DEV Community: Ravi Rai

Claude Code vs Human Developers: Where Each Actually Wins in 2026

Ravi Rai — Thu, 23 Apr 2026 22:28:00 +0000

Every client conversation in 2026 starts with the same quiet question: 'Do we even need a developer anymore? Can't Claude Code just do this?' It's a fair question and it deserves a fair answer. We use AI coding tools every day on client work — we also hire and get hired as humans every week. This is the honest view from inside that overlap: where Claude Code and similar tools have genuinely replaced developer tasks, where they've not, and where the whole industry is fooling itself.

What AI coding has actually gotten good at

Stop reading what AI evangelists tweet. Here's what we've observed across real client projects in the last year:

Scaffolding a brand-new Next.js + Tailwind + Shadcn project with a landing page, contact form, and basic auth — 15 minutes of prompting vs 4-6 hours of developer time
Writing the first pass of CRUD endpoints for a Laravel or Node.js API when the schema is clear — genuinely 80% of the way there
Translating a Figma screenshot into passable Tailwind + React — not pixel-perfect, but good enough to ship as a v1
Writing tests for existing code — arguably Claude is now better at unit tests than the average mid-level developer
Writing Shopify Liquid templates and WordPress plugin boilerplate from a spec
Migrating a Next.js 14 project to Next.js 16 — routine chore work the AI handles well when given the changelog
Generating SQL schema + seed data + ER diagrams from a plain-English product description

If a freelancer's main offer in 2023 was 'I build landing pages fast,' that person is in trouble in 2026. The tools are faster and cheaper, and the client can run them in a Cursor window while drinking coffee.

What AI coding still gets wrong — at scale, in production

Here's where AI coding breaks down on actual client engagements:

Architecture decisions that don't come out in a single prompt — choosing between headless Shopify and fully-custom Next.js, deciding if Laravel or a serverless stack fits the compliance profile, estimating database cost at 100k users
Debugging complex production issues — when the bug is in the interaction between three services, Claude still suggests plausible-but-wrong fixes about 40% of the time
Client-specific context that was never written down — 'Why is this feature a priority?', 'What does the sales team actually do with this data?', 'What did the previous dev do and why was it removed?'
Navigating vendor limits that aren't well-documented — Stripe rate limits for B2B billing, Shopify's undocumented webhook retries, AWS account-level quotas
Knowing when NOT to build something — a senior developer will tell you a feature is a bad idea; AI will happily build the bad idea well
Long-term code hygiene — AI generates code that's idiomatic but inconsistent across sessions; keeping a codebase coherent over six months still needs human ownership
Hard client conversations — explaining why the timeline slipped, negotiating scope, saying no to feature creep, managing stakeholder expectations

AI coding tools are extraordinary interns. They're fast, tireless, and they never push back. Which is also the problem — on real client projects, you desperately need someone who pushes back.

How we actually use Claude on client work

We use Claude Code every day at buildbyRaviRai. We're transparent with clients about it. Here's the exact split:

Scaffolding: Claude writes the first 80% of most new features. A human reviews, rewrites ~20%, and takes ownership of the shipped version.
Unit tests and types: mostly Claude, reviewed by human.
CRUD and admin panels: Claude scaffolds, we refine — speed increase here is genuinely 3-4x.
Architecture and system design: entirely human. Claude is a rubber duck, not a decision-maker.
Client communication, scoping, requirements gathering: entirely human.
Code review: Claude reviews pull requests as a second opinion. Human has final say.
Bug triage in production: human owns, Claude assists with search and hypothesis generation.

Net effect: we ship about 2.5x more code per month than we did in early 2024. Not 10x, like the AI hype cycle promised. Real productivity gains exist — they're also much more modest than LinkedIn suggests.

When a client should NOT hire a developer — they should just use AI

We turn down client work every month because the honest answer is 'you don't need us.' If your situation matches one of these, you're genuinely better off with a Cursor license, 20 hours of your own time, and a decent tutorial:

A simple brochure website for your consultancy — 5 pages, contact form, modest traffic. Claude + Vercel template + a weekend will give you something that competes with a ₹1L agency build.
An internal tool used by 5 people where UX polish doesn't matter — a spreadsheet is usually better than a custom app anyway, but if you insist on an app, AI will build it.
A static marketing site for a personal brand — the scaffolding is trivial, and the content matters far more than the code.
A proof-of-concept you'll throw away in a month — paying for production quality is a waste when you're validating an idea.

