Which channels produced projects you won?

Commercial construction runs on relationships, repeat clients and reputation. That is exactly why most firms conclude marketing cannot be measured here, and stop trying.

It can be measured — it just will not flatter advertising. A correct report for this segment shows a large Direct / Unknown bucket, because most awards genuinely come from relationships rather than campaigns. What the attributed portion tells you is which of your marketing activities actually reaches new clients, and that is a smaller but far more honest number than any tool willing to redistribute the rest.

Awards are lumpy. One quarter can contain a single project worth more than the previous year, which makes month-to-month comparison actively misleading.

Comparing whole years, or the same quarter across years, is the only comparison that means anything here. Because nothing had to be installed at the time, those comparisons can be run over periods that closed long ago.

Most commercial work is awarded from a shortlist, which means the decisive event happened earlier: getting onto the bid list, the framework, the approved-supplier panel. A firm that is not prequalified never sees the opportunity at all, and no quantity of inquiry generation changes that fact. Marketing in this sector is therefore doing something narrower and more specific than producing leads — it is making the firm a name that survives a shortlisting meeting nobody from the firm attends. That is a different objective from generating inquiries, and in most firms it is funded from the same budget as though it were the same thing.

That work still leaves records behind. A prequalification submission has a client contact on it. A framework application has an owner on the other side. A tender portal keeps a list of the bodies that invited you to bid. Any of those can go into the lead file, tagged with its source, and the awards that followed are ranked against it, which may be the first time a bid team's pipeline has been read as an acquisition channel rather than as workload. The bid team usually keeps these lists already, in a spreadsheet maintained for reasons that have nothing to do with marketing.

The comparison tends to be uncomfortable in a useful way. A portal subscription producing dozens of invitations a year and one award in three is expensive per win once the bid costs are counted. A single framework that quietly delivers work for five years may never have been counted as marketing at all. Both facts become visible the moment awarded value rather than opportunities seen is what orders the table. It is worth running across several years rather than one, because a portal that produced nothing for two years and one award in the third is a different proposition from one that produced nothing at all.

Repeat clients are the backbone of commercial contracting, and they are also why firms conclude the work cannot be measured. The second project came from the first, the first came from a relationship, and the relationship came from somewhere nobody wrote down. Part of that is true and will show as unattributed. Part of it is not: the first project had a traceable origin, and nobody ever thought to connect it to the fourth one. The firm is not wrong that relationships won the work. It is wrong that the relationship had no traceable beginning anybody could have recorded.

Matching on the client contact makes that connection automatically across a long enough range. A firm running three years of awards through one report frequently finds that a meaningful share of what it thinks of as relationship work traces back to a single inquiry that came through a channel still being funded — or, more awkwardly, through one that was cut two budgets ago on the strength of its inquiry cost. Three years is the minimum useful range here, and five is better where the client base is small and the individual projects are large.

Variations and change orders deserve the same treatment. On a large project they can exceed the original award, and they are recorded as contract value in the project system rather than as new sales anywhere. Exporting them as further rows against the same client credits them where the original award was credited, which is the reading that reflects how the firm actually earns rather than how its paperwork is filed. Whether to include them is a decision to make once and keep, because a year with variations counted and a year without them are not comparable in either direction.

The unattributed share is large, and that is the finding rather than a failure of the method. Most awards may trace to nothing in the marketing data, because they came from a relationship, a repeat client, or a recommendation passed between two people who have never filled in a form in their lives. Seeing that printed at its real size is precisely what makes the remaining share worth reading at all. A report claiming to attribute most of a contractor's awards to advertising would be describing a different kind of business entirely.

Inside that remaining share the ordering is usually a surprise. Industry events and association activity, funded on faith and reported on attendance, often outrank digital spend on awarded value. Paid search, defended at every budget meeting on the cost of an inquiry, may turn out to be producing inquiries from clients too small to be worth the bid team's time. Neither result is a reason to act the same afternoon. Both are a reason to look at what a channel produced rather than at what each inquiry cost to acquire.

A third finding is about periods rather than channels. Awards are lumpy enough that the answer changes depending on where the range is drawn, so the discipline is to fix the range first and read the result afterwards. Whole financial years, or the same quarter across several of them, are the only comparisons that survive one large project landing on the other side of a boundary. Fixing the range afterwards, once the shape of the result is already known, is how a report stops being evidence and becomes an argument.