Most of this work is data engineering.
Using AI is the easy part. The hard part is designing a page somebody wants to read and getting your data into a shape a model can actually use.
How to tell which aspects of SEO you truly need.
Don’t get sold an “SEO” package. Find someone who can help you diagnose the real problem and fix it.
A site with 40,000 URLs, where Google spends its visits on filter pages nobody searches for. That site doesn’t need more content.
Technical SEO →A site with twelve good pages and no coverage of how its market searches. That site needs a new foundation.
SEO Strategy →Location pages that would sell the same thing anywhere else once you delete the city name.
Programmatic SEO →Nothing gets updated after it ships, so the traffic arrives as a spike.
The services we offer.
Roughly ordered by where programs usually start.
SEO Strategy
What to build and in what order, from how your market searches and what data you actually hold. You get a number of pages to build and the order to build them in.
Read → 02Technical SEO
Google only visits so many of your pages. The work is making sure those visits land on the ones that make you money.
Read →Programmatic SEO
Thousands of pages built off a real data layer, so each one has something true to say. Delete the city name and read what’s left: if it would sell the same thing anywhere else, Google won’t rank it.
Read → 04Content Production
Genuinely good pages made at a volume that would once have needed a room full of writers. AI-first rather than AI-exclusive, because a New York Times op-ed is still better done the old way.
Read →Generative Engine Optimization
Being the source an AI answer cites, which so far looks like the same fundamentals as search. Be wary of anyone claiming to have this figured out, because the models are black boxes.
Read → 06Contractor SEO
Home services and the trades, where a homeowner searches the problem when something breaks and doesn’t yet know who to call. The hard part is never page volume, it’s getting real evidence out of a few hundred owners.
Read →Link Building
What earns links, in our experience, is original data rather than outreach volume, so that is what we build the campaign around. When somebody reports a link scheme to Google, your domain is the one at risk, not the vendor’s.
Read → 08Conversion Rate Optimization
Turning organic landing traffic into contacts, including the awkward part: small improvements need a lot of traffic to prove. Most page groups can’t carry a valid test, so the first job is finding which of yours can.
Read →Two engagements, in numbers.
Sherpa Auto Transport ships cars and Franzy sells franchises, and both had data nobody had turned into pages.
A distant challenger, neck-and-neck with incumbents 10× its size
Sherpa Auto Transport had route and vehicle data nobody had turned into a page, so we built coverage across the lanes they actually run, with real pricing.
Source: SEMrush estimates, U.S. database, pulled June 18, 2026.
Read the case study → Case study · Franchise marketplaceFranzy: thousands of thin pages, unstuck
Thousands of franchise profile pages sat thin and largely unindexed, and the problem was the data behind them rather than the design, so we rebuilt every block.
Source: Franzy engagement reporting, A2 Analytics, 2026.
Read the case study →Three questions worth asking before you sign anything.
Including us. If a provider can’t answer these three plainly, we’d keep looking.
What data does this run on, and where does it come from?
If the answer is keyword research, the pages get built from the same public inputs everybody else has. The version that works runs on something you own: jobs, routes, inventory, pricing, reviews.
What happens to these pages in month nine?
A page is never more accurate than the day it went live, so a program with no refresh built in is buying a spike. Ask who’s watching for decay and what triggers a rewrite.
What would you tell me not to buy?
Everybody’s constraint is one thing at a time, so a provider that recommends its whole catalogue hasn’t diagnosed anything.
Our own answers
What it runs on: your data layer, usually messier than anybody expected. Month nine: a scheduled refresh, because we’ve watched pages decay. What not to buy: often the content, because the constraint is technical and it can wait a quarter.
Tell us what’s in the way and we’ll tell you where we’d start.
Send us the site and what you’ve already tried, and we’ll tell you which of these we’d start with and why. If you don’t need us yet, we’ll say so.