The website: our home base

Eight pages, and nothing a search engine can build on yet

vallacis.co.uk is live, tidy and legally complete. It isn't yet doing any marketing work.

What's already right

• Company number, registered office and ICO registration are in the footer, and the Terms are unusually clear about what the product does and doesn't do. The Terms give us claims-safe wording we can reuse: no bank feed, we never hold or move money, and Solo keeps MTD-ready records but doesn't file to HMRC.
• It runs on Next.js on Vercel, which is fast and can easily host a content hub, a glossary and calculators.

What needs fixing before anything is added

• It sells a product with a different name. The site describes "CIS Portal" for contractors and accountants. Our ads, store listings and plan all say Solo (and Zebra for now).
• Two addresses for one site. vallacis.co.uk redirects to www, but the sitemap, robots file and social tags all declare the non-www address, and no page states a canonical URL. It's a small fix, and it needs making before we build links.
• Nothing links out to us or from us. There are no social or app links in the footer, so Google has nothing to connect the company to.

Four jobs, one site

Small, easy CIS searches, and a big self-employed layer around them

About 690 UK keywords, pulled from Ahrefs on 15 and 23 September 2026. The pattern is clear: CIS-specific searches are small but almost uncontested, and the higher-volume general self-employed topics surround them.

What competitors' content tells us

• Nobody owns CIS content. The single best CIS page across nine competitor sites is Taxd's CIS calculator, at about 450 UK visits a month. Most CIS pages earn tens.
• The page types that win are calculators, glossary and explainer pages, and templates. RIFT's and Powered Now's "what is CIS" pages, Crunch's invoice template and Hnry's separate glossary hub all show the same pattern.
• Making Tax Digital content earns almost nothing yet (Coconut's whole MTD hub is under 30 visits a page), so there's room for us there in January.

Hubs and spokes, with the glossary as the spine

Each subject gets one hub page that covers it fully, plus a spoke page for each question inside it. Every page links up to its hub, across to related spokes, into the glossary for any term it uses, and out to the product page that solves the problem.

Rules every page follows

• One question per URL. "Why is my CIS deduction 30%?" gets its own page instead of a paragraph in a long guide. That's the shape AI answers quote (section 07).
• Every page ends at the product. The deduction page leads to "see it on every invoice, free", with store badges linked through Singular.
• No orphan pages. Each page has at least three links in (hub, glossary, one related spoke) and links out to its hub.
• Their words in headings, the correct terms in the body. Headings say "payslip", "the 20%" and "stopped"; the text then explains "payment and deduction statement". Both get searched.

The topical map: 105 pages in twelve clusters

Every planned URL, its lead search term (UK, Ahrefs), when it's built and what has to clear first. Search volumes are measured, never estimated; "not pulled" means we haven't measured that term yet.

Searches count each keyword once across every page. That's the reachable demand, not a traffic forecast. A new site earns a share of it over months, not all of it. About 56,000 of the total is general self-employed topics (register, tax year dates, NI, expenses) where gov.uk holds the top spot; the specific questions around them are the realistic target.

Home and product

The pages that sell: what Solo is, what it costs, who is behind it. Every other page links here.

CIS for subcontractors

The core hub. Everything a subbie needs to understand the scheme, one question per page.

Invoices and statements

The closest fit to what Solo does. Weak search results held by tiny sites.

Tools

Calculators and checkers. The page type that wins in this niche and earns links.

VAT and the reverse charge

7,000 searches a month at difficulty 0–2, and Solo already handles it on the invoice.

Getting paid

Payment notices, late payment, retentions. Practical, low competition, sits next to invoicing.

Self-employed or employed?

Employment status, CEST, umbrellas and the new right-to-work rules for subcontractors.

Self-employed tax basics

The big-volume layer (register, UTR, tax year, NI, expenses). Timed for January.

Making Tax Digital

Jan–Mar 2027 surge ahead of the £30k threshold. Gated on what Solo can honestly claim.

Trade hubs

One hub per trade that pulls the other clusters together in that trade's words.

Glossary

84 sourced terms. The spine every other page links into.

Contractors and accountants (later)

Hub-side content. Parked until Hub marketing opens, listed so the structure has room for it.

