title: "The 48-Hour Before/After: How Our Website Demo Works" slug: the-48-hour-before-after-how-our-website-demo-works description: "AtlasForge ships a custom Next.js rebuild of a prospect's homepage in 48 hours and walks them through the deltas live. Here's exactly how the demo works." pillar: speed-proofs author: rj-murray publishedAt: "2026-04-25T00:00:00Z" tags: ["sales", "demo", "rebuild", "speed-proofs"] coverImage: /posts/the-48-hour-before-after-how-our-website-demo-works/cover.png coverAlt: "AtlasForge 48-hour demo before/after split" featured: false faq:
- q: "What does the 48-hour website demo actually include?" a: "A custom Next.js 16 rebuild of the prospect's homepage and one inner page, deployed to a Vercel preview URL. It ships with real Lighthouse scores, JSON-LD schema, an llms.txt file, and a side-by-side Lighthouse delta vs. the live site. The walkthrough is a 45-minute live screen-share, not a recorded video."
- q: "Is the 48-hour demo free?" a: "Yes. We ship the demo at our cost as the qualifying step in the sales process. The deliverable is real production code on a live preview URL, not a Figma mockup. It exists so the buyer can see the work before paying, and so we can decide whether the engagement is a fit on our side."
- q: "Do I get to keep the demo code if I don't sign?" a: "No. The Vercel preview URL stays live for 14 days for review, then is decommissioned. The demo is a sales artifact, not a deliverable. If you sign within 30 days, the rebuild we ship inherits the technical decisions made during the demo, and the work continues from that branch rather than restarting."
- q: "Why 48 hours and not a slower, more thorough proposal process?" a: "Because most agency proposals are decks, and a deck cannot show you that the agency can build. We can. Forty-eight hours is long enough to ship a real homepage and short enough that the buyer is still in the buying motion. It also filters our pipeline to teams who want to move."
- q: "What pricing tier does the demo lead into?" a: "Foundry ($12K-$18K + $1.5K/mo) for a 12-20 page rebuild, Atlas ($35K-$60K + $3.5K/mo) for a 30-60 page rebuild with pSEO and a content engine, and Empire ($80K-$150K + $8K-$15K/mo) for full marketing-surface rebuilds with multi-axis pSEO, GEO pages, AEO work, and a managed content pipeline. The demo itself does not change the tier; the scope of the live site does."
The 48-hour demo is the sales motion. A prospect sends us their URL, two days later we hand them a Vercel preview link with a custom Next.js rebuild of their homepage and one inner page, plus a side-by-side Lighthouse delta against the live site. The walkthrough is 45 minutes, live, screen-share, no deck. It costs us a couple of agent-days and it closes more business than any cold email ever has.
What the 48-hour demo actually contains
The deliverable is real production code on a live preview URL, not a Figma board and not a recorded video. The homepage is rebuilt in Next.js 16 with Tailwind v4, deployed to Vercel, with a clean JSON-LD graph (Organization, WebSite, BreadcrumbList, and the relevant business-type schema from schema.org). One inner page comes with it, usually the highest-traffic service or product page from their existing analytics.
We ship the page with the same stack we ship every client on. Next.js 16, Tailwind v4, MDX for any long-form content, Payload v3 wired up to MongoDB Atlas if the prospect needs a CMS in the production build, Resend for any form handling, PostHog for the analytics. The demo runs on the same code paths as the live engagement. There is no demo-only shortcut.
The package the prospect receives is four things in one email. The Vercel preview URL. A one-page Lighthouse delta with their current scores next to ours, captured the morning of the call. An llms.txt file at the root of the demo so we can show what AEO discoverability looks like under the new stack. A scoped quote for the full rebuild, tier-mapped to Foundry, Atlas, or Empire.
What is intentionally not in the demo: a copy rewrite of every page, a fully migrated blog, a finished design system, or 30 pages of content. The demo is a structural and performance proof. It shows the buyer what shipping with us looks like, not what their finished site looks like.
