it only if it is in your feed and in a campaign. So every product has four states you can measure: in the catalogue, shown in ads, given budget, sold. Planning the assortment for such a store means managing all four.

Count products the way Google does: by SKU, the single item with its own ID in the feed. One model of trainers in seven sizes is seven SKUs, and each of them can sell differently.

This guide works from ad data: what Google showed, what got budget and what sold. That data exists for every product you advertise. Margin, returns and stock levels live in your own systems, so bring them into the decision yourself.

How much of a catalogue takes part in Google Ads?

In our study of 1.4 million products, fewer than half of the products took part in advertising. We looked at 1,404,808 products in 130+ stores over 13 months, and only 632,826 of them, or 45.0%, got any ad budget at all.

So the ads never tested more than half of the products in the study. The share of the catalogue that takes part in advertising is called assortment coverage. Check it before any assortment decision: a product with no impressions has told you nothing yet, good or bad.

Spend also concentrates on a few products. In the median store of the study, the top 1% of products took 27.7% of spend and the top 10% took 68.1%. Your assortment in ads is much narrower than your catalogue. Planning starts with knowing which product sits in which layer.

Will adding more products grow your sales?

Not by itself. When we compared months with a similar change in spend, more active products showed no link with revenue from ads. The rank correlation between the change in active products and the change in revenue sat between −0.057 and +0.040. That means no link at all (GetProfit data, 899 store-months, June 2025 – June 2026).

We also followed 144 cases where a store grew its number of active products by more than 20% in one month. We counted that growth over and above the median store’s change in the same month (GetProfit data, June 2025 – June 2026). Each figure in the table is a change on the previous month, measured against the median store in the same calendar month, so season and holidays cancel out.

Revenue jumped in the month of the expansion and then stopped growing: over the next two months it moved by +0.5% and −1.6%. Why does it look as if more products sell more? Stores that add products often add budget at the same time: in our data the two moved together, with a correlation of +0.259. Compare months with a similar change in spend, and the link disappears.

One caveat: we count a product as new when it gets its first impression, so a change of IDs in the feed also looks like a new product. The full breakdown is in whether more products mean more sales.

The same applies to a supplier’s catalogue uploaded in one go. Thousands of new products then share the same budget, and each of them needs its own test. How to handle such a catalogue is in supplier catalogues in Google Ads.

“Add everything and cut later” is sound advice when it describes a flow: a steady stream of new products and a steady cut of those that failed their test. It misleads when it means one bulk upload and no selection.

Growing stores renew their assortment rather than enlarge it

Renewal goes with growth: a steady share of new products and a habit of keeping the ones that sell. In our analysis of 100 stores, we compared the first three months of a 13-month period with the last three. Then we measured what share of current revenue came from products that were not in the assortment at the start.

GetProfit data, 100 stores, June 2025 – June 2026. Across all 100 stores, the correlation between this share and the change in the store’s revenue was +0.208. Three checks rule out the obvious explanations:

• Not just spend. The link held among stores that cut spend (correlation +0.205), kept it flat (+0.302) and raised it (+0.266). The share of new products was barely linked with spend growth (+0.106).
• Not just arithmetic. In stores that doubled, revenue from old products also grew, to 126% of its starting level. New products came on top. In stores that declined, revenue from old products fell to 28%.
• Not the number of tests. Stores that declined added more new products than stores that doubled, 168 a month against 61, and got a smaller share of revenue from them. They were also larger, so the careful reading is that the number of tests alone did not go with growth.

Read these numbers as observations, not an experiment. The stores are our clients, and our algorithm already manages product rotation in their ads. That narrows the gap between stores that renew and stores that don’t, so the real effect is probably larger than the measured one.

The median store got 35.7% of its revenue from new products, and 25 of the 100 stores got less than 20%. How to measure your own share, and what to do if it is low, is in catalogue renewal and new products.

Renewal works alongside spend. In the same data, spend growth still had the strongest link with revenue growth (correlation +0.786). How spend and ROAS fit together is in growing a store with Google Ads.

Where should new products go: deeper, wider or up in price?

Mostly deeper, with a small share kept for new directions. Deepening means adding models or price points to a category that already sells. Widening means adding a category you have never carried.

Growing stores put 91.1% of their new products into categories with earlier sales, and declining stores put 98.7% there. In other words, growing stores sent 8.9% of their new products into new categories, and declining stores 1.3% (GetProfit data, 100 stores, June 2025 – June 2026).

A single new product rarely sells fast in either place. Of products added to categories with earlier sales, 4.0% sold within three months; in categories with no earlier sales, 3.5%. In 39% of stores it was the other way round, so treat this as a rough guide (GetProfit data, 760,842 new products, June 2025 – June 2026).

