A website project fails long before a single line of code is written. It fails in the kickoff call that ended with 'we'll figure the content out later'. It fails in the vague brief that says 'make it look modern.' It fails in the missing sign-off on the sitemap. Over eight years and 200+ launches, we've learned that shipping fast and shipping well are not opposites — they are both consequences of a process that forces the ugly conversations to happen on paper, not in production.
This is the exact eight-step process we run for every website engagement at RoTechX, whether the client is a solar installer in Bhavnagar or a D2C brand shipping across India. It is not glamorous. It is not novel. It is not agile-with-a-capital-A. It is a straight line from an unclear brief to a launched site that actually moves a business metric.
Timeline
Small marketing sites (5–10 pages) ship in 3 weeks. Standard business sites (10–25 pages) ship in 4–5 weeks. E-commerce and multi-language ship in 6–8 weeks. These timelines assume the client is responsive on approvals; the process pauses if they are not.
Step 1 — Discovery
The discovery call is a 60-to-90-minute working session with the founder and, wherever possible, the person who owns marketing. We do not send a questionnaire. Questionnaires get filled out on autopilot and lie to us. A live conversation catches the throwaway comment — 'oh we tried Meta ads last year and burned ₹4L' — that changes the entire strategy.
By the end of discovery we have written answers to four questions: who is the buyer, what is the buying journey, what business outcome is the site expected to move, and what is the honest six-month history of everything the business has already tried. We share the notes with the client the same day. Any factual disagreement gets resolved before we spend another rupee of their money.
Step 2 — Research
Research is the step most agencies skip because it has no obvious deliverable the client can point at. We do it anyway because it is the difference between a site that looks like a competitor's and a site that beats one.
Competitive teardown
We pick five direct competitors and three out-of-category benchmarks — companies who are excellent at something we want to borrow. For each, we document positioning, primary CTA, information architecture, top three traffic channels (via SEMrush), and one thing they do that we should steal.
Search-intent mapping
The top 30 keywords the business could plausibly rank for, grouped by intent — informational, transactional, navigational. This map becomes the skeleton of the sitemap. It is also the first place we push back on the client's original page list, because half the pages they wanted are not what buyers are searching for.
User interviews (when budget allows)
Three to five 20-minute calls with the client's existing customers. Cheap, uncomfortable, and the single highest-ROI research activity we do. Assumptions kill conversion rates. Customers, told the right way, will tell you exactly what would have made them buy faster.
Step 3 — Strategy
The strategy deliverable is a single-page document. One page, deliberately. If a strategy needs more than a page to explain, it isn't a strategy — it's a wishlist. The page contains: the one-line positioning, the primary CTA, up to two secondary CTAs, the information architecture (as a bullet list, not a fancy diagram), and the 'so what' of every planned section.
The strategy doc is signed off in writing before any design work begins. This sign-off is the most important checkpoint in the entire process. Every disagreement that gets skipped here becomes a design revision, a development revision, and eventually a re-launch.
Sign-off ritual
We literally ask the client to reply 'approved' to a specific email. It sounds bureaucratic. It has saved us from six-figure scope disputes.
Step 4 — Design
High-fidelity design happens in Figma. We ship the homepage first as a directional exercise — three visual directions, one round of feedback, then commit. Committing to a direction early prevents the death-by-a-thousand-tweaks that kills so many design projects.
Every interactive state is designed: hover, focus, loading, empty, error, disabled. If a state is undesigned, a developer will invent it, and that invention will not match the brand. Empty states in particular are where cheap sites reveal themselves. A well-designed empty state on a search page is worth more than an animated hero.
Design system, not screens
For every project longer than five pages, we build a lightweight design system: colors, typography scale, spacing scale, button variants, form elements, and card patterns. Screens are then composed from the system. The upfront cost is one extra day. The downstream cost saving is roughly 30% of development time, and the site stays consistent as it grows.
Step 5 — Development
Our default stack is React with TanStack Router, Tailwind, and TypeScript. Every project is custom-built. We do not use page builders. We do not install 40-plugin WordPress themes. This is not stack snobbery — it is a business decision. Custom builds ship faster to Core Web Vitals targets, are cheaper to maintain past year one, and do not carry the ongoing security burden of a plugin ecosystem.