When you should hire a human (still)

You're building a real product that will have real users, real payments, and real support tickets. AI writes the code; a human still owns it when things go wrong at 2am.
Your product touches money, health, compliance, or personal data. Regulatory context is not well-captured in training data — it's in documents that aren't public and conversations that aren't recorded.
Your requirements are ambiguous and you need someone to help you figure out what to build. AI is excellent at building specs; it's mediocre at forming them.
You're replatforming, migrating, or modernizing — the hard part is understanding the existing system and the business it serves, not generating new code.
You need a partner who can explain technical tradeoffs to your non-technical board, cofounders, or customers.

The economics — what does hiring cost now vs AI-plus-your-time

Rough numbers from clients who chose each path:

AI-only path (founder or marketer doing it themselves)

Cursor Pro or Claude Code subscription: roughly ₹1,800 to ₹3,500/month
Your time: 40-80 hours to ship a real product v1 if you're not a developer
Typical outcome: something that works for your first 10-100 users, breaks by 1,000
Cost of rewrites once you outgrow the AI-scaffolded version: usually 1.5x to 2x what a proper build would have cost

Freelance agency hybrid (what we offer)

₹1.5L to ₹6L for a production-grade v1, depending on complexity
Your time: 15-25 hours total across 4-8 weeks (requirements, review, decisions)
Typical outcome: something that survives 10x traffic growth and can be extended for another 18-24 months without a rewrite
Ongoing retainer: ₹30k to ₹80k/month for maintenance, occasional feature work, and being on-call

If your project's expected revenue is above roughly ₹50L/year, the human-led path almost always pays back within 6-12 months. Below that, DIY with AI and move faster; pay for humans when the stakes go up.

What changed for freelance developers this year

Five real shifts we've lived through at the agency:

Junior-level freelancers doing template implementations got squeezed hard. Rates are down 30-50% for that tier since late 2024.
Senior freelancers with architecture and product thinking command MORE than they did — clients now see the gap between 'someone who can code' and 'someone who can design a system' more clearly.
Specialization wins. 'Shopify developer' out-earns 'full-stack developer' because the product context is harder for AI to substitute.
Project managers, QA, and ops disappeared from most small teams. A single senior developer with AI assistance now covers roles that used to need 3 people.
Client expectations on speed went up. What used to be a 4-week MVP is now often expected in 2 weeks. Delivering on that without quality loss is the new senior-dev superpower.

The bottom line

In 2026, the question 'AI or human developer?' is the wrong framing. The right framing is 'which parts of my project benefit from AI leverage and which parts need a human owner?' For most real products the answer is 'both, deliberately split.' That's how we work, how our best clients work, and — honestly — it's how most serious software teams will be working for the next several years.

If you're weighing whether to hire a freelance developer or just use Claude yourself, we'll tell you honestly which path fits your project — even if the honest answer is that you don't need us. That's the only way the relationship works long-term.

Not sure whether you need a developer or just AI + your time? → Get a Free Honest Assessment

Building a Laravel Admin Dashboard: Architecture Lessons from 5 Client Projects

Ravi Rai — Thu, 23 Apr 2026 22:21:54 +0000

Admin dashboards are one of the most consistently underestimated pieces of any web app. The client-facing store gets the investment, the marketing polish, the design love. The admin panel gets whatever is left over — and then breaks under the weight of every feature added in year two. Over five client Laravel admin builds in the last three years, we have settled on a set of architectural patterns that keep these panels maintainable past year one. This is the condensed version.

Start With the Right Toolkit, Not a Blank Slate

The single biggest mistake we see in Laravel admin projects is building the dashboard from scratch with Blade and raw Eloquent queries. It feels cheaper at first and ends up costing 3x once you are two months in and writing the 14th 'list with filters' page.

