Twenty-five competitors, one document, one morning: how counting words gives you both an audit and a statistic your trade press does not have yet

1. What is the Commodity Phrase Index, in one paragraph?

Take the twenty-five closest competitors in one vertical. Paste every homepage and every services page into a single document. Count phrase frequency. Then delete from your own site every phrase that appears more than twice. That is the entire method. What it produces is a ranked inventory of the ground no one in your category can own, with counts attached. That beats a positioning statement, because counts are harder to argue with than taste.

A committee will debate whether “trusted partner” feels right for six weeks. Show them that nineteen of their twenty-five competitors say it word for word and the debate ends before lunch.

2. What are the exact steps?

1. Pick twenty-five real competitors, meaning the ones that show up in your losses. If you can only name eleven, that is your first finding.

2. Pull the homepage and the top services page from each. Roughly fifty pages, usually 40,000 to 60,000 words of text.

3. Count two-word through five-word phrases. Do the stop-word cleanup after the count, never before, or you will destroy the phrases you are hunting.

4. Rank by document frequency, meaning how many of the twenty-five use it. Raw counts mislead. A phrase used thirty times by one company is a nervous tic, while a phrase used twice by twenty-one companies is a category.

5. Set the kill line at two. Anything above it comes off your own site this week.

6. Save the ranked table. You just built the audit and the publishable asset in the same sitting.

The counting is six lines of Python, a pivot table, or one long paste into a model with instructions to return frequencies as JSON. The scarce ingredient is the willingness to look at the answer.

3. Why does this matter more in 2026 than it did in 2016?

Because the reader changed. SparkToro found that 68 percent of United States Google searches ended without a click in early 2026, and that the share of searches producing at least one click fell 9.51 percentage points between 2024 and 2026, a 22.9 percent relative decline. Ahrefs data cited in that study puts AI Overviews on more than 20 percent of searches, with click-through falling roughly 60 percent when one appears. Pew Research Center watched 68,879 real searches: people clicked an organic result 8 percent of the time with an AI Overview present, against 15 percent without.

Now add the buyer. A Gartner survey of more than 1,100 B2B customers found that 64 percent could not tell the difference between one B2B brand’s digital experience and another’s. That was measured in December 2020, years before a generative writing tool sat inside every marketing department in America. The commodity phrase problem has been compounding at machine speed since.

If 64 percent of buyers cannot tell you apart, the words doing the telling are the ones you share with everybody.

4. What does Walmart have to do with any of this?

Walmart is the cleanest available demonstration that scale does not save you from commodity language, and that specific numbers do.

Look at what Walmart puts on its leasing site when it wants a business to sign: 90 percent of all Americans live within 10 miles of a Walmart store; 38.5 million people shop at Walmart every day; more than 4,500 stores carry leased space; more than 350 brands and services operate inside them. Four claims, four numbers, zero adjectives. Every one of those sentences is unavailable to Target, to Kroger, to Dollar General, and to the appliance dealer on Route 51.

The financial page reads the same way. For fiscal year 2026, ended January 31, 2026, Walmart reported total revenues of $713.16 billion, consolidated net income of $22.27 billion, 10,955 retail units, and global e-commerce of $150.4 billion, up 24 percent. Roughly 280 million customers and members visit each week across 19 countries. In February 2026 it became the first traditional retailer valued above $1 trillion.

Here is the part that should make you feel better about your own website. Walmart is not immune. Its press boilerplate calls the company a “people-led, tech-powered omnichannel retailer,” a phrase that would clear a Commodity Phrase Index only because so few firms are large enough to say it with a straight face. The lesson is sequencing. Walmart leads with the ten-mile number and lets the adjectives ride in the back seat. Most of your competitors have the seating chart reversed, and so, probably, do you.

5. What will the results actually look like?

Predictable, and that is the point. Across staffing, insurance brokerage, HVAC, managed IT and industrial distribution, the top of the list is almost always some arrangement of these:

1. “Trusted partner” and “your partner in,” usually the single most duplicated construction in professional services.

2. “Industry-leading,” “world-class,” “best-in-class,” which are three ways to say the same unverifiable thing.

3. “Tailored solutions,” “customized approach,” “we take the time to understand your business.”

4. “End-to-end,” “full-service,” “one-stop shop,” which tell a buyer you have no edges.

5. “Passionate,” “dedicated,” “committed to excellence,” which describe an emotional state rather than a service level.

Run the count and you get the local version with real numbers attached. Saying “everyone uses this language” is an opinion. Saying “21 of 25 firms in this market use the exact phrase trusted partner, and 17 use it above the fold” is research.

6. If machines are reading, why does deleting phrases help?

Because retrieval rewards distinctiveness and density, and both are measurable. The analysis published at 1000startups.com in “You Lost the RFP Before Anybody Wrote It” collected the relevant figures in one place, and they are worth memorizing:

1. Pages carrying 19 or more discrete statistics averaged 5.4 citations against 2.8 for thin pages, close to double.

2. Sections running 120 to 180 words between headings averaged 4.6 citations against 2.7 for other lengths.

3. Domains absent from G2, Capterra and Trustpilot averaged 1.8 citations; listed domains averaged 4.6 to 6.3.

4. Median citation half-life across 3.5 million citation events is 4.5 weeks, and ChatGPT churns fastest at 3.4 weeks, so publishing in bursts means going dark in the gaps.