We enforce performance budgets from day one: Largest Contentful Paint under 2.5 seconds on a mid-range Android phone on 4G, Cumulative Layout Shift under 0.1, Total Blocking Time under 200ms. Every pull request runs Lighthouse in CI. If the budget is broken, the PR does not merge. This discipline is why our sites score 95+ on Lighthouse a year after launch, not just on launch day.
Step 6 — Quality Assurance
The QA checklist is 60 items long and shrinks by roughly two items every project — because we fix root causes, not symptoms. QA covers cross-browser rendering (Chrome, Safari, Firefox, Edge), cross-device (iPhone SE up to iPad Pro, low-end Android up to Samsung flagship), accessibility (WCAG 2.1 AA), content accuracy, form validation, 404 handling, redirect map correctness, and analytics event validation.
The launch-day fire we've stopped having
Two years ago we launched a site with a broken contact form. The client lost 40 enquiries in 48 hours. Now, form submissions are QA'd end-to-end — including the email hitting the client's inbox — before we go live. Every time.
Step 7 — Launch
Launch day is deliberately boring. If launch day is exciting, something has gone wrong. Our launch checklist runs in a specific order:
- 1Freeze the new site in a staging environment identical to production
- 2Redirect map: every legacy URL mapped to a new URL, with 301 status codes, verified with an automated crawl
- 3DNS TTL lowered to 300 seconds 48 hours before the cutover
- 4Analytics validated on staging (GA4, Meta Pixel, GTM containers)
- 5Search Console submission of the new sitemap the moment DNS propagates
- 6First-hour monitoring: real-user metrics, error logs, form submissions
Zero-downtime cutovers are the norm, not the exception. If a launch requires downtime, we schedule it at 2 AM the client's local time and communicate it three days in advance. Silent downtime is how you lose customer trust.
Step 8 — Growth
The launch is the start line, not the finish line. Every project includes a Month 1 review — a 60-minute session where we look at real user behavior, heatmaps, form conversion, top exit pages, and the two or three highest-impact iteration opportunities. From that review comes a prioritized backlog.
For clients on our growth retainer, we ship one iteration cycle every two weeks: hypothesis, change, measure, decide. This is where a good website becomes a great one. Sites that are treated as 'launched and done' quietly decay — content goes stale, offers become outdated, and the CMS turns into a graveyard. Sites that are treated as a living asset compound.
Why This Process Exists
Every step in this process was born from a specific project that went sideways. The one-page strategy came from a client who signed off on a 40-slide deck and then argued about it for three months. The design-state discipline came from a launch where the empty search state was, in production, the word 'undefined'. The QA form-submission check came from the lost 40 enquiries. Process is scar tissue. We are sharing ours so you don't have to earn it the hard way.
Frequently asked questions
How long does a typical project take?
3 weeks for a small marketing site, 4–5 weeks for a standard business site, 6–8 weeks for e-commerce or multi-language. Timelines assume the client is responsive; the process pauses when approvals slow down.
Do you work with existing brand guidelines?
Yes. If a brand system exists, we extend it. If it doesn't, we build a lightweight one during design. We rarely recommend a full rebrand as part of a website project — it doubles the timeline and dilutes focus.
Can you migrate content from our old site?
Yes. Content migration is scoped in Step 3 based on what actually deserves to be migrated. Roughly half of every legacy site's pages should not survive the move.
What happens after launch?
Every project includes a 30-day defect-fix warranty and the Month 1 review. Ongoing changes are either ad-hoc (hourly) or on a growth retainer with a bi-weekly release cadence.
Do you use WordPress or a custom stack?
Custom by default. We only use WordPress when the client has strict CMS requirements and the editorial team is already trained on it. For everything else, custom is faster, safer, and cheaper past year one.
Conclusion
A website launch is not a design exercise; it is a business project with a design phase inside it. The eight steps above exist to keep the business goal in the middle of the conversation, from the first discovery call to the Month 12 review.
Written by
Ajay Bhojani
Co Founder & CTO · RoTechX