Our default stack in 2026:

Filament PHP — our go-to admin panel framework. Resource-based, TALL stack under the hood, great extensibility. Ships 80% of what you need in week one.
Laravel Sanctum — for API token-based auth when the admin has a companion mobile app or third-party integration.
Spatie's laravel-permission package — roles and permissions, battle-tested, widely understood by the Laravel community.
Spatie's laravel-activitylog — audit trail for every admin action. We install this on project one, not when a client first asks who changed a record.
Laravel Horizon — queue monitoring when we have background jobs. Free, bundled, works.

Honorable mention: Nova is excellent but commercial ($199/developer/year) and we have shifted most new builds to Filament because it is free and the community has surpassed Nova in extension count.

Authorization Is Not Just Login

New Laravel devs often treat 'can this user log in' as the whole authorization story. Production admin panels have at least four layers:

Authentication — who is this user? Laravel's defaults handle this well.
Authorization — what roles do they have? spatie/laravel-permission gives you roles and permissions tables.
Policy — which records of a given type can they see or edit? For example, a regional manager should only see orders from their region. Laravel policies + Eloquent global scopes are the right tool here.
Data masking — which fields of a record can they see? An ops person might see a customer's name but not their full payment details. This usually means attribute accessors that check the current user's role.

Wire all four layers in week one even if your MVP only has two roles. The cost of adding them later, after 200 features reference the old assumption of 'any authenticated admin can do anything', is enormous.

Audit Logs Are Non-Negotiable

At least once per project we have had a client ask, months after launch, 'who approved this order?' or 'who changed this customer's address?' If you did not install an audit log on day one, you cannot answer. Install Spatie's activitylog on project day one, turn it on for every sensitive model, and surface a Filament resource that lets authorized staff browse the log.

Good audit log conventions:

Log who, what, when, and the full before/after diff of changes — disk is cheap
Capture IP address and user agent for security-sensitive models (user accounts, permission changes, financial records)
Purge after 2 to 5 years depending on industry compliance — never purge without a documented retention policy
Never log secrets in clear — password changes should show 'password changed', not the password itself

Data Tables: The Feature That Eats Your Month

Every admin panel has 10 to 40 'here is a list of X with search, filters, sorting, and pagination' pages. Building these one at a time is a trap. The pattern:

Use Filament's Table Builder — it handles pagination, sorting, filters, bulk actions, column visibility, and CSV export with one configuration object
Standardize on three filter types: date range, enum (status), and relationship (belongs-to). Everything else is usually premature.
Add server-side search only on fields backed by a database index. Full-text search without an index is the first thing that kills a dashboard at 100k records.
For tables larger than 500k rows, switch to Meilisearch or Typesense indexes rather than LIKE queries

Background Jobs: Where Dashboards Fall Over

Admin panels accumulate long-running operations: 'export all orders from last month' , 'email all customers in segment X', 'recompute tier for every user'. Running any of these in a request cycle is the fastest way to a 60-second timeout and a bad day.

Our default setup: Laravel Queues with Redis as the driver, Horizon as the monitor, and a convention that every admin action that might take more than 2 seconds is queued with a progress indicator in the UI. For exports specifically, Spatie's laravel-excel + chunked queries + a download notification is what we reach for every time.

Performance: The Mistakes That Hurt At Scale

N+1 queries — install Laravel Debugbar in development and Telescope in staging. N+1 is the single most common performance issue we find in client code reviews.
Eager load relationships in admin list views aggressively — joins are cheap, hundreds of round-trips are not
Use chunk() or lazy() for any loop over large result sets — loading 100k Eloquent models into memory at once will kill your process
Cache aggregate dashboards — the homepage of most admin panels runs 10+ count queries that rarely need to be real-time. Cache for 5 minutes.
Use Eloquent accessors/casts sparingly in list contexts — they run per-row and can quietly make tables 10x slower

Deployment: Boring Choices Win

We deploy Laravel admin panels to Laravel Forge + a DigitalOcean or AWS droplet 80% of the time. Forge costs $12/month, provisions the server, handles SSL, queues, deployments, and monitoring. For projects that need it, we upgrade to Ploi or Envoyer — but Forge is the sensible default.

We do not deploy production Laravel admin apps to serverless (Vapor, Lambda) unless the client has specific reasons. The cold start penalty on admin actions and the cost model rarely beat a small Forge-managed VPS for the kind of traffic admin panels see.