The full map, with secondary keywords and volumes, is in topical-map.csv. The keyword source files are available on request.

84 terms, each one sourced to gov.uk or legislation

The glossary is the spine of the site: every guide links into it, and every term links back to the guide that goes deeper. It's also the page type AI engines lift definitions from most readily.

How it's built

• One index page, /glossary: A–Z, filterable by group, with every definition on the page and an anchor for each term. 21 terms also get their own URL at launch because people search for them (gross payment status, UTR, reverse charge, pay less notice and others). Other terms graduate to their own page when they earn one.
• Each entry follows the same shape: a 40–70 word definition written for a tradesperson, a "why it matters to you" line, related terms, the gov.uk source link, and a "last checked" date. Figures that change on a date (rates, thresholds, mileage) carry the date they took effect.
• Structured data: DefinedTermSet on the index and DefinedTerm on each entry, so machines read it as a glossary.
• Maintained from one data file (glossary-terms-2026-09.json), so a rate change is one edit, and the "last checked" stamp is honest.
• Reviewed by a qualified person before it goes live. Every definition was drafted from a primary source, but a named human signs it off. That signature appears on the page as the reviewer.

The terms (draft definitions, not yet reviewed)

Two places where official sources disagree are noted in the data file. The gov.uk subcontractor guide still says you wait one year to reapply after losing gross payment status, while Finance Act 2026 sets five years for some cancellations. The contractor and subcontractor guides also differ on fuel as a deductible cost; the glossary follows the CIS 340 guide.

New sites are already being quoted, if the page has the right shape

We put 20 real subbie questions to three AI engines on 23 September (60 answers) and checked Google's AI Overviews for 12 terms. The answer is encouraging: small, new CIS sites are already being cited.

Engines tested: Perplexity (its own search), plus two web-search stand-ins for ChatGPT and Gemini, so treat those two as rough signals. In the 30 "which app should I use" answers, QuickBooks and FreeAgent were each named 18 times; the small CIS apps TapTax, Kletta and DoneCIS 8–9 times each. Nobody named Hnry.

So every guide page is built the same way

1. The answer in the first two sentences, in plain words, before any background.
2. A table wherever there's a comparison: rates, who pays what, deadlines.
3. A worked example with real numbers: an invoice showing labour, materials, VAT and the deduction. The most-cited single page in our test was an accountant's worked labour-and-materials example.
4. Numbered steps for anything procedural (registering, applying for gross status), following gov.uk's order.
5. Sources and a date: links to the gov.uk page, "last checked", and a named reviewer.
6. Their questions as sub-headings, taken from Google's "People also ask" and the forum research ("Do you only pay 20% tax on CIS?").

Right now the company is only a Companies House record

An "entity" is a company that search engines recognise as one known thing, with a name, founders, products and an address, rather than as a string of words. You build one by saying the same facts in the same way everywhere, and pointing everything back to the website.

What exists today (checked 23 September)

Names we can be confused with

• Valla Ltd (valla.uk), an Edinburgh legal-tech company that already has the "Valla" LinkedIn and Crunchbase profiles and shows up in searches for us.
• Valla Homes Ltd, Sam's property company at the same registered office. That's a genuine sibling, which we link through Sam as a founder, not as the same company.
• Zebra: Zebra Technologies (its own apps in the UK App Store) and The Zebra (US insurance). Solo: Dext Solo, Kletta Solo, ApariSolo and Traddie in the same tax and trades space. Hub: an app called "CIS The Hub" is already in the UK App Store.

That's why the legal name Valla CIS Limited is the anchor. Every profile carries the full legal name, company number and website, so the product names can change without losing the entity.

The build, in order

Dollar costs converted at 0.74832 (22 September 2026). Not recommended: a Google Business Profile or Bing Places listing. An online-only business at a registered-office address isn't eligible, and a listing risks suspension. Wikipedia isn't realistic this year.