Price is the third direction. Stores that doubled their revenue over the year raised their average order value by 72.7% (GetProfit data, 95 stores). Stores that stayed flat raised it by 12.3%, and stores that declined by 0.3%. In 57 of 92 stores, new products had a higher median order value than the old ones (GetProfit data, June 2025 – June 2026).

Where can you find products worth adding?

Start with the categories that already sell for you, then look at the market through Google’s best sellers report. According to Google’s Merchant API reference (BestSellersProductClusterView), the report ranks product clusters by popularity on ads and organic surfaces. A product cluster is one product across its offers and variants. Google sets the rank within a chosen category and country, based on the estimated number of units sold.

Rankings come weekly or monthly. The product category in this report follows Google’s own product taxonomy, so the same product can sit in a different branch than in your store’s menu.

Each product in the report also carries an inventory status. The status says whether the product is in stock in at least one of your countries, out of stock in all of them, or not in your inventory at all. The last group is a ready list of popular products you don’t carry. How to read it before you call a supplier is in reading the best sellers report.

How do you give a new product a fair test?

Give it enough clicks before you judge it. In the study of 1.4 million products, the minimum fair test was 15–20 clicks, roughly $3–4 at the sample’s average cost of $0.20 per click. The threshold differed by category, from 15 clicks for bags to 50 for hobby products.

Many products never get even that. Of 1,196,357 products that started the 13 months with zero spend, 74.1% never got a single impression, and 2.59% sold at least once. Those that did sell took a median 4 months to their first sale.

So a new product needs two things: a place in a campaign where it gets impressions, and a rule for when its test is over. In Performance Max you control the first through listing groups. According to the Google Ads API documentation (Listing Groups for Retail), listing groups decide which products each asset group includes or excludes.

The same page notes that listing groups work best for groups of products, for example by custom label or brand, rather than single products. Custom labels come from your Merchant Center feed; in the Google Ads API they are called custom attributes. The full launch process, from the feed to the keep-or-drop decision, is in launching new products in Google Ads.

What should you keep, limit or cut?

Keep what sells at your target, limit what spends without return, and let untested products finish a fair test. Cutting on a spend threshold alone loses future sales.

In our study, the rule “no sale after N spent, switch it off” was right 88.1% of the time past $5 of spend. Past $500 it was right only 66.2% of the time. In the same study, 35.4% of products that had spent at least the cost of one conversion without a sale went on to sell, after a median 2 months.

Winners also change fast. Of the products that ever won, 64.7% did it only once, with a single sale in a single month. A steady bestseller held its place for a median 3 months. A list of “our best products” made once a year describes a catalogue that no longer exists.

The long tail brings more revenue than it seems. In the median store and month, products with one or two sales made up 64.9% of selling products and brought 57.0% of revenue (GetProfit data, 116 stores, June 2025 – June 2026). The tail also turns over: 75.0% of a month’s tail products had no sales the month before, against 12.5% for products with six or more sales.

So judge groups, not single products. The portal does this with labels: it looks at each product’s conversions and ROAS over the last 12 months against the account’s goal. Then it puts the product into one of eight groups.

The same logic works one level up, for categories. Some categories earn above target on little budget, some spend without return, and some never got enough impressions to judge. How to sort them is in category performance analysis. The decisions that cost stores the most, from cutting too early to never testing at all, are in assortment mistakes in Google Ads.

Is your store a warehouse or a showcase?

The shape of your catalogue decides where most of the work lies. We grouped 114 stores by how their catalogue behaves in ads and got two shapes: 73 “warehouses” and 41 “showcases” (GetProfit data, June 2025 – June 2026).

A one-question self-test: did at least 30% of your catalogue sell at least once in the last year? If not, you are closer to a warehouse. This rule matched our grouping in 87% of stores. The two shapes blend into each other, so a store near the line can read both columns.

For a warehouse, most of the work is choosing which part of a wide catalogue gets budget. For a showcase, it is choosing what to add next. Both are covered in warehouse or showcase.

How does your category change the rules?

Categories differ in how much spend goes to products that never sell and in how long a test takes. These figures come from the same study of 130+ stores over 13 months.

A furniture product sells rarely, so a test there needs more clicks before “no sale” means anything. Car parts and tools carry huge catalogues of similar products, and a large share of their spend lands on products that never sell.

Print-on-demand stores are a special case: thousands of design variants on a few base products. Whether the usual rules apply to them is in print-on-demand on Google Shopping.

Which numbers show whether your assortment works?

