San Francisco · Web development
Web development agency in San Francisco
Your engineers should be building the product, not the marketing site. We build the site so your team does not have to, and hand it over in a state your marketers can run without filing tickets.
Get a free site reviewWhat does a web development agency in San Francisco do?
A web development agency in San Francisco builds and maintains the site your company sells from, as distinct from the product itself. The work is making pages fast, making them readable by search engines and AI crawlers, and making them editable by marketers without consuming engineering time.
Two things people mean by "rebuild the site"
Both are legitimate projects. They cost different amounts and return different things, and the expensive mistake is buying the first while expecting the second.
A redesign
Changes how the site looks
- New visual language, new brand expression, updated components
- Success measured by whether the team is happy with it
- Page structure and the conversion path usually carried over unchanged
- Traffic typically holds; enquiries typically do not move
- Genuinely worth doing when the brand has outgrown the design
- Does not, on its own, fix speed, accessibility or crawlability
A rebuild
Changes what the site earns
- Templates reworked for speed, accessibility and machine readability
- Success measured against enquiries, not against internal preference
- The path from arrival to contact deliberately shortened and tested
- Platform chosen so marketing can publish without engineering
- Worth doing when the current site limits what you can do next
- Usually includes a redesign, because you are in there anyway
We will tell you which one you actually need during the review, including the cases where the answer is neither and three weeks of targeted repair would do it.
What we build and why it matters here
Four requirements. The second is the one Bay Area companies get wrong most often, and it is the one with the largest downstream cost.
Speed measured on real visits
Core Web Vitals are assessed on the connections and devices people actually use, not on a laptop on office wifi. Most slow marketing sites are slow because things accumulated and nothing was ever removed — the fix is more often subtraction than architecture.
Usually a subtraction problemRendered so machines can read it
Client-rendered marketing pages are the single most common technical fault we find in this market. Google may eventually index them; AI crawlers execute far less JavaScript and frequently will not. Server-rendering the commercial pages is not a preference, it is the price of being found.
The expensive mistake hereA conversion path that was designed
The route from landing to enquiry, made deliberately shorter — fewer fields, clearer next step, no interstitial asking people to book a call before they know what you do. This is the cheapest lever on the whole site and it is usually inherited rather than chosen.
Cheapest lever you ownIndependence from your engineers
A CMS your marketing team can publish and build pages in without opening a ticket. If shipping a landing page needs sprint capacity, you will stop shipping landing pages — and that constraint quietly kills more marketing programmes than budget does.
Protects engineering time
How a build runs with us
Review before we quote
Performance on field data, how much of the site a retrieval crawler can actually see, accessibility, and where visitors currently give up. Often this shows targeted repair gets most of the value, and we would rather say so than sell a project.
Templates first, pages second
Get the handful of templates carrying the site right on performance, accessibility and structure, then populate. Templates are what stop a site degrading the moment your team starts publishing into it at volume.
Launch, then measure
Redirects mapped, analytics and search console verified from day one, field performance checked against real visits rather than a lab score. Then fix the friction the design did not anticipate, which there always is.
No amount of rebuilding creates demand that was not there. If very few people are arriving, a rebuild gives you a nicer destination for traffic you still have to go and find — the design was never the cause. We would rather point that out before you commit budget than after launch, even when it means a smaller project for us.
Questions
- Why not just have our own engineers build the marketing site?
- They can, and some do it well. The question is whether marketing-site work is the best use of engineering capacity, and whether the result will be something marketers can update without you. In practice internally built marketing sites tend to be technically sound and permanently dependent on the person who built them.
- Our site is a single-page React app. Does that matter?
- For the product, no. For the marketing pages, a great deal. Search engines render JavaScript inconsistently and AI retrieval crawlers do far less of it, so content that only exists after hydration may be effectively invisible. Server-rendering the pages that need to be found resolves it without changing your product stack.
- Which CMS will we end up on?
- It depends on your publishing cadence, not on our preferences. If you ship content weekly and need marketers to build pages independently, a headless CMS earns its complexity. If you update quarterly, that same complexity is a cost with no return. We recommend after we understand how your team actually works.
- How do we relaunch without losing organic traffic?
- Careless redirect and content handling is how most rebuilds destroy value, so it is the part we plan first. Done properly — URLs mapped, structure preserved, content carried across intact — rankings usually hold and often improve, because the technical foundations underneath them are better than what they replaced.
- What is a realistic timeline?
- Expect eight to fourteen weeks for a focused marketing site. The build is rarely what sets the date; copy and sign-off are. Projects slip because copy is not ready or nobody is empowered to sign things off, so we agree who decides what before the first sprint rather than discovering it in week nine.
Sources
- Semrush US database, September 2026 — the 2,400 monthly search volume cited above.
- Google web.dev documentation on Core Web Vitals and field versus lab measurement.
- Google Search Central documentation on JavaScript SEO and rendering.
- W3C Web Content Accessibility Guidelines (WCAG) 2.2.
Related
Find out what your site is costing you
A free review: field performance on real visits, how much of your site an AI crawler can actually read, and the points where visitors give up before contacting you.
No obligation. Often the recommendation is repair, not rebuild.