Doing things that don’t scale and why the lighting industry still has to, but when is it too much?

‍The Case for Doing Work That Doesn’t Scale

In 2013 Y Combinator co-founder Paul Graham argued that the most important thing for a new company isn’t building automated systems but doing whatever it takes to make customers happy, even if those tasks “don’t scale.” Manual, unscalable work teaches you what customers actually need, allows you to start immediately without a huge investment in automation and often costs less while you iterate. Graham pointed to the early days of Airbnb, where the founders personally photographed apartments, and to Jeff Bezos packing and driving books to the post office. The point was not to stay manual forever but to learn enough to automate the right things later. Sketchplanations summarised the benefits of this approach: it helps you build only what’s needed, enables quick experimentation and lets you discard ideas that don’t work.

‍

Complexity Behind a “Simple” LED

The lighting industry has lived in this world of unscalable work for decades. A “simple” LED tape light is actually a flexible printed-circuit board lined with LEDs; it often has an adhesive backing so it can be placed into a metal extrusion and can be cut to length. Once you start specifying a project you discover that there are hundreds of possible variables. There are end-feed configurations, tape widths, colour temperatures and outputs, and then there is the extrusion: profiles may be suspended, surface-mounted or recessed, and each profile supports a catalogue of lenses.

Some product families, for example, offer lenses that sit proud or flush, clear lenses that maximize light, grazing lenses, optical lenses that control the beam, encapsulated outdoor-rated versions, louvers and even micro-louver versions. Others go further, offering a dozen or more different lens options, multiple beam angles, asymmetric lenses, forward-throw lenses and varied levels of diffusion. When you combine that with diffusers (frosted, prismatic or lens types) that change beam shape and reduce glare, drivers, wiring, mounting hardware and custom lengths, the possible combinations run into the thousands.

‍

Expanding Options and Complexity

The rise of LED systems has only accelerated this complexity. Suspended lights often incorporate track systems with modular heads that can be swapped or repositioned. Downlights now come in countless interchangeable trims, optics and beam spreads. Custom curved extrusions or flexible systems allow designers to bend light into unique architectural forms. Each of these advancements increases creative potential but also multiplies the specification challenges.

Historically there were far fewer choices. Some fixtures still reflect that simplicity, offering just a handful of lens options with a single flat profile. The appeal of such products is obvious: having only a few options simplifies specification and reduces the chances of ordering mistakes. Designers might fondly recall a time when a cove light meant picking a wattage, a colour temperature and perhaps a frosted or clear lens. Today, specifying a linear run requires a spreadsheet.

‍

Why Customization Doesn’t Scale

The calculus of all these variables does not scale well. Each custom configuration demands calculations: verifying LED strip wattage against driver capacity, checking lens compatibility with the extrusion depth, and ensuring that the end-feed and mounting clips match the installation. Because there are hundreds of permutations, manual quoting and ordering often result in errors. Complex LED installations can suffer from flickering and power issues if the wiring and components aren’t set up correctly, and poor-quality lenses and reflectors can reduce efficiency. Ordering guides and spec sheets exist to help, but the underlying complexity remains. Unlike software, you can’t refactor a cut piece of extrusion, you have to reorder it.

‍

Manual Work as an Industry Standard

What does this have to do with doing things that don’t scale? In a sense, the lighting industry never left that phase. Custom projects still require humans to coordinate thousands of variables. This hand-holding is expensive but it adds value: designers get exactly what they want, and manufacturers learn which combinations work and which don’t. Over time, those lessons inform better systems. Some companies are now building configurators and modular approaches to reduce errors, but they only exist because of years of unscalable, bespoke projects.

‍

Balancing Simplicity and Possibility

So is there anything wrong with the “old days” of just a few lens options? Simplicity makes life easier for engineers and electricians, but it also limits the designer’s ability to shape light. The proliferation of options reflects how far the industry has come: linear LEDs are now used for grazing, indirect uplighting, tunable white, RGB effects and more. The frustration isn’t that choice exists but that our tools haven’t fully caught up to manage it. The answer isn’t to wish away options but to embrace the early-stage startup mindset. Do the manual work to understand what designers need, then automate wisely. Unscalable effort is a feature, not a bug. The end will come when systems are robust enough that specifying a linear light feels as straightforward as those “old days,” even though the underlying choices are far richer.

‍