5. Per Ahrefs, across 26,283 source URLs, “best X” listicles account for 43.8 percent of the page types ChatGPT cites, and 28.3 percent of its most-cited pages have zero Google organic visibility.

Read those together and the strategy writes itself. A phrase shared with twenty-four competitors gives a retrieval system no reason to pick you out of the pile. A number that exists on exactly one page in the world gives it a reason. The same piece reports, per 6sense, that 94 percent of buyers used an LLM somewhere in the purchase and 95 percent bought from a vendor already on the day-one shortlist, up from 85 percent a year earlier. You are optimizing for a list written before the phone rings.

7. How does an audit turn into a publishable asset?

Name it and ship it. A “Trusted Partner Index” for staffing, or for insurance brokerages, or for regional HVAC, is a number that does not currently exist anywhere. It costs an afternoon of scraping. It is exactly the sort of specific, citable statistic that trade press and answer engines pick up, because writers need a figure to cite and there is no other source for this one.

The link economics support it. Ahrefs reports that 95 percent of web pages have zero backlinks. Backlinko and BuzzSumo, across 912 million blog posts, found content over 3,000 words earned 77.2 percent more referring domains than content under 1,000 words, with original research beating opinion pieces because a journalist needs a figure to cite. Editorial.Link has digital PR named the most effective tactic by 48.6 percent of SEOs, guest posting second at 16 percent.

Walmart, again, is the model. Walmart Data Ventures packaged the company’s first-party data as Scintilla and, at its Inspire 2025 event, published the finding that Scintilla subscribers grew total omni sales 15 percent versus non-subscribers, backed by more than 30 supplier case studies across the United States, Mexico and Canada. Kimberly-Clark and Reynolds are named users. That is a company turning its own exhaust into a citable statistic and a product line at once. Your phrase count is the same trick at one ten-thousandth of the budget.

One morning of counting produces two deliverables: a deletion list for your site and a statistic the internet does not have.

8. What should you refuse to delete?

1. The category noun. “Commercial HVAC contractor” appears on all twenty-five sites because it is what you sell, and retrieval needs it. Kill the adjectives, keep the nouns.

2. A true claim that happens to be common. Replace it with the arithmetic underneath it rather than throwing it out.

3. Geography, licensing, certifications, anything a buyer filters on. Those are qualifying facts.

4. Anything you can prove and source. Proof is never commodity, even when the claim it supports is.

9. What replaces the phrases you delete?

Swap adjectives for arithmetic. Four examples that hold up in any vertical:

1. “We are a trusted partner” becomes “41 of our 60 clients have been with us more than five years.”

2. “Industry-leading response times” becomes “Median time to first qualified resume: 26 hours.”

3. “Tailored solutions” becomes “We staff third-shift maintenance only. For office roles we will refer you to three firms we trust.”

4. “World-class service” becomes the name of the person who answers the phone, with her direct line.

Walmart prints the old price beside the new one and lets the subtraction do the persuading. A refusal, a median, a named human, and a number are all things a competitor would have to lie to copy. That is the working definition of ownable ground.

10. What do you measure afterward?

1. Kill count: how many distinct commodity phrases came off the site.

2. Uniqueness ratio: the share of homepage word count that survives the index. Under 40 percent is a bad morning.

3. Criterion presence: run twenty buyer prompts monthly and log whether your differentiator shows up as a criterion.

4. Citation counts and refresh cadence, given that 4.5-week half-life.

5. Pickup of the published index: press mentions, links, and requests for next year’s edition.

11. What does it cost, and what is the alternative?

One morning to count, one week to rewrite and publish. Compare that with the status quo. Loopio’s research across more than 1,500 teams, cited in the same 1000startups.com analysis, has the average organization submitting 166 RFPs a year at 33 hours each, roughly 5,478 hours of labor, against a 39 percent win rate, with only 13 percent of losses blamed on proposal quality and 55 to 61 percent on price and competition. Teams lose on the criteria, then spend five thousand hours arguing inside them.

A morning of counting will not fix a bad business. It will tell you, with numbers, which sentences on your site are doing no work at all, and it will hand you a statistic worth publishing on the way out. Both come from the same document. Very few exercises in marketing pay twice.

Sources cited in this article

Named in the body above and listed here with links.

1. Claude Penland, “You Lost the RFP Before Anybody Wrote It,” 1000 Startups (retrieval multipliers, 6sense and Loopio data): 1000startups.com/2026/09/04/you-lost-the-rfp-before-anybody-wrote-it/

2. 1000 Startups, AI Search & Visibility archive: 1000startups.com/category/ai-search-visibility/

3. Search Engine Land on the SparkToro 2026 zero-click study: searchengineland.com/google-zero-click-searches-2026-study-479717

4. Gartner, on seller-assisted digital buying (64 percent cannot tell B2B brands apart): gartner.com newsroom release

5. Walmart leasing statistics (90 percent within 10 miles; 38.5 million daily shoppers): leasing.walmart.com

6. Walmart Inc. FY2026 Annual Report, filed with the SEC: sec.gov EDGAR filing

7. Walmart Data Ventures at Inspire 2025 (Scintilla, 15 percent omni sales lift): corporate.walmart.com news release

8. Pew Research Center (July 2025); Backlinko, BuzzSumo, Ahrefs, Editorial.Link data compiled at thestacc.com/blog/backlink-statistics/

9. Retail Dive on Walmart and Sparky in ChatGPT: retaildive.com