Thinking tool: Shorten cycle times, reduce batch sizes

Headshot of Adam Jones

Adam Jones

Most people reading this will be doing work that is in some way "new" or "creative" - not necessarily in the painting-and-berets sense, but in that the exact thing you're about to do hasn't been done before. For example, you might:

Meme: Willem Dafoe as Norman Osborn in Spider-Man (2002), smiling, captioned "You know, I'm something of a CREATIVE myself"

In this kind of domain it's easy to go off track, and the situation itself keeps changing, so reacting fast is valuable.

Short cycle times and small batches massively help here: they cap how much time and material you invest in potentially the wrong thing.

Reducing cycle time also tightens the feedback loop, so you can self-improve and react to what users are telling you quicker.

Small batches also let you pay more attention to quality, which is particularly useful in product development: it's much better to build something a few people love than something a lot of people find okay. (Especially as the space for generic solutions is usually already occupied by an incumbent.)

All of this is common in startup literature, agile methodology and change management theory. Software practices like Kanban set work-in-progress limits to optimise overall flow and end-to-end latency (which also helps optimise the constraint). Reducing work in progress eliminates a bunch of other costs too: PR merge conflicts, bugs and incidents from complex parallel changes, and general team cognitive load.

Nor is any of this new to software: traditional manufacturing advice also focussed on adapting to change and minimising work-in-progress inventory (which ties up money in materials and storage).

Examples

This technique works across lots of domains - startups, canning, and the race to the first aeroplane.

Y Combinator

Y Combinator's first piece of advice to every startup is "launch now": because "launching a mediocre product as soon as possible, and then talking to customers and iterating, is much better than waiting to build the “perfect” product". Then:

"Once launched, we suggest founders do things that don’t scale. Many startup advisors persuade startups to scale way too early. This will require the building of technology and processes to support that scaling, which, if premature, will be a waste of time and effort. This strategy often leads to failure and even startup death. Rather, we tell startups to get their first customer by any means necessary, even by manual work that couldn’t be managed for more than ten, much less 100 or 1000 customers. At this stage, founders are still trying to figure out what needs to be built and the best way to do that is talk directly to customers."

This echoes the advice here: if you consider your cycle the standard lean startup "build -> measure -> learn" cycle, a short cycle time means shipping things to users, getting feedback and learning from it very quickly. And similarly a small batch size (for one customer) is preferable to scaling the wrong thing too early.

Crown Cork & Seal

Looking down on rows of identical unopened drinks cans packed tightly together, filling the frame

Most of the canning industry is built around long, efficient production runs for giant customers.

Crown realised that competing here wouldn't work. Instead, they took the short-run, hard-to-handle and rush orders, and optimised its plants for low setup costs so that small batches were profitable.

The big producers couldn't follow, because their whole strength - long-run efficiency at scale - came with huge setup costs. This is often also given as a case study for strategy that uses leverage against competitor weaknesses.

The Wright brothers

In the race to the first aeroplane, Samuel Langley worked in one big batch: over $50,000 of War Department and Smithsonian funding went into a single grand, full-size flying machine, tested by catapulting it (and its pilot) off a houseboat roof. It crashed into the Potomac twice in 1903, and the project ended there.

Langley's full-size flying machine, a slender monoplane with long curved fabric wings, sitting on a catapult track atop a two-storey houseboat superstructure. Photo: Library of Congress, public domain

Photo: Langley's full-size flying machine

The Wright brothers by contrast spent a fraction of that, but cycled much faster: kites, then three successive gliders, hundreds of short test flights, and a homemade wind tunnel that let them test dozens of wing shapes in weeks. Each pass through the "build -> measure -> learn" cycle allowed them to make progress. Nine days after Langley's final crash, they flew the first successful aeroplane.

Trade-offs

Shorter cycle times and smaller batch sizes usually reduce immediate throughput: more time goes on planning and coordination, and you get less from economies of scale.

The trade-off is well worth it in new, uncertain spaces. Scaling up production of the wrong thing is a much more common mistake than improving your solution for too long.

That said, if you're very confident in exactly what to produce and how much, larger batches are reasonable. 3M producing tape in normal times can sensibly run large batches and slow cycles; 3M producing FFP2 masks during the COVID pandemic would want smaller batches, since demand and product-market fit are much less predictable.

This is part of my thinking tools series. Also consider: