Fast vs Balanced Cutting Optimization: When to Switch

Most cut-list software gives you one optimizer and one speed. CutOptim gives you a dial. The question this guide answers is when to turn it — and, more often, when to leave it exactly where it is.

What the Setting Actually Controls

The default optimizer is a guillotine multi-strategy search: it generates many candidate layouts — several packing strategies, sort orders and first-cut directions — scores each one, and keeps the best. Most of that work is cheap. One step is not.

That expensive step is a per-board re-search: after an initial layout is found, the optimizer goes back and tries to repack individual boards to squeeze out the last few percent of yield. On a small job there is almost nothing to gain there, so it finishes instantly. On a large, varied job with hundreds of distinct sizes, it does real work — and takes real time.

The reason the cost grows so sharply with size is that the number of ways to rearrange parts on a board explodes as parts are added. A handful of parts has a handful of arrangements worth trying; a few hundred distinct sizes has an enormous number, and the re-search has to look through many of them to find the marginal gains. That is why the step is invisible on an everyday job and dominant on a big one — the same code, doing far more work because there is far more to examine.

Speed vs density is the switch for that one step. Balanced runs it. Fast skips it. Everything else about the search — the strategies, the guillotine rule, the scoring — is identical between the two.

The Measured Trade-off

Generic percentages are worthless, so here is a single measured job rather than a marketing claim. On a demanding run — about 1,550 parts on a 2.07 × 5.6 m board — the two settings came out like this:

Balanced spent roughly two and a half times as long and got twenty fewer boards’ worth of density out of it. Whether that trade is worth making depends entirely on the size of the job and how the seconds feel while you wait.

The chart plots the same trade-off: Fast sits lower and to the left (quicker, slightly less dense), Balanced higher and to the right. On the small-to-medium jobs most shops run every day, the two points sit almost on top of each other — the expensive step Fast skips has little to find, so skipping it costs almost nothing and saves almost nothing.

Fast Is Never Denser Than Balanced

This is the property that makes the choice safe. Fast searches a strict subset of the layouts Balanced searches. It considers the same candidates minus the one expensive re-search, so mathematically it can only ever match Balanced’s yield or fall a little short — it can never beat it.

That has a useful consequence: you never have to A/B the two settings hoping Fast got lucky. If Fast gives you a plan you are happy with, Balanced would give you the same plan or a marginally denser one. On a budget-limited large job the two can even come out identical, because the search is capped before the expensive step would have changed anything.

When to Choose Fast

Reach for Fast in one situation: a very large job is taking long enough to interrupt you. Concretely, that usually means several hundred parts or more with many distinct sizes — the shape of job where the per-board re-search does the most work and therefore costs the most time.

• Iterating on a big job. If you are adding parts, changing kerf and re-running a several-hundred-part cut list repeatedly, Fast keeps each run snappy while you experiment; switch back to Balanced for the final plan you send to the saw.
• A rough answer, right now. If you need a board count and a ballpark yield to quote a large job on the phone, Fast gets you there in a fraction of the time, and a couple of points of yield will not change the quote.
• Batch or automated runs. Through the Engine API, where many large jobs are solved back to back, Fast can be the sensible default for throughput — with a Balanced pass reserved for jobs that actually go to production.

For everything else — the daily cabinet run, the small batch, the one-off — leave it on Balanced. A few dozen to a couple hundred parts solves in well under a second either way, so there is nothing to buy by going faster and a little yield to lose.

Where to Find It

In the app, open Advanced settings and set Speed vs density to Balanced or Fast. The choice is per project and applies to 2D panel jobs; the default is Balanced.

On the Engine API, the same choice is an effort option on the POST /v1/optimize/2d request — "balanced" (the default) or "fast". It changes only how hard the default optimizer searches; the response format, the guillotine guarantee and everything else are unchanged. See the Engine API reference for the request shape.

Not the Same as the “balanced” Engine

One naming collision is worth clearing up, because it trips people up. On the Engine API there is a separate balanced engine — a completely different optimizer that uses MaxRects free nesting instead of the guillotine search. That is not what this setting is about.

• Fast and Balanced (this guide) are two search depths of the default guillotine optimizer. Both produce saw-cuttable plans.
• The balanced engine is a different packer you opt into on the API. Its layouts are often not guillotine-cuttable, so it is for callers running CNC routers or lasers, not panel saws.

They are unrelated choices. If you cut on a panel saw, you want the default optimizer at either Fast or Balanced depth — never the free-nesting engine.

Try It on Your Own Numbers

The best way to feel the trade-off is on a job of your own. Enter your parts and stock below; on a small job you will see the two settings agree, and the difference only opens up as the part count climbs.

For a full cut list with kerf, trim and exports, open the optimizer, or start from the panel cutting optimizer.