Here's an example:
Practically all entry level chinese multimeters have manual ranges. I wonder, how having a big rotary switch with so many contacts is cheaper than having a simpler switch and letting an IC take care of the ranges?
The switch is made of molded plastic parts. The case, the cam (if any) and the knob. Any number of positions can be molded in, at no cost to the production of the molds or parts. The contacts are spring metal on PCB, and the PCB of course can be made with any number of positions etched on it, at no added cost.
The cost of selector, case and PCB is borne by all competitors in that space (you can't have a multimeter -- at least one marketably recognizable -- without any of these three fundamental aspects), so the cost comes to materials reduction (what's the cheapest per strength resin, and how little of it can they get away with?) and BOM cost. Saving half a cent on the non-autoranging ASIC is the clear winner, then.
And that chip, in turn, is probably something like an ICL7107 -- ancient LED/LCD display driver with analog input, with the added features of volts/amps/ohms ranges (depending on how many resistors they afforded to place around it). Such chips have been around for decades, so basically no one is paying NRE on them, they're as cheap as the silicon they're made on.
Which is a general rule of semiconductors -- compare an IRF540 to a modern device of the same (Vds, Rds(on)) ratings; the new device has far smaller capacitances because of process optimization and miniaturization. The die might be 1/4 the size, but the cost is the same, because just damned everybody makes those ancient parts!
I recently saw a nice looking radio-MP3 player in an aluminium enclosure on ebay. It was selling for just $10, shipping included!
If I wanted to make something similar myself, 10 bucks would just barely cover the cost of the enclosure alone. How can the Chinese sell stuff that cheaply?
NRE: enclosure for example. Maybe you'd spend $500 getting the 3D model machined from bulk aluminum. Or $200 for a handful of CNC cut / bent parts. Or between, say, $0-10 per part getting them molded (sand castings), but you still need $200+ for the machined or molded (stereolithography) pattern to start from. Or upwards of $20,000 for a real serious die set, made of machined and polished tool steel, for pressing/punching (sheet) or molding (die cast) parts. Now, this last method, making the dies, has per-part costs hardly above the cost of material alone, so it's the obvious winner in quantity. But there's no freaking way you can get anywhere close to that in single quantities, even if you spend three months with a die grinder making the molds in your free time!
Supply chains: volume pricing. In large enough quantities, parts may come in straight from the manufacturer, no distributor middle-man/men at all. This saves usually 40% straight away (business / retail markup), plus distributor inventorying and transportation. Though it also costs your own production's stocking overhead and all that.
Manufacturer preferred pricing. Often, one manufacturer will attempt to sell you their alternatives to competing parts you're already using. And they will undercut their catalog prices for you, even at moderate quantities. One example I've seen: DAQ board, 12 bit 80MSa/s ADC, two channels, migrating to four. Old part: National, 3.3V CMOS, dual, 0.8W, like $40 in 100s. Proposed alternative: Analog Devices, 1.8V CMOS, quad, 0.3W, like $37 in 100s. Digikey stock price was something like $60 each -- still a win over using two of the old device, but the new price (which was quoted through Arrow I believe) was unequivocally better, besides the other gains like power dissipation (not that that was a concern on this project, but it still helps with thermal management and routing).
Tim