Sustainability in software development means treating environmental responsibility as a quality attribute in its own right when making development decisions. The Pink Index rates products across seven dimensions: manufacturing quality, necessity, frequency of use, durability, energy consumption, end-of-life recovery, and strengthening engagement. Applied to software, that translates into caching, image compression, lean cloud sizing, and deliberate feature design.
Key Takeaways
- The Pink Index rates products on more than CO2 equivalents. It uses seven dimensions: manufacturing, necessity, frequency of use, durability, energy consumption in operation, end-of-life recovery, and strengthening engagement.
- Training your own AI models uses far more electricity than running them later, so companies should question whether they really need their own GPU cluster.
- Computers draw substantial power even when idle. Sizing the runtime environment conservatively is more sustainable than provisioning generous, underused nodes.
- Many software features are rarely or never used, yet they keep consuming compute, memory, and bandwidth, which makes them an avoidable sustainability problem.
- A goal that looks out of reach stops people from taking the first step. Small improvements from many people add up to more than perfect solutions from a few.
Green IT and Green by IT Are Two Different Levers
When people talk about sustainability in software development, the conversation usually lands on green IT: building software that needs less memory and less power, so data centers draw less energy and less data travels across the network. That makes sense, but it covers only one side.
Green by IT asks a different question: what can software do for sustainability as a whole? Dominik Rost and Marcus Trapp see the bigger potential there.
Fleet planning is a simple example. If software helps a company get by with one car less, that pays off for the environment and for the budget. Nobody needs to debate win-win cases like that for long. You just do them.
What the Pink Index Measures
The Pink Index rates a product across its entire life cycle instead of calculating a single figure to the eighth decimal place. The name comes from the German for “pragmatically precise information on sustainable consumption,” and the index belongs to the company Full Flamingo.
The usual approach in Germany digs deep into CO2 equivalents: manufacturing, packaging, delivery, component supply, and raw materials. That is thorough, but it ignores many other dimensions.
The Pink Index turns the logic around. Instead of getting ever more precise in one dimension, it looks at all of them with pragmatic precision. That brings in questions beyond CO2:
- Do you really need the product, or is it a convenience or luxury item?
- How often do you use it?
- How durable is it? Is it a throwaway product, or does it last?
- How much energy does it need, not to make, but to run?
- What happens at the end? Is it recycled, taken apart, or passed on?
Dominik and Marcus call one separate, almost esoteric dimension “strengthening engagement.” It describes how a product helps you behave more sustainably yourself. A water-saving shower head is the example: even if its manufacturing is not ideal, you save water every time you use it.
How the Index Gets into the Shopping Cart
Full Flamingo runs neither its own shop nor a public calculator. It plugs into existing online stores instead. The sustainability rating for your specific purchase shows up in the cart, in the same area where you might be offered insurance for an electrical appliance.
The rating displays the seven dimensions of the Pink Index as a traffic light system, so you get an immediate sense of how sustainable your purchase is.
An amount is tied to the rating, roughly 4 to 6 percent depending on the purchase. You can donate it to a regional organization that drives sustainability locally. It is opt-in, with no subscription and no obligation.
The regional focus is the goal, not where things stand today. Germany has thousands of repair cafés, urban gardening projects, and food-sharing initiatives, often with little visibility and no common organization. What they all need is money. Dominik and Marcus see everyday shopping as the biggest lever for getting it to them.
How to Make Software Development More Sustainable
A sustainability company has to hold itself to its own standards. That is why Dominik and Marcus apply the logic of the Pink Index to their own software development. They see themselves as a tech company for sustainability, so looking inward is part of the job.
Applied to software, manufacturing becomes everything that happens at development time. Much of it has nothing to do with code at first. Does the company buy genuine green electricity? In Germany, only about ten providers actually supply it. Are the lights on all night in empty offices?
Closer to the technology, there is the question of test systems. Does every single commit have to go through the entire test infrastructure, even when only the README changed? From a testing perspective that may make sense. From a sustainability perspective, not automatically.
Training Your Own AI Models Is a Question of Power
Training your own models takes an enormous amount of electricity. AI already uses a lot of energy in everyday use, and training the models takes far more.
Before everyone sets up a GPU cluster of their own, it is worth asking whether that is really necessary. The reflex to build something in-house because everyone else is doing it comes at a high price in sustainability terms.
Why Conservative Cloud Sizing Is Often Better
Computers draw substantial power even when they are doing nothing. So it is better to run machines at high utilization than to let several nodes in a cluster idle along at low base load.
Cloud providers make it easy to book the next bigger CPU or more memory. Developers like the buffer because everything runs smoothly. For sustainability, the opposite approach works better: start with conservative sizing and add capacity only when real performance problems show up.
The same goes for unused hardware. Leftover budget at the end of the year quickly turns into computers nobody is sure they need.
Every Feature Costs Energy for as Long as It Exists
Many features built into software are rarely or never used. Studies point in that direction. Once built, a feature usually stays, even though it takes up CPU, memory, and data transfer.
The same applies to technologies and data. Do you really need another Redis instance next to the existing stack, or will something you already run do the job? Do you have to collect as much data as possible, data lake style, because you might find an interesting correlation for a dashboard later? Every bit of data you collect, move, and store costs energy.