The Non-Code Stuff That Matters Most

The engineering patterns above make a dashboard maintainable. But what actually makes a dashboard successful is the things we do outside the code:

Interview the actual users before design — the intern doing returns, the customer success lead doing bulk refunds. Build for them, not for imagined power users.
Deploy early and iterate — a good-enough admin panel in production beats a perfect one on staging for six weeks
Keep the admin UI extremely boring and predictable — this is not where you experiment with glassmorphism
Document every non-obvious admin action in a 1-page Notion doc linked from the admin header. The person who replaces the current ops lead will thank you.

Building a Laravel admin panel and want a second opinion on architecture? → Book a Free Architecture Review

Next.js E-commerce vs Shopify: When a Custom Build Actually Wins

Ravi Rai — Thu, 23 Apr 2026 22:21:53 +0000

Every six months a wave of 'Shopify is dead, go headless on Next.js' content washes across Twitter, LinkedIn, and YouTube. Every founder who sees it seriously reconsiders their stack. This is an honest look at when a custom Next.js e-commerce build is the right choice — and the much larger number of cases where it is a bad decision dressed up as a good one.

What 'Next.js E-commerce' Actually Means in 2026

There are three distinct flavors of Next.js e-commerce and the economics of each are very different:

Headless Shopify — Shopify handles checkout, inventory, and payments. Next.js renders the storefront via the Storefront API. Best of both worlds for some cases, expensive overengineering for others.
Headless Medusa, Vendure, or Saleor — open-source commerce engines behind a Next.js front-end. Maximum control, no platform fees, you are fully responsible for the checkout.
Fully custom — Next.js + Stripe + your own database. Works for specific niches (subscriptions, creator stores) but generally a rewrite of problems Shopify solved a decade ago.

The Commercial Reality Most Posts Skip

A fully-loaded Shopify store (theme, apps, development) costs a founder roughly ₹1.5L to ₹4L in year one for a mid-size store. A decent headless Next.js + Shopify rebuild costs ₹6L to ₹15L in year one. The question is not whether Next.js can match Shopify features — it can. The question is whether the specific benefits of custom justify 3x to 5x the investment.

Going headless is almost never about technology. It is about whether a 2-second faster page load, a more memorable brand experience, or a content-heavy marketing layer will add enough incremental revenue to pay back the rebuild.

When Next.js E-commerce Wins (Legitimately)

You have an existing content operation — a blog, editorial brand, media property — and commerce is an add-on to content rather than the other way around. Gymshark, Allbirds, Rhone are examples of brands where the content-commerce blur justifies custom.
You are in a category where brand experience is the product. Luxury fashion, limited-drop streetwear, bespoke furniture, curated subscription boxes. Shopify themes simply do not produce the feel these brands need.
You have specific performance requirements that Shopify cannot hit — product pages that need to render 10+ configurable variants dynamically, real-time pricing from a backend ERP, or tight integration with a proprietary recommendation engine.
You are running internationally with currency, tax, and language complexity that Shopify Markets still does not handle cleanly for your countries.
You plan to sell the commerce platform itself — a few SaaS companies have genuinely bootstrapped new Shopify competitors from their own Next.js commerce code.

When Next.js E-commerce Loses (More Often)

You are a new founder and need to ship in weeks, not months. Shopify will have you selling in 7 days. Custom Next.js cannot beat that.
You do not have a developer on retainer or a comfortable budget for ongoing maintenance. A custom e-commerce site without a maintainer quickly becomes a security and SEO liability.
Your catalog is under 500 SKUs and standard in structure. Shopify's defaults handle this perfectly. Every feature you would build from scratch already exists in a Shopify app.
Your team does not include a designer with e-commerce sensibilities. A bespoke Next.js store with mediocre UX performs worse than a default Shopify Dawn theme. Always.
You are doing this because 'headless is the future' and not because of a specific pain you cannot solve on Shopify. The future of commerce is not headless — it is whatever matches each brand's commercial needs.

The Headless Shopify Middle Ground

The most commercially defensible Next.js e-commerce setup for most growing brands is headless Shopify. You keep Shopify's checkout, payments, fraud detection, tax, and inventory. You build the product and content experience on Next.js. You pay the Shopify plan plus your development costs, but you do not reinvent checkout or lose out on Shop Pay conversion.

Hydrogen — Shopify's official React/Remix framework for headