SEPTEMBER 2026 — Edition #8

the

pixel

insider
by WEB DESIGN TEAM

September was about doing more of the work ourselves. We stopped buying stock images and stopped paying for background removal, replacing both with tools built inside the team, with Francisco's cut-out engine going live on the same day the old service ended. Acquisition started ordering every marketing piece through our platform, and the team delivered 1,035 tasks, the busiest month of the year so far.

From stock images to images made in-house
KNOW MORE +
Images AI Cost
From stock images to images made in-house
Problem
Stock imagery cost €5,369 a year, was capped at 350 images a month and rarely showed the products we actually sell.
Solution
The team now generates lifestyle, service and category images in-house, with templates that keep every product family consistent.
Impact
344 images in two weeks
The stock subscription is suspended. The team produced 344 images in two regular working weeks, at about a tenth of the cost per image.
Background removal built by Francisco
KNOW MORE +
Cut-out In-house Images
Background removal built by Francisco
Problem
Cutting products out of their background depended on remove.bg, a paid external service that was not renewed, and three apps relied on it.
Solution
Francisco Ferreira built an in-house engine that removes backgrounds directly in the browser, now hosted on our hub and connected to Banners, Peças MKT and Images.
Impact
97% clean cut-outs on common products
No licence, no cost per image, and the image never leaves the computer. Since 29 September every product image is cut out automatically.
Acquisition orders pieces through Peças MKT
KNOW MORE +
Peças MKT Acquisition Workflow
Acquisition orders pieces through Peças MKT
Problem
Marketing pieces were requested by message and rebuilt by hand, with the rules living in a few people's heads.
Solution
Since 14 September, Acquisition orders, previews and approves every marketing piece in Peças MKT, and Design produces from the same request in Asana.
Impact
One flow for both teams, and the improvements now come from the people who use it every day.
Apple configurator in one click
KNOW MORE +
Apple Metis Automation
Apple configurator in one click
Problem
Updating the Apple configurator took two scripts and 21 copy-pastes into the CMS every time the commercial team sent a new file.
Solution
A new app reads the commercial Excel, checks it and publishes all seven highlights with one button.
Impact
21 → 1 manual steps
Faster updates, and a broken file is now stopped before it reaches the store.
Enriched PDPs for a second brand
KNOW MORE +
PDP Marketplace AI
Enriched PDPs for a second brand
Problem
Enriched product pages were built one by one, and each new brand meant starting from zero.
Solution
PDP Lab now serves Becken and Merach, a marketplace seller, from the same flow: SKU in, enriched content and images out, published to the CMS.
Impact
+20% buy-to-detail on first Becken PDPs
A repeatable model for more brands before peak season.
Francisco Ferreira
Francisco Ferreira
Replacing a paid service

Built the background-removal engine that replaced remove.bg in three apps on the same day the old service stopped, running in the browser with no licence and no cost per image.

Vítor Cardoso
Vítor Cardoso
Peças MKT in real use

Main Design tester of Peças MKT from its first week with Acquisition. Caught a piece that would have shipped without products and drove multi-select, zoom and the KV version, while closing 199 tasks in September.

António Pinto
António Pinto
Leading the image tools

The most active user of the team's image tools, with more than a third of all image and cut-out calls in the first weeks of measurement, and 113 tasks closed in September, four times the July volume.

Team Activity
KNOW MORE +
Team Activity
1,035 tasks delivered in September
September delivered 1,035 tasks, 17% more than August and 26% more than June, with 897 owned, the highest of the year so far. AI prompt images became the second largest project of the month, right after the weekly board, followed by GW24 Crédito 10x and Banners App AI.
Images made in-house
KNOW MORE +
Images made in-house
344 AI images in two weeks
For the first time we can measure image production. In two regular working weeks, seven people across four apps generated 344 images and 191 background removals, more than double the monthly ceiling of the old stock plan.
2,702 calls in four weeks
Over the following four weeks, from 4 September to 2 October, the counter recorded 2,702 calls from 13 people: 2,123 to the image AI and 579 background removals through the paid service, most of them before Francisco's in-house engine replaced it at the end of September. PDP Lab led with 1,170 calls, followed by Images (957), Banners (327) and Peças MKT (248).
01
A lost supplier can be a push

When remove.bg was not renewed, the in-house engine was ready and performed better on the hardest case.

02
New users find the real bugs

Experienced users learn to work around problems. People on their first day report them.

03
Check the delivered file

A piece can look right on screen and still ship empty. The final file is the proof.

04
Input files are a contract

A small change in a source spreadsheet can silently break a page, so the tool must check it first.

🛒
PDP
Merach catalogue before peak season

Deliver the seller's enriched catalogue by the end of October and turn it into a model for other top sellers.

📊
Data
Measuring time and cost saved

Count pieces produced and time saved per app, measure the real cost of image generation, and test the cut-out engine on Windows PCs.