As most people have said, estimates of calorie burn by fitness devices/cycling computers/etc are really not very accurate, for a variety of reasons - with the biggest variation coming from individual body weight and fitness.
One thing in particular I've always considered useless is calorie counts on gym exercise machines - they have no clue about the weight of the exerciser, which is the biggest contributor to calorie burn.
Having said that, I think it can still be very useful, but only when consistently using a single device and comparing it against itself. You can get a feel for day-to-day energy usage, even if the actual calorie count is not accurate. And I think the most useful devices for it are 24-hour wearable fitness trackers.
I wear a fitbit tracker, which has 24-hour tracking of my heart rate and all physical motion, it knows my age, and I tell it my weight every week so it can adjust for that. When I have a "moderately active" week (as in the usual calculator speak), it tends to estimate around 2,700 kcals usage per day (which is 100 kcals more than an online calculator I just tried).
I set myself a target to average 500 kcals more than that per day (fitbit kcals - they're arbitrary units, so it doesn't really matter if they're not quite the same as real kcals). I also try to keep my daily kcal intake (based on estimates for each food item) to 500 to 1,000 kcals less than my fitbit kcals burn rate.
I suspect the fitbit overestimates kcal burn when exercising hard as I have some surprisingly high-cal burn days on longer or harder rides (or harder gym sessions). But the net result of my approach is that I'm losing weight at around 0.5kg per week, and my fitness is definitely improving (rest HR is down to around 56bpm, from around 64 about a month ago, and in the 70s at New Year).
So, yes, if you treat it as an arbitrary number, and use the same device so that the arbitrary number is consistent, it does give you something to measure and compare against over time.