How we run the rebuild in 48 hours
The first eight hours are recon. Our recon-crawler walks the prospect's site, captures the marketing surface (skipping blog pagination, archives, and obvious noise), pulls full-resolution media from any CDNs, and writes a state file we can resume from. We cap the crawl at 30 pages because the demo only needs structure, not the entire archive. By the end of recon we have a pages map, a content extract per page, and a media library.
The next twelve hours are the brand pass. The brand-forge pipeline runs intake, then direction, then tokens, then the component scaffold. We do not invent a new brand for the demo. We extract the prospect's existing palette, type stack, and logo from their live site and reproduce them at production fidelity in Tailwind v4 tokens, with the heading rules inside @layer base so they actually take effect. If the existing brand is genuinely broken we flag it on the call rather than silently replacing it.
Hours 20 to 36 are the build. Homepage and one inner page in Next.js 16, hand-written, no template fork. JSON-LD inserted server-side, Core Web Vitals instrumented from the first deploy, llms.txt generated from the page graph, sitemap.xml emitted statically. The page ships with the same component primitives our production clients ship with, including the gallery module that pulls Instagram archives when a trades client doesn't want a second photo shoot. That same module is what we used on the Karpentor rebuild.
The last four hours are audit. We run our internal /visual-audit against the preview URL and require a 9 of 10 or higher before we send the link. We capture the Lighthouse delta on a clean profile, headless Chrome, mobile and desktop, both runs from the same machine on the same network so the comparison is honest. If the demo fails the audit we hold the call for a half-day rather than send work we would not ship to a paying client.
What the live walkthrough covers
The walkthrough is 45 minutes and runs four things in order. Lighthouse delta first because it is the most concrete. We open the live site, run Lighthouse from the audience side of the screen, and put it next to the demo URL running the same audit. On every demo we have run the deltas land in roughly the same range: Performance from the 30s to 95+, Accessibility from the 70s to 100, Best Practices from the 80s to 100, SEO from the 80s to 100. The actual numbers are in the real-Lighthouse-scores post for six of our shipped rebuilds.
Schema is second. We open the live site in Google's Rich Results Test, find the four or five schema errors that almost every WordPress site carries, then open the demo URL and walk the JSON-LD graph node by node. The point is not to humiliate the existing site. The point is to show that schema correctness is a production concern we treat as non-optional, and that our build ships with it from day one. AEO citation surfaces, ChatGPT, Perplexity, Claude, Gemini, depend on schema being clean.
Cover-the-pages is third. We open a Google Search Console export of their top 50 pages by impressions, then walk down the list and show how each page type maps onto the new architecture: service pages, geo pages, case studies, product pages, blog. This is where the buyer realizes the demo is not a homepage replacement but a structural template for the whole rebuild. It is also where we catch scope. If the GSC export shows 400 indexed product pages, that is a tier conversation, not a homepage conversation.
Mobile is fourth and last because mobile is where most mid-market sites quietly fail. We open the demo on the buyer's phone (we ask them to pull it up live), then their existing site, in that order. The contrast is usually more visible on the phone than on the desktop comparison, especially for sites still on a 2018 WordPress theme. Core Web Vitals changed in 2025, and INP is the metric most existing builds will never pass without a rebuild.
Why this works as a sales tool. What it filters for.
The demo wins because most agency pitches are decks. A deck cannot show that the agency can build. We can, so we do, and the prospect can see the work in their own browser inside two days of the first call. That collapses the part of the buying cycle where the buyer is calling references and asking for portfolio links, because the demo is the portfolio link, except it has their logo on it.
It also filters for the buyers we want. A CMO who reschedules the demo call twice and ghosts the third one is a CMO who is not buying. A CMO who shows up on time, brings their CFO and a head of growth, and asks pointed questions about retainer scope is the buyer who closes. We have run roughly 40 demos to date. The close rate inside 30 days is north of 35 percent, which is several multiples of what a deck-based pitch ever produced.
The demo also filters us. We have walked away from demos where the prospect's existing site is on a custom Java stack with deep CMS coupling that we are not the right team to migrate, and where the brand position is so confused that a rebuild would not fix the underlying business problem. Walking away from a deal in the demo phase costs us two agent-days. Walking away three months in costs everyone real money, so the demo is also a cheap insurance policy on engagement quality.
