The tableware sells. But which pieces?

Imagine a supplier has sent the owner of an online tableware shop a new catalogue. On screen are mugs, bowls, plates and serving dishes. The buying terms look good, but she cannot take everything.

The owner remembers packing a lot of bowls lately. Should she order more of those?

She opens the orders. Bowls do sell. But one blue model accounts for most of the shipments. Several others sit alongside it in the catalogue, and their sales histories look quite different.

This is a fictional example built around a simple question: what exactly are you planning to expand — a successful model, a clear customer need, or simply a section of the website with a good total?

Group the tableware around a buying decision

The shop’s total revenue will not tell her what to choose from the supplier’s catalogue. The owner starts with product types: mugs, bowls, plates and serving dishes. Material, colour and price range can wait.

This is a useful order to work in: first compare recognisable groups, then look inside the one that gives you a reason to continue. If you start with a table of “type × colour × material × price”, you may end up with lots of cells containing a single purchase. You will still be choosing almost blind.

For the first review, she uses one completed period across all groups. She notes when products were out of stock and does not compare a newcomer added near the end of the period with a model that was available throughout.

If you repeat this, decide what you are counting first. The number of different models, units sold and orders are different things. A customer can buy several bowls in one order, and a single model can come in several colours.

Two purchases put a segment in the lead

The owner adds ad spend, orders and their values to the groups. Serving dishes lead the chart by a long way: each euro spent on ads corresponds to €12 in order value. Plates bring just €0.50. She is tempted to order more serving dishes and put plates aside.

But the long bar has two orders next to it, and the short one has one.

Fictional example1–30 September 2026

Serving dishes lead.
But there are only two purchases.

ROAS · order value per €1 of ad spend

Mugs

3×

24 orders

Bowls

5×

18 orders

Serving dishes

12×

2 ordersLittle data

Plates

0.5×

1 orderLittle data

0×6×12×

A high return ≠ a reliable return.
With 1–2 orders, the next purchase can change the picture considerably.

Fictional example for 1–30 September 2026. All ad clicks and orders in the example fall within this period. Each completed order is assigned exactly once, to the segment of the product in the last ad clicked before purchase. Its value includes the whole order, including other products; amounts are after discounts and exclude VAT and delivery. There are no returns in this example. This is a calculation rule defined specifically for the example, not a universal description of how Google Ads product reports work.

Serving dishes look stronger than the rest, but the entire result rests on two purchases. A different basket would have changed both average order value and ROAS considerably. Plates look weak at the moment, but one order does not describe every model and all its customers. The money has already been spent, though: a small number of purchases is no reason to extend a test indefinitely.

The owner does not turn this table into a final purchasing list. She notes two questions: why the orders attributed to serving dishes were so valuable, and what the spend on plates has actually tested. A good result is a reason to investigate; a poor one is a reason to check the conditions and spending limit. Neither promises the same outcome next month. Even 12× is order value divided by ad spend, not profit.

The serving dish costs €40. Why is the order €120?

The owner checks the dish price against the order calculations: €40 excluding VAT. There is no price error. She then opens the two orders that make up the €240. All the prices below also exclude VAT.

The first contains a serving dish at €40, two salad bowls at €25 each and serving spoons at €30. Total: €120.

The second contains the same serving dish at €40, six dessert plates at €10 each and a set of napkins at €20. Another €120.

The high order value now makes sense. The dish itself did not get more expensive: customers added other things for the table. Across these two orders, serving dishes accounted for €80 and the other products for €160. It would be wrong to count all €240 as revenue from the serving dishes themselves.

There are plates in the second order, too. Under the rule chosen for the chart, that whole order is assigned to serving dishes. So the short bar for plates does not mean only one plate sold during the month. The chart answers a question about orders following ads for a particular group; the order lines show what people actually bought.

Two purchases do not establish a bundle everyone should be offered. But they do give you something to check: which additions appear in other orders containing the dish, whether they are in stock and whether they help the customer with the same task. You can start with a few relevant recommendations on the product page, then see whether people choose them and what remains after costs. A bundle discount is not required.

We explain how to turn that review into an offer for the customer in “Product Bundling Around a Bestseller to Raise Order Value”. For now, the owner saves the combinations she found and returns to her main question: which tableware is actually worth adding to the next stock order?

Check what sits behind a good total

Within bowls, the owner breaks sales down by model. She adds stock availability dates and notes on returns alongside them. Now she knows what to find out next.

No row says “buy it” on its own. Each points to where the missing answer might be: orders, stock history, ads or a conversation with a customer.

In our example, the other bowls do not yet justify buying a whole new range. The first job is to understand the blue model and decide whether to test a similar option.

Find out what customers mean when they buy a “bowl”

The owner calls a customer after delivery.

“Have you had a chance to use them yet?”

“Yes, we’ve been using them for soup. They’re just right, not huge.”

“What were you using before?”

“Ordinary soup plates. I wanted something that would fit more easily on the children’s little table.”

“What did you look at when you were choosing? The size, the colour?”

“Size first. Colour came later. You had the measurements on the page, so I compared them with mine.”

The conversation is fictional. A real call may produce a different answer. That is what you need to hear, rather than steering someone towards the explanation you have already chosen.

The owner notes the need and the criterion. For now, it is one customer’s reason for choosing. She reviews other enquiries, speaks to customers who are willing to talk and checks their answers against the orders.

If the same need keeps coming up, she has a specific question for the supplier: which other models have the right dimensions? She will need to check the sample’s measurements and other properties. A similar photograph is not enough.

Ad return helps you decide where to look

You can add ad spend and the purchase value the system attributes to those ads. Take both from one consistent report for the same period. Do not divide the group’s entire revenue from shop orders by one campaign’s spend and call the result that campaign’s ROAS.

Shares are easy to misread, too. For example, if a group received 10% of ad spend and 30% of recorded purchase value, the ratio of those shares is three. Its ROAS is therefore three times the overall ROAS in the same dataset. It does not mean every rouble spent on ads returned exactly three roubles.

Even a high return does not tell you how much is left after stock purchases, delivery, returns and all the other costs. It gives you a reason to examine the group more closely. Whether you can profitably add new products to it still needs testing.

Go back to the supplier with a specific request

The owner now opens the catalogue with a different task. She is looking for models that suit the use she has uncovered, rather than just any bowls. She checks dimensions, requests samples and asks about order terms and replenishment lead times.

Alongside that, she calculates how much money she will need before the first sale and how much might remain from an order after costs. If the minimum order is too large for a test, the sales history of a similar model does not remove that risk.

The review may lead to a small purchase of selected models. Or it may lead to a decision to improve the product pages for items she already carries, or to keep the blue bowl in stock. Expanding the assortment right now is not compulsory.

Before your next stock order, choose one category you would like to expand. Look at which models account for its results, who needs them and what already sits alongside them without a clear test. Your question to the supplier will then be far more precise than “what’s selling well for you at the moment?”