Kind: Measured report
By Asif Hussain, Founder and CEO, SpecIndex
A rep checking a fire alarm section in a tower manual finds the whole contest in one line: "Acceptable manufacturer for this project shall be Edwards, Notifier, Siemens, or approved equal." Three names are already in, and every other maker has to argue its way in.
That line is from the project "Anchorage Airport Traffic Control Tower (ANC ATCT) and Anchorage Terminal Radar Approach Control/Base Building (A11 TRAC" (document "Attachment J-3c ANC ATCT TRACON Final Spec Vol 4.pdf", pages 400 to 414). It carries the message of this article in plain words. Substitution fights are decided at the approved list, and the list is short enough to work name by name.
Facts at a glance
- The headline figures come from the State of the Spec enumeration published on this desk on October 14, 2026. (From: Method note.)
- Its population was 4,838 citable positions on 1,421 projects, made up of 622 basis of design positions and 4,216 approved equal positions. (From: Method note.)
- That count gives 6.8 approved equal positions per basis of design position, a median approved list of 2 manufacturers, and 41.2 percent of basis of design sections naming one manufacturer or fewer. (From: Method note.)
- The fresh bounded check was run on October 16, 2026 with the SpecIndex tool find_brand_positions. (From: Method note.)
- Across those 600 rows, the approved manufacturers count was 0 on 2 rows, 1 on 145 rows, 2 on 52 rows, 3 on 29 rows, and 4 or more on 372 rows. (From: Method note.)
The count says the list is the contest
We published the count behind this in the State of the Spec enumeration on October 14, 2026. In that population the median approved list names 2 manufacturers. About two in five basis of design sections name one manufacturer or fewer.
For each basis of design position there were nearly seven approved equal positions. Read that the right way round.
The rare prize is the design basis. The common position is a place on a list beside a small number of named rivals, and that place is where most selling either holds or slips.
A long list would spread that work thin and make it generic. A list of two names does the opposite. It tells a rep exactly whose product the section favors, exactly who else is already accepted, and exactly what proof an outsider would need to get in.
A second document shows the same gate in different words
The Anchorage line names the gate as acceptable manufacturers. A second document names it as equipment that will be considered.
On the project "NONPERSONAL SERVICES TO PROVIDE FULL ELECTRICAL FUNCTIONALITY TO THE H2 CONEX AT ANNETTE ISLAND WEATHER SERVICE OFFICE (" (document "scope (b) HYDROGEN container drawings NWWV41002200019 .pdf", pages 10 to 12), the substitution language reads in full: "EQUIPMENT FROM CUTLER-HAMMER, SIEMENS ENERGY & AUTOMATION, GENERAL ELECTRIC, OR ALTERNATIVE SYSTEMS WILL BE CONSIDERED PROVIDING THAT SUFFICIENT DOCUMENTATION IS PROVIDED TO SATISFY THE CONTRACTING AGENCY THAT THE EQUIPMENT MEETS THE REQUIREMENTS OF THE SPECIFICATIONS, AND MATCHES THE BASIS OF DESIGN ON ALL POINTS WHICH ARE PERTINENT TO THE PROJECT."
The specified product on that row is "Square D by Schneider Electric". The structure is the same as Anchorage, even though the words differ. One product sets the standard.
A short set of named makers will be considered. Anyone else carries the burden of documentation, and the test is a match on all points that matter to the project.
That burden is the point a generic name count misses. Appearing in a section is not one thing. Being named on the list, being the product the list is measured against, and being outside the list with a folder of proof are three very different places to stand.
A fresh bounded check shows why the field needs care
On October 16, 2026 we read the first hundred rows returned for each of six brands, six hundred rows in all, and recorded the approved manufacturers count on each row. About a third of those rows named one or two manufacturers. The full split by brand is in the method note below.
That check does not replace the enumeration, and the method note says why in plain terms. The rows returned first are not a random draw from the corpus, and some rows from large manuals carry counts in the dozens.
Treat the bounded check as a look at the rows the tool returned, in the order it returned them. Treat the enumeration, built on its stated population of citable positions, as the source for the median and the shares quoted above.
The contrast is still useful for a working rep. It shows that a single field in a single row can describe a whole manual rather than one tight section list, so the work is to open the section, read the list as written, and count the names that are actually there.
My judgment is that teams start this fight too late
Substitution fights are decided at the approved list, and the list is short enough to work name by name. That is the judgment this measurement supports.
By the time a substitution request is being argued, the list has already done its work. The specifier chose the names, the contractor priced around them, and the outsider is asking busy people to reopen a settled choice. The cheaper fight, in time and in credibility, is the earlier one to be written onto the list while the section is still being drafted, or to be ready with point by point proof against the named product before the bid goes out.
A short list makes that earlier fight practical. A rep does not need a campaign aimed at everyone. She needs to know, section by section, who the two or three names are, which firm wrote them in, and what her product has to prove to join them or displace one of them.
What to do with a short list
Start with the sections where you are absent. Read the approved list in each one and write down every name on it. That list, not a territory wide rival set, is your working list for that project.
Next, sort your effort by list length. A section naming one or two makers is a direct contest you can prepare for name by name, with the named product data sheet open beside yours. A section with a long list is a different job, and it should not absorb the time the tight sections deserve.
Last, take the list to the specifier before the next project, not to the contractor after this one closes. Ask what a maker has to show to be named.
Bring the proof that answers that question for the exact product class the section covers. That conversation is where the next list gets written.
Method note
The headline figures come from the State of the Spec enumeration published on this desk on October 14, 2026. Its population was 4,838 citable positions on 1,421 projects, made up of 622 basis of design positions and 4,216 approved equal positions. That count gives 6.8 approved equal positions per basis of design position, a median approved list of 2 manufacturers, and 41.2 percent of basis of design sections naming one manufacturer or fewer.
The fresh bounded check was run on October 16, 2026 with the SpecIndex tool find_brand_positions. We paged the first 100 rows returned for each of Trane, Eaton, Armstrong, Siemens, Zurn and Victaulic, in calls with a limit of 10 or below, for 600 rows in all, 100 rows per brand. Across those 600 rows, the approved manufacturers count was 0 on 2 rows, 1 on 145 rows, 2 on 52 rows, 3 on 29 rows, and 4 or more on 372 rows.
The split by brand was as follows. Trane: 0 on 1 row, 1 on 33 rows, 2 on 11 rows, 3 on 10 rows, 4 or more on 45 rows. Eaton: 0 on 1 row, 1 on 33 rows, 2 on 10 rows, 3 on 9 rows, 4 or more on 47 rows.
Armstrong: 1 on 21 rows, 2 on 18 rows, 3 on 4 rows, 4 or more on 57 rows, with no rows at 0. Siemens: 1 on 42 rows, 2 on 7 rows, 3 on 4 rows, 4 or more on 47 rows, with no rows at 0.
Zurn: 1 on 13 rows, 2 on 5 rows, 3 on 2 rows, 4 or more on 80 rows, with no rows at 0. Victaulic: 1 on 3 rows, 2 on 1 row, 3 on no rows, 4 or more on 96 rows, with no rows at 0.
The limits of the bounded check are stated here plainly. The first rows returned are not a random sample and were not weighted to match the corpus, so their split is not a population estimate and does not replace the enumeration median. Some returned rows carry counts that reach into the dozens, which reflects how the count is recorded on large manuals.
Both exhibits above are quoted from the substitution language field of a returned row for Siemens in that check, with the project name, the document title and the page span as returned. All counts in this article describe public bid documents SpecIndex has read. Work that is bid in private does not produce a public specification and is not represented here.