The structured data, in outline

{
  "@context": "https://schema.org",
  "@graph": [
    { "@type": "Organization", "@id": "https://www.vallacis.co.uk/#org",
      "name": "Valla CIS", "legalName": "Valla CIS Limited",
      "identifier": { "@type": "PropertyValue", "propertyID": "Companies House", "value": "17216592" },
      "url": "https://www.vallacis.co.uk/",
      "address": { "@type": "PostalAddress", "streetAddress": "Neville House, 66 High Street",
                   "addressLocality": "Henley-in-Arden", "postalCode": "B95 5BX", "addressCountry": "GB" },
      "founder": [ {"@id": "https://www.vallacis.co.uk/about#sam-taylor"}, … ],
      "sameAs": [ "https://find-and-update.company-information.service.gov.uk/company/17216592",
                  "https://www.linkedin.com/company/…", "https://apps.apple.com/…", … ] },
    { "@type": "MobileApplication", "@id": "https://www.vallacis.co.uk/solo#app",
      "name": "Solo", "operatingSystem": "iOS, Android", "applicationCategory": "FinanceApplication",
      "publisher": { "@id": "https://www.vallacis.co.uk/#org" },
      "offers": [ { "@type": "Offer", "name": "Free", "price": "0", "priceCurrency": "GBP" }, … ] }
  ]
}

The full example (with founders, ICO number and the website) is in org-schema-example.json. Names update when the brand is settled; the legal name and @id stay the same.

Lists, data, trade press and the people who do the books

A mention is any page elsewhere that names us. The AI engines build their answers from those pages. So we go where they already look, and give journalists something new to write about.

1 · Get onto the lists AI engines already quote

All 25 targets, with who they currently list, are in ai-listicle-targets-2026-09.csv.

2 · News only we can tell, all from HMRC's own figures

3 · Who we pitch

Contact routes for each outlet still need checking; that list was cut short and is the next piece of research. On paid press-release wires: a Muck Rack study reports press releases make up about 1% of AI citations, against about 27% for journalism. We haven't re-checked it, but it matches our own test, which cited no press releases. So a wire release is a launch formality at most. The money goes into the survey and direct pitches.

4 · Accountants

Lena's point that accountants drove Xero's UK growth holds for AI answers too: accountancy blogs are 15% of what gets cited. Accountant partners who recommend Solo to subbie clients and write about it on their own sites earn mentions in the places the engines read. This joins up with the /accountants page later.

Foundations at launch, depth before Christmas, the MTD push in January

Pages per wave: 28 at launch, 31 in Nov–Dec, 18 in Jan–Mar, 7 later, plus the 21 glossary pages alongside the launch set. Use the filter buttons in section 05 to see each wave's pages. This is a publishing order, not a guarantee of rankings. A new domain usually takes months to rank, which is why this starts before launch and why paid carries months 1–3.

Four early signals, then traffic, then installs

The month-3 review (already in the plan) resets the order of waves 3 and 4 using real Search Console data instead of forecasts.

Eight answers that unblock the build

1. Which domain is home? vallacis.co.uk, or the brand's own domain from day one.
2. Who builds the site, and who writes? My role is the plan, not the content. We need a named writer or agency and a developer for the Next.js site.
3. Who reviews for accuracy? A qualified person (an accountant or tax adviser) named on each page as reviewer. This is what makes the content trustworthy, and it's how every figure stays sourced, as Lena asked.
4. Who signs off wording against the Claims Register, and how fast? 19 glossary terms and about 25 planned pages need it.
5. Should the no-competitor-names rule cover comparison pages? It's the page type AI engines quote most for "which app" questions (section 07).
6. Is CIS300 filing on Hub's roadmap, and what's Solo's MTD route? Getting onto HMRC's lists depends on it, and every AI "which app" answer points to those lists.
7. Can the founders go on the record? Photos, short bios, a quote each, and Luke's trade story confirmed.
8. Is @vallacis on Instagram ours? And is there budget for a proper survey sample (a panel firm) alongside the On the Tools community?

What's measured, and what's judgement

Judgement, not measurement: the clusters, page order, waves and which terms get their own glossary page. Not yet checked: contact routes for press outlets, the HMRC listing process for a records-only app, whether CIS income counts before or after deductions for the MTD £30,000 test (needs an adviser), and the ChatGPT and Gemini columns of the AI test (web-search stand-ins, not the products themselves).