Keyword Frequency, Density, and Distribution in SEO Copywriting
by Rob Laporte
Visibility Magazine – March 2009
(original article reprinted below)

SEO professionals have long debated whether one can ascertain and benefit from an ideal keyword frequency, density, and distribution (FDD) in SEO copywriting. Some assert that FDD was more heavily weighted by SEs in the past than today. The SEO people at DISC have not found a convincing refutation to the argument that

  • Search engine algorithms must assign some weight to the FDD of key phrases and conceptually related phrases;
  • SEO copywriters do write in key phrases, with or without an FDD ideal in mind;
  • strategically repeated key phrases in SEO writing do not take a lot more time than producing good copywriting anyway;
  • therefore, having at least a loose FDD guideline, preferably one supported by FDD reverse engineering, is better than having none at all.

Even if FDD accounts for only 25% of a page’s search engine position, that 25% is relatively easily controlled, thus making FDD deliver high ROI. And what if FDD accounts for more like 50%? The verbal components of a site’s CMS-SEO can be written correctly once and for perpetuity. If a given page’s copywriting does not have to be changed frequently, the ROI is even better.

Reverse Engineering FDD

In theory, one can reverse engineer FDD by compiling data from equivalent, top-ranked sites, and then using spreadsheets (perhaps helped by software like GRKda) to derive normative values. Given enough data, one can arrive at a useful, if not precise, FDD for all verbal parts of a web page. (Explaining how to factor out PageRank, number of incoming links, age of the site, internal links, key phrase distribution in link anchor text, and several other confounding variables exceeds the scope of this essay). DISC has done this reverse engineering yearly for about 10 years. In the early years of SEO, there were less confounding variables because SEs weighed fewer off-page factors than they do today. You can imagine that the typical repetition of a phrase when humans write about a topic would not change much because human brains and the deep grammar of language have not changed much, if at all. DISC’s research confirms this logical supposition.

Other SEOs have reverse engineered FDD, and come up with widely varying results. DISC has found published estimates of ideal density ranging from 2% to 10% (not counting wild outliers), with an average in the 4% to 6% range.

This percentage applies to one key phrase, so a related question is how many phrases to target on a page. After all, if you’re targeting six phrases, each of which will occupy about 6% of the text, then more than a third of the text would have to be key phrases. Although one can and should include shorter, more competitive phrases within one long-tail phrase, this simple math suggests a practical limit to the number of key phrases to target per page.

Even if reverse engineering could arrive at a precise and unchanging FDD – and it can’t – SEO copywriters should still diversify around that FDD when writing several pages in a web site, which means that an imprecise formula still serves the practical goal of guiding SEO copywriting.

Concept Clustering and Natural Language Algorithms 

Search engines strive continuously to make their ranking algorithms mimic humans’ understanding of language. This effort has produced such terminology as latent semantic indexing, term vectoring, concept clustering, concept space, data clustering, term-document matrix, and the more general natural language processing. The upshot is that increasingly the SEs rank a page not merely by the occurrence of the searched term, but also by the amount and kind of conceptually related phrases. In theory, a search for “car” could produce a top-ranked page that contains no mention of “car” but plenty of “auto,” “automobile,” and even “Chevy.” Would occurrences of “engine” or “gas mileage” also help? While there are ways to determine what phrases the SEs link in such clusters, and thus what other phrases one should write into a page, the time/effort/cost in doing this research for each page is high, and in the end one would do what one should do regardless of this research: write rich, relevant copy that contains a variety of key phrases which augment the core ones and which ring bells of relevance in human readers. After all, natural language algorithms aim to mimic human language processes. In theory, the day will come when just good writing will constitute by far the single best rule of SEO copywriting.

Concept clustering makes reverse engineering so much more difficult that, arguably, the best guides to FDD have to based on research done in the late ‘90s and early 2000s.

Human Art and Search Engine Science

What practical, cost-effective actions can SEO copywriters take away from this web of intricacies surrounding FDD and its ostensible foundation on reverse engineering? As with any art, in performing SEO copywriting according to FDD, discipline and experience are foundational: A scientifically researched FDD is the sketch or the music score. Upon that foundation, the heights of excellence and effectiveness in SEO copywriting are limited only by the artistry of the writer, as he or she strives to please human readers and the increasingly human algorithms. An SEO Einstein would beat the best of us every day.