The structural reason this works at all is that we built the production pipeline before we built the demo. The demo runs through the same /website-rebrand and /website-forge flows that ship our paid client work. There is no demo team and there is no production team. Two-day turnaround is possible because we rebuilt 20 sites in 90 days using this same engine, and the demo is a 1-page slice of that work.
Pricing tiers it leads to
The demo does not change the price. The scope of the live site does. We map every demo to one of three tiers on the call, before the prospect asks.
Foundry is $12K to $18K plus $1.5K per month, scoped at 12 to 20 pages, no pSEO, single-vertical service business. This is the right tier for the trades shop, the boutique medical practice, the law firm with one office. Burris and Sons Heating ($1917 family HVAC, 30-page rebuild) was Atlas-scoped because of the eight neighbourhood geo pages. Knightsbridge Doctors was Foundry-scoped because the booking system stayed and the surface was 19 pages.
Atlas is $35K to $60K plus $3.5K per month, scoped at 30 to 60 pages with one pSEO axis, a content engine seeded with 10 to 20 launch posts, and case-study production wired up. Most of our 2026 closes land in Atlas. Karpentor sat at the upper end of Atlas: 36 pages in 24 days, 10 geo pages for the towns they actually serve, 8 launch blog posts on scope, permits, and material choice, plus the Instagram-archive gallery module so they did not need a second photo shoot.
Empire is $80K to $150K plus $8K to $15K per month, scoped at full marketing surface with multi-axis pSEO, GEO pages at scale, AEO work, and a managed content engine with weekly publish cadence. This is the tier for manufacturers with brand-by-product matrices, regional service businesses with 50+ markets, and SaaS companies with serious organic ambition. Raydon Accounting is our first Empire-shaped retainer, and the day-45 case study will be our first publish with full Search Console data and attributable demo requests.
The recurring number is not a maintenance fee. It is the content engine, the reporting framework, and the ongoing pSEO and AEO work that takes a launched site and turns it into a ranking site. Most agencies bill the rebuild and disappear. The retainer exists because the rebuild is the start of the engagement, not the end.
When the 48-hour demo is the wrong fit
The demo is wrong when the buyer is not ready to talk numbers. If the call is exploratory, if there is no budget conversation on the other side, the demo is wasted on both ends. We screen for this in the first 15-minute discovery, before we burn agent-days. The signal is whether the prospect can name the line item the rebuild lives under and roughly what range it sits in.
The demo is wrong when the existing stack is not WordPress, Webflow, Squarespace, or static HTML. We build on Next.js 16. We can migrate from headless WordPress, classic WordPress, Webflow, Squarespace, Wix, and most static sites cleanly. Drupal, Sitecore, AEM, and bespoke Java CMSes are real engagements, and we either decline or scope them as a multi-month migration with a different sales motion. The demo is for the WordPress-and-Webflow majority of the mid-market.
The demo is wrong when the buyer wants Shopify. We do not build on Shopify for the kind of mid-market B2B marketing sites we specialize in. The right fit there is a Shopify-native agency, and we send the referral without running the demo.
The demo is wrong when the brand position is unresolved. If the buyer cannot say in one sentence what their company does and who it sells to, a rebuild is not the work that needs to happen first. Brand and positioning work happens before the demo, not during it. We have had demos derail because halfway through the call the marketing lead and the CEO disagree about who the customer is. That is a real problem. It is just not a problem the demo solves.
The demo is also wrong if the buyer has more than one stakeholder who has to be sold separately. The demo is a sales artifact, and a single 45-minute live walkthrough that ten people cannot attend together is a worse asset than a written proposal with screenshots. For those engagements we send a Loom recording of the demo plus the preview URL, with paid search killed off the budget conversation entirely, and we wait for the room to align before we re-engage.
If the demo is the right fit, you will know within the first 15 minutes of the discovery call. Send us the URL. We will hand back a preview link in two days. The walkthrough is on the calendar inside a week. If you are deciding which tier the rebuild lands in, the demo is the cheapest way to find out, and a 90-day organic growth plan is the second-cheapest. Either way, the next move is to book the call.