Six Names Make the Shortlist. The Forty That Got Cut Were Sorted by the Buyer's Software, Not a Person.

Six Names Make the Shortlist. Forty Got Cut Before Anyone Read a Deck

Every agency review you have ever lost had two rounds you never saw. Round one is the long list — thirty to fifty shops a marketing director or a procurement analyst assembles from directories, partner listings, a peer's recommendation and whatever the search turns up. Round two cuts that to six who get the RFI. Only after those two rounds does anyone look at your work.

Through 2025, round one was a person with browser tabs open, and the shops that got on the list were the ones with a memorable name, a recent award, or a friend inside. In 2026 round one is increasingly done by software working on the buyer's behalf: an assistant that reads forty agency sites in an afternoon, pulls what it can from Clutch, Agency Spotter, the Google Partners directory, the HubSpot Solutions Directory and Meta's partner listings, and hands its human a comparison table with columns the human asked for — verticals served, size, platform certifications, conflicts, minimum engagement, location.

If your agency cannot be read by that process, you are not losing pitches. You are not being invited to them, and nobody at the client will ever be able to tell you why.

What the Buyer's Assistant Is Actually Looking For

It is not looking for beauty. It is looking for facts it can put in a column, because its human asked for a table. Do you do paid social for regional healthcare systems? How many people, and where? Are you HubSpot-certified, and at what tier? Do you already work with a competing health system in the same DMA — the conflict question that kills more introductions than bad work ever has? What does an engagement start at? Who would actually be on the account, and what does the account director do besides attend the pitch?

Now go look at your own site with that list in hand. Most agency sites answer none of it. The capabilities page says "integrated brand experiences." The case studies are a scroll of full-bleed images with the results baked into the picture as text a machine cannot read. The good numbers are in a PDF behind a gated form. The team page is initials and hover states. There is no statement of what you do not do, no vertical list, no minimum, and the only way to reach you is a form that a human answers within two business days — by which point the shortlist is done.

Being findable by another company's software means publishing, in plain readable text on your own site, the facts a buyer would have asked you on a first call: what you do, for whom, in which categories, at what size of engagement, with which certifications, and how to book time with a human this week. That is the whole discipline. It is not a rebrand.

flowchart TD
  A["VP of Marketing sends her assistant to find shops"] --> B{"Does the site say which verticals you serve?"}
  B -->|No| Z["Cut before anyone opens your work"]
  B -->|Yes| C{"Certifications and headcount stated in text?"}
  C -->|No| Z
  C -->|Yes| D{"Conflict with a competing account?"}
  D -->|Yes| Z
  D -->|No| E["Added to the six-name shortlist"]
  E --> F["Chemistry call booked from your own calendar link"]

The Seven Facts Most Agency Sites Refuse to State

Write these as text, on pages, without a form in front of them. Verticals you actually serve, named by industry, with the ones you decline stated too. Headcount and offices. Platform certifications with the year — Google Partner or Premier Partner status, Meta Business Partner, HubSpot tier, Klaviyo, Shopify, Microsoft Advertising. Service lines in the words buyers use, so "paid search management" beats "performance ecosystems." A minimum engagement, even as a range. A named leadership team with real titles. And a conflicts statement: the categories where you already hold an exclusive, which reads as confidence, not weakness.

Then fix the case studies. A case study a machine can read is text: client type, category, the problem in a sentence, what you did, and a number with a time period on it. If the client will not let you name them, "a 14-location orthopedic group in the Southeast" is legible and honest; a logo wall is neither. If your results live only in a sizzle reel or a Behance carousel, they do not exist to the process that builds long lists.

The third fix is the one agencies resist hardest: make booking possible without a human. A calendar link for a 20-minute introduction, and a phone line and web chat that answer at 8:40 p.m. when a marketing director is finally clearing her list. The assistant doing research will happily take a booking if one is available; if it is not, it moves on and your name never enters the conversation.

