We here at BikeRadar have spent lots of time testing different power meters over the years, and much of that has been dedicated to evaluating their accuracy and precision. A power meter's readout should precisely quantify your output in watts, and that figure should line up with other meters, right?
That may be true, but ultimately neither really matters for most of us. The most important thing, like with a bathroom scale, is day-to-day consistency so you can track performance relative to your past efforts.
Power meters are complex and intricate little widgets, as they damn well better be given their asking prices. Yes, at their most fundamental state they're typically just a bunch of strain gages bonded to chunks of aluminum and connected with some wires to circuit boards, batteries and antennae. If that's all they were, however, every DIYer with a soldering iron, Dremel tool, Krazy Glue, and a Digi-Key account would be rolling around on a homemade PowerTap.
It's how those gages are arranged, the way those chunks of aluminum are shaped, and the elaborate algorithms used to process the resultant signals that require so much R&D work. Nearly all of that effort is directed toward making these things as accurate and repeatable as possible – a watt should be a watt should be a watt given that power is far and away the most important metric when it comes to measuring one's performance on a bike.
That accuracy is what you're paying for – and what you're convinced you need to have.
My functional threshold hovers around 99 Awesomes. What's yours?
Yet for many everyday users, power meters serve but one function: to quantify your output for short- and long-term analysis. Does it really matter if the 'real' number is 230 watts or 240? Or does it only matter more that you were putting out ten more watts at threshold today than you did two weeks ago, and that those figures are reliable over a wide range of conditions?
In that sense, does it even matter that a power meter displays output in watts at all? Couldn't we just use some arbitrary figure like 'Owwees,' 'Awesomes' and 'Donuts'? Think about it: you could still quantifiably prove that today's ride hurt much more than yesterday's (because you were working harder), that you really were more awesome on the bike that day than the day before (because your 'awesome' meter said so), and yes, you do deserve that extra donut after that mega-hard ride.
Either way, the usefulness of the device in terms of training is the same. In that context, it doesn't matter where that needle is pegged; only where it is today relative to where it was before.
There are downsides to this scenario, of course. Precision is the key element in terms of training, but ultimate accuracy – plus a standard unit of measure – is still needed to compare one individual's performance to that of another rider. You can say definitively that your 3W/kg threshold won't set any records on Mont Ventoux, for example, but you can't say that your 547 Awesomes trumps your buddy's 322 Owwees.
Ultimate accuracy is also necessary to provide some sense of where you sit in the general pecking order. So you cranked out a personal-best 3.7 W/kg on your local climb? That's fantastic – but it's still only about half what a top pro can muster.
Variations in ultimate accuracy have implications if you switch from one power meter to another, too. These days, the range of variability is pretty small and certain types/brands of meters seem to trend consistently. In other words, if you started out with a crank-based meter and decide to upgrade a few years down the road, for example, I'd recommend sticking with a similar system to reduce the number of variables.
So what's the takeaway? Most consumers shouldn't get too hung up on how ultimately accurate a power meter might be if you're primarily using it for training. Comparing FTPs with your buddies is nice and all but all any of you really care about at the end of the day is who got to the town line sprint first anyway.
Be realistic about what your needs and goals are for a power meter and then budget from there. Weekend warriors whose main goal is completing that annual charity ride probably don't need the four-data-packet-per-second transmission rate of an SRM. As long as your choice is internally consistent and repeatable, that'll still get the job done for most of us. Make sure that with whatever meter you're using currently or decide to purchase that it's set up properly, too, and that it's zeroed out and/or calibrated as needed to ensure repeatability.
And either way, I'd still recommend that extra donut.