You may recall I asked a little while back for Carrera serial numbers to help advance something I was working on. Well, those have dried up now, so I'll share some initial results in the hope of eliciting a few more, on the basis that the more we have, the more accurate we can be.

I'll also stress that, in the absence of solid production figures from Heuer itself, we are reduced to... well, not quite guessing but deriving numbers from known serials. As with extrapolating any target number from a known, but smaller, sample size, there are some margins for error and a number of assumptions have to be made. But I have proceeded as rigorously and consistently as I could, given a sample of 199 Carrera and 354 other serial numbers.

The method basically assesses numbers in a given range. It is quite clear from the numbers we do have that cases are batched by model pretty consistently, which makes dating accurately by comparing one model with another a little trickier than it could be. However, the only occasion where the batch sizes and model offsets are very large and could cause real dating confusion is a known period, with the launch of the Chronomatic movements, helpfully minimising the issue.

For the most popular models by volume (the Carrera and the Autavia basically), batches are quite often in the thousands. Smaller models are often in the hundreds but also sometimes in quite large batches too - in this instance, we can see that very few batches are ever made of some of those models. Whereas there are at least 6 large batches made of the Carrera 1153, as an example. In the process I assumed that the complete 1000 of each block of serials was used - not too far a stretch, we have serials in the low 00xs as well as the high 99xs, but a possible scope for some overestimation. Where another model shares a block of 1000 with a Carrera, I have assumed that the Carrera batch ends at the 100 preceding the other model (so if we have a Carrera xxx357 and a Verona xxx654, I assumed that the Carrera ran to xxx599). Again some potential for overestimation there, but still fairly minor in the scheme of things - I don't think there is more than say 10% scope for overestimation.

There is potential underestimation too. There are several ranges of 1000 where we don't have any serials at all. I have not made any guess as to what these ranges contain - even where they are bracketed by known models, it is a potential shaky assumption to count the "missing" ranges as belonging to the same model. The scale of this potential underestimation is larger than the potential overestimation, but still within reasonable limits - I estimate somewhere around up to 15%. So the margin for error is somewhere around +15%/-10%.

Before I finally finish outlining my method and start talking numbers, I'll talk about one more number I used as a check and balance. The exercise is largely about assessing and assigning batches of serial numbers, which inherently includes a fair degree of extrapolation. Against this I also use a measure which looks only at known serial numbers to derive an absolute minimum number produced. This makes no assumption about the potential maximum size of a batch, just subtracting the lowest number in a range from the highest. For example, if we have one Carrera model at xxx157 and one at xxx356, this number assumes the batches stop and start with those numbers. If there is only 1 serial known in a potential batch range of 1000, it assumes that a batch of a single watch was produced. This isn't going to be the case in reality, but this number deliberately underestimates to provide a minimum baseline to compare against the extrapolated maximum. The real goal is to close the gap between the two numbers as much as possible - my last request got the minimum number from about 1/4 of the maximum to not far under half. The closer the two numbers are, the less margin for error is involved.

Anyway, more than enough preamble, let's talk numbers!

The minimum amount of Carreras produced across the first four generations (I can't find serials on the fifth gen Lemania 5100 Carreras) is: 31344.

The derived maximum amount of those four generations is: 77500.

I consider the latter number much closer to the probable reality, but also feel it is important to have the former as a check too.

So, even allowing for the error potential, the total production of vintage Carreras ever is not into six figures. I feel that's quite an interesting perspective for Heuer's big selling model, compared to supposed rarity of some other houses, and drives home the point about Heuer being primarily about the timers and stopwatches.

The data lets me break numbers down by generations too, which is quite interesting.

First gen (2447, 3647 etc): 39500
Second gen (1153, 73453 etc): 29900
Third gen (barrels 110.573 etc): 6000
Fourth gen (quartz 371.253 etc): 2100

So we see first gen Carreras as the big seller in the range at the time, more competition from other models during the Chronomatic era but still healthy sales, relatively limited penetration of the more unusual barrel Carrera and an inkling of Heuer's (and the Swiss watch industry as a whole) problems at the onset of the quartz era. Incidentally, I feel the third and fourth gen numbers have most scope to change, so serials for those are particularly welcome.

I hope the numbers are of interest to the community at large. I wouldn't want to see them be used in sales ads or anything, but it is an interesting insight into the sort of volumes we are dealing with. And it's a process I won't ever close the door on so as you see and acquire new Carreras (or any model, each one helps to better delimit batches of all watches), please feel free to send the serial to Jeff or myself. We have enough data to repeat the exercise for Autavias and Monacos but not really on any of the other models, but it's a lot of analysis and number crunching so don't expect to see anything on those models any time soon!