Four numbers from your ad data describe the assortment better than one ROAS figure for the whole account. The reference points in the table are observations from our data, not targets.

In our study, the share of spend on products with no sales separated strong stores from weak ones. Both groups had products with no sales, but the bottom quarter by ROAS spent 71.8% of its budget on them, against 39.8% in the top quarter.

Compare like with like. A share of selling products counted over one month looks far smaller than the same share over a year. How these measures relate to retail’s stock measures is in sell-through rate and other assortment KPIs.

How often should you review the assortment?

Every month. A steady bestseller lasts a median 3 months, and three quarters of the tail turns over each month, so a quarterly review comes too late. Go through it in this order:

1. Coverage. Which products and categories got no impressions, and why: not in a campaign, a feed problem or no budget.
2. New products. Did last month’s additions get impressions and clicks, and which of them have finished their test?
3. Spend without sales. Which groups spent without return? Limit them, but keep a small budget on products still in their test.
4. Winners. Which products moved into the top or dropped out of it? Protect the budget of the current ones.
5. Renewal. What share of revenue came from products new this year, and what goes into next month’s batch?

The full checklist, with what to look at in each step, is in the monthly assortment review checklist.

Where does merchandising happen when shoppers meet you in an ad?

On your own site you choose what sits on the home page. When most shoppers first see your products in Shopping ads or Performance Max, merchandising starts earlier. It starts in the feed: which products get into campaigns and how you group them. That shift is covered in merchandising for Google Shopping.

A worked example: one store’s renewal share

Example store, not client data.

A tableware shop with 3,000 products earns 180,000 a month from ads. Over the last three months it earned 540,000. Of that, 135,000 came from products that were not in its assortment in the first three months of the year. So its share of revenue from new products is 25%.

That puts it next to the declining stores in our data (26.1%), whatever its ROAS says today. Its next step is a monthly batch of new products, mostly in the categories that already sell. Each product gets a test budget and a date for the keep-or-drop decision.

What to do this month

1. Split your catalogue into layers for the last 12 months: no impressions, spend without sales, sold. Write down the share of products and of spend in each.
2. Calculate your renewal share. Take revenue for the last three months and find the part that came from products you didn’t have a year ago. Compare it with 26.1% and 47.3%.
3. Choose where to add. Put most new products into categories that already sell; keep a small share for one new direction.
4. Add in batches with a test rule. Decide in advance, for each category, how many clicks a product gets before you judge it.
5. Sort the rest by group. Scale what sells at target, limit what doesn’t, and move budget rather than switching products off before a fair test.
6. Put the review in the calendar for the same day each month.

For your own account, the portal’s assortment section covers the first step and the sorting in the fifth. It walks the catalogue through six stages over 90 days, from “In the catalogue” to “Paying off at target efficiency”. It then names notable product groups in four lists: “Working well”, “Expand”, “Limit and repurpose” and “Give a test”.

For one store’s history of this work, see the published case where ROAS rose from 243% to 650%.

How much of the catalogue reached sales — and what the rest is doing. The portal walks the whole catalogue through the stages — from the feed to sales, shows where the budget goes by product group, and answers whether you really have a season. The portal changes nothing without your consent.

Break down your catalogue →

Sources

• BestSellersProductClusterView — Merchant API reference — the best sellers report ranks products by popularity on ads and organic surfaces in a category and country, based on estimated units sold; weekly or monthly rankings; categories from Google’s product taxonomy; inventory status in stock, out of stock or not in inventory. Checked 2 October 2026.
• Listing Groups for Retail — Google Ads API — listing groups set which products each Performance Max asset group includes or excludes; they work best for groups of products by custom label or brand; custom attribute equals the Merchant Center custom label. Checked 2 October 2026.
• Product mix — Wikipedia — width and depth of a product mix. Checked 2 October 2026.
• Category management — Wikipedia — a category managed as a strategic business unit, with a cycle of steps ending in review. Checked 2 October 2026.
• GetProfit data: study of 1,404,808 products in 130+ stores over 13 months — catalogue layers, spend concentration, test thresholds, switch-off accuracy, later conversions, bestseller lifetime, categories.
• GetProfit data: 100 stores, June 2025 – June 2026 — share of revenue from new products by store group, checks against spend, where new products were added, new versus old order value.
• GetProfit data: store-months and 144 expansions, June 2025 – June 2026 — number of active products versus revenue at a similar change in spend.
• GetProfit data: 95 stores — average order value of stores that doubled; 116 stores — the long tail; 114 stores — warehouse and showcase shapes; 760,842 new products — sale within three months.
• GetProfit portal — product labels and the assortment section.