How Vivense Lifted ROAS 32% by Not Advertising Products It Couldn't Sell

The Client

Vivense is one of Turkey's largest online furniture and home retailers. Sofas, beds, dining sets, garden furniture, mattresses, storage — the full range, sold online at scale.

Furniture e-commerce has a particular shape that makes it unusually demanding to advertise. Order values are high, so a wasted click costs more than it does in most categories. Consideration cycles are long. And, most importantly, the catalogue multiplies: a single sofa model is not one product but dozens, once you account for fabric, colour, size, and — in the case of corner sofas — orientation. Left corner and right corner are different SKUs with different stock levels.

The result is a catalogue that runs to tens of thousands of individual variants each with its own availability, price and delivery status, all of it changing daily.

The Problem

Catalog ads work by pulling products from a feed and matching them to users automatically. That automation is what makes them efficient at scale — and it is also what makes feed accuracy non-negotiable. The platform advertises whatever the feed tells it to advertise. It has no independent knowledge of what is actually in the warehouse.

For a catalogue the size of Vivense's, that creates a persistent and largely invisible leak.

Out-of-stock variants keep getting impressions. A grey left-corner sofa sells out. Until the feed reflects that, the ad keeps running — and it keeps running well, because the algorithm has learned that this product performs. Budget flows toward it precisely because it used to convert. Every impression is now spent on an item nobody can buy. Worse, the users who click land on an unavailable product page, which costs Vivense not just the media spend but the trust of a high-intent visitor.

Budget spreads evenly across a catalogue that performs very unevenly. In a catalogue of tens of thousands of variants, a small share of products drives the majority of revenue. Without product-level performance visibility, spend distributes according to the platform's own logic rather than the business's — and a lot of it lands on long-tail variants that were never going to carry the return.

Scale makes both problems impossible to manage manually. With a few hundred products you can spot the leaks. With tens of thousands of variants updating daily, no team can audit the catalogue by hand. The waste isn't dramatic in any single instance; it accumulates quietly, item by item, across the whole catalogue.

Put together, the challenge Vivense faced was a specific one: scale the catalogue in ads without scaling the waste that comes with it.

The Solution

Working with Vivense's partner D-DAT, we approached this as a data problem before a creative one.

Real-time stock and availability control in the feed

The first step was making the feed an accurate representation of what Vivense could actually sell, continuously rather than periodically. Availability at the variant level flowed into the feed in real time, so products that sold out stopped being advertised — automatically, without waiting for a manual export or a scheduled sync.

This sounds like housekeeping. Its financial effect is not. In a high-AOV category, every impression redirected away from an unavailable product is budget moved to one that can convert, and the effect compounds daily across the entire catalogue.

Product-level performance data and scoring

Removing waste is only half the equation; the other half is deciding where the recovered budget should go. Feedance's product-level performance data made it possible to see which variants were actually generating return rather than simply generating clicks — and to structure the feed so that budget concentrated on them.

Products were segmented by performance and stock depth, so campaigns could push the items that both converted well and had inventory behind them to sustain the demand. The catalogue stopped being a flat list and became a prioritised one.

Creative that answers the furniture buyer's real questions

With the data layer in place, Creative Suite generated branded creatives for the catalogue automatically — each one carrying the information a furniture shopper needs before they will click:

• The Vivense brand lockup, consistent across every placement
• The original price struck through and the cart price shown — a discount is only persuasive if the shopper can see both numbers
• A "Hızlı Teslimat" (Fast Delivery) badge where applicable — in furniture, delivery time is a genuine purchase objection, and answering it on the creative removes a reason to hesitate
• Seasonal campaign styling — spring, winter and promotional templates applied across the catalogue at once, so ads matched the campaign calendar without a design project each time

Because these were generated from feed data, they stayed correct on their own. A price change updated the banner. A product going out of stock removed it from circulation entirely — creative and targeting moving together rather than one lagging behind the other.

The Results

ROAS increased by 32%.

The mechanism behind that number is worth being precise about, because it wasn't a bidding change or a budget increase. The same catalogue, the same channel and — as far as the media plan was concerned — the same operation produced a third more return, because the budget stopped being spent in two specific ways: on products that were unavailable, and on products that were available but weren't converting.

Alongside the performance gain, Vivense's team gained operational leverage. Managing creative and availability across tens of thousands of variants moved from a recurring manual burden to an automated process — which is what made the result sustainable rather than a one-off campaign win.

What other large-catalogue retailers can take from this

Feed accuracy is a media efficiency lever, not an IT task. Out-of-stock products in a live feed don't announce themselves. Performance dashboards show them as ads that once worked. The waste only becomes visible when someone connects inventory data to ad spend — and by then it has been accumulating for months.

The bigger the catalogue, the more the long tail costs you. Every retailer knows a minority of products drives the majority of revenue. Fewer act on it in their feed, because acting on it requires product-level return data rather than campaign-level data. Without that, budget distribution defaults to the platform's judgement rather than yours.

In high-AOV categories, the maths is harsher. A wasted click on a ₺200 product is an annoyance. On a ₺40,000 sofa, the cost per click is high enough that a small percentage of misdirected impressions becomes a meaningful line item. Categories with expensive products and long consideration cycles have the most to gain from feed discipline.

Variants are where feeds break. Colour, size, fabric, orientation — every dimension multiplies the catalogue, and every variant carries its own stock level. Feeds that treat a product family as a single item advertise products that are partly unavailable. Feeds that handle variants properly advertise only what can be shipped.