Should a manufacturer add machine capacity?

Capital equipment decisions in small manufacturing shops are made against two things: the current backlog, and the owner's read of the market. The backlog is real and it is the wrong instrument, for a reason that is easy to state and hard to feel. It is a record of enquiries that arrived. It contains no information about the enquiries that went to a competitor because your quoted lead time was long, or that were never sent because a buyer filtered you out before making contact. A backlog measured while quoting long lead times is a measurement taken through the constraint being evaluated.

There is a denominator available and it is unusually good for this specific decision, because it is published by employment size band rather than as a single count. Census County Business Patterns reports establishments in an industry and county broken into size bands, which means you can see not only how many shops operate in your market but roughly how many are of your scale. That distinction is what makes the series usable here: a county with many very small shops and two large ones is a different competitive situation from a county with a dozen mid-sized ones, and a total count reports them identically.

Two properties of that series must be understood before quoting it, and both are easy to miss because the file does not shout about them. First, some size cells are withheld to protect the disclosure of individual establishments, so a band that reads as empty may be suppressed rather than genuinely zero — and in a concentrated industry in a small county, the suppressed band is frequently the one you care most about. Second, a number of establishments are published as unclassified by size, so the bands do not necessarily sum to the total. A shop that adds the bands and finds them short of the headline has not made an arithmetic error; it has found the unclassified residual.

Alongside it, covered employment for the same industry and county gives the scale of employment in your market, subject to its own definitional quirk: it counts jobs covered by unemployment insurance rather than people, so a working proprietor drawing no covered wage does not appear. In a market composed largely of owner-operated shops, that omission is not marginal.

None of this produces a demand forecast, and it should not be presented as one. What it produces is a plausibility check on the story behind the investment. If the case for a new machine rests on capturing work from shops of a certain size in a certain area, the count of such shops is a fact you can obtain in an afternoon, and it will either support the story or quietly demolish it. Most capital cases in this sector are never subjected to that test.

The other test worth running before financing anything is on the quoting path itself. If quotes are going out slowly because there is one estimator, then the constraint that lost the work was the quote, not the machine. Adding capacity behind a bottleneck that sits upstream of it produces a faster machine standing idle while the same estimator quotes at the same rate, which is an expensive way to discover where the queue actually was.