Small Effort, Big Savings: Data over the Wire
Between server and client there is no simple wire. There is a chain of DNS servers, routers, and firewalls, and every one of those devices has to run and process whatever data passes through. The less you send, the better.
Small measures pay off quickly here:
- HTTP caching takes manageable effort to set up.
- Image compression and responsive images cut the amount of data transferred considerably.
- Static HTML pages often replace a full content management system.
The static page is a good example of unnecessary complexity. Many business websites don’t need a CMS with a database. In one case at Fraunhofer, the only reason left for a CMS was that each branch had different opening hours.
Web standards today are powerful enough to deliver a good user experience with HTML and CSS. A static page is fast, secure, and gets by without much JavaScript.
What Happens at the End of a Device’s Life
The last dimension asks where a product goes once it is retired. In IT, you see it in the piles of old notebooks that build up because new devices were bought, even though the old ones would still handle normal day-to-day business.
Throwing them away is the worst option. Proper recycling is the minimum. A second life is better: donate them to schools or clubs, or send them to the Global South through organizations such as Labdoo.org, which transport, repair, and refurbish the devices in a CO2-neutral way.
The sharing economy and the circular economy follow the same idea. A lawn scarifier gets used once or twice a year, yet one sits in many basements. Products like that are made for borrowing. Second-hand platforms give things a second or third use once supply and scale fit.
Sustainability Is a Quality Attribute with Trade-Offs
In software development, sustainability belongs next to a system’s other quality attributes and should be weighed explicitly. The point is not to do everything at once. The point is to consider it at all.
Not every measure clearly pays off. Moving workloads to a region with green electricity sounds good, but it takes a lot of effort and means shipping data and logic halfway around the world. Whether it is worth it depends on the context. There is no one-size-fits-all answer.
The first step matters more than a perfect vision. If you set up an unreachable ideal, you make sure nobody gets started.
“If each and every one of us makes a small contribution somewhere, then in the end we’ll all be better off.”
(Marcus Trapp)
Full Flamingo’s own website shows that the principle works on a small scale too: 92 images, displayed large rather than as thumbnails, add up to 2.72 megabytes in total. Thought through early, easier to keep lean later.
Frequently Asked Questions
Why isn’t the carbon footprint alone enough to evaluate a product?
Because it calculates only one dimension with ever-greater precision while ignoring others. The approach commonly used in Germany delves deeply into CO2 equivalents for manufacturing, packaging, shipping, and raw materials. The Pink Index, used by the company Full Flamingo in 2023, turned this logic on its head: seven dimensions assessed with pragmatic precision, including necessity, frequency of use, durability, energy consumption during operation, and end-of-life recycling.
Which has greater impact: more efficient code or software that reduces emissions elsewhere?
The greater potential lies in software that saves resources elsewhere, in other words, “Green by IT.” “Green IT,” which focuses on reducing the application’s own storage and power consumption, covers only one aspect. One example is fleet planning: If the software eliminates the need for one car, it benefits both the environment and the bottom line. No one needs to debate such cases at length.
What does “strengthening engagement” mean as a sustainability criterion?
It describes how a product supports its users in behaving more sustainably. Take a water-saving showerhead as an example: even if its manufacturing process isn’t ideal, every single use saves water. This dimension falls outside of pure manufacturing assessments because it considers the effect during the usage phase and in user behavior.
Where does sustainable software development begin, if not in the code?
It begins with everything that happens during development time. This includes the question of whether the company uses genuinely green electricity: In 2023, there were only about ten providers in Germany that actually supplied genuine green electricity. It also includes lights left on at night in empty offices. Closer to the technical side is the question of whether the entire test infrastructure needs to run with every commit, even if only the README file has been changed.
Is it worth training AI models in your own data center?
Only after careful consideration. AI already consumes a lot of energy during operation, and training the models requires significantly more electricity. Before setting up your own GPU cluster, you need to ask whether it’s actually necessary. The impulse to set something up just because everyone else is doing it is costly from a sustainability perspective.
Should cloud resources be provisioned generously or conservatively?
Conservatively. Computers consume a substantial amount of energy even when they’re idle, so a fully utilized machine is better than multiple nodes with a low baseline load. Cloud providers make it easy to upgrade to the next-largest CPU or more memory, and developers like having that buffer. It makes more sense to take the opposite approach: start with a conservative allocation, then scale up only when actual performance issues arise.
Why are rarely used features a sustainability issue?
Because they constantly consume CPU, memory, and bandwidth. Studies suggest that many built-in features are rarely or never used, and once built, a feature usually remains in the product. The same applies to technologies and data: Do we really need an additional Redis, and does data that might later reveal an interesting correlation need to be collected in a “data lake” style?
What measures reduce the amount of data transferred with manageable effort?
HTTP caching, image compression, and responsive images deliver quick results. Often, a static HTML page can replace a full-fledged content management system, because there is a chain of DNS servers, routers, and firewalls between the server and the client that must process every bit of data. In one case at Fraunhofer, the only reason for using a CMS in the end was that each branch had different opening hours.
What should you do if a sustainability measure does not yield a clearly positive result?
Like any other quality attribute, it must be explicitly weighed. Shifting workloads to a region of the world with green electricity sounds good, but it requires a lot of effort and means transporting data and logic all the way around the world. Whether it’s worth it depends on the context; there’s no one-size-fits-all solution. The first step is more important than a perfect end goal.