Your Directory Profiles Are Doing More Work Than Your Homepage

Uncomfortable but true. A profile on Clutch or Agency Spotter or a platform partner directory is already structured — categories, service breakdown percentages, minimum project size, locations, reviews — which is exactly the form a research assistant consumes best. Half the agencies reading this have a profile last touched in 2023, listing services they stopped selling and a founder's email that bounces.

Spend an afternoon on this. Update the service percentages so they reflect what you actually sell now. Fill in the minimum project size honestly — filtering out the $3,000-a-month enquiries is a feature. Ask three clients for reviews, because review text is where a research assistant finds the specific words that match a buyer's question. Claim and complete the platform partner listings; the certification badge on your footer is worth far less than the directory entry that says which platforms you are certified in and when it was verified.

What One More Long List a Quarter Is Worth

Illustrative, at a mid-size independent. Use your own averages.

• You are invited to about 18 pitches a year and win one in four.
• Average win: $14,000 a month, retained for about 16 months, so roughly $224,000 in fee.
• Suppose being legible to the research step adds four long-list appearances a year, and half of those turn into an actual pitch invitation.
• Two extra pitches a year at the same one-in-four win rate is half a win.

Half a win a year is a soft-sounding number until you notice the cost line. And unlike a rebrand, you can check it: watch whether your inbound enquiries start arriving already knowing your verticals, your minimum and the name of the person they want to speak to. That is the signature of a buyer who arrived pre-briefed.

What This Cannot Do, and What You Should Not Publish

Being legible gets you onto the list. It does not win the pitch, and no amount of readable text substitutes for the chemistry meeting where a CMO decides whether she wants to sit with your account director every Tuesday for two years. Strategy, taste and the relationship remain entirely human, and the shops that treat this as a marketing project rather than a plumbing project will be disappointed.

Be careful what you make readable. Client results under an NDA stay unnamed; get written permission before publishing a number tied to an identifiable client. Do not soften a claim you cannot substantiate — your own advertising is advertising, and the same standards you apply to a client's claim apply to "the fastest-growing agency in the Midwest." An unsupported claim is worse now than it was in 2024, because software will repeat it back to a prospect in a comparison table as though you had certified it.

And keep a person on the first reply. An assistant that books a meeting is fine; an assistant that negotiates your scope or quotes a rate is not. Publish the range, let a human hold the number.

Frequently asked questions

Isn't this just search engine optimisation with a new name?

It overlaps, but the target is different. Search work is about ranking for a phrase. This is about whether a research assistant reading your page can answer six specific questions about your firm without a form, a PDF or a phone call. Plenty of agencies rank well and are still unreadable.

Do we really have to publish a minimum engagement?

You do not have to, but you pay for the silence in two ways: unqualified enquiries you spend hours on, and comparison tables where your row says "unknown" next to competitors who answered. A range — "most retainers start at $8,000 a month" — costs you nothing and does the filtering for you.

How would I even know an assistant looked at us?

Your site logs already show it: visits that read three or four pages in seconds with no images loaded, arriving from no search term. Ask whoever runs your site to report those separately each month. The trend line matters more than the count.

What about the search consultants who run big reviews?

Consultants still run the largest reviews, and their databases still matter — keep your record with them current. But the mid-market accounts, the $8,000-to-$25,000-a-month retainers most independents live on, do not hire a consultant. Those clients do the search themselves, and increasingly they do not do it by hand.

The Afternoon That Starts This

Sit down with your new-business lead and write the seven facts on one page, in plain sentences. Publish that page. Then rewrite three case studies as text with numbers and a time period, refresh two directory profiles, and put a real booking link where the contact form is. That is one afternoon and one developer morning, and it can be live this week.

The last step is the one agencies underinvest in: what happens when the buyer, or their assistant, actually reaches out at 9 p.m. on a Tuesday. CallSphere builds AI voice and chat agents that answer the agency line and website chat around the clock, ask what category and budget range the enquiry is about, book the introductory call on the right principal's calendar, and pass the transcript to your new-business lead before morning. It will not get you on the long list — your published facts do that. It makes sure being on the list turns into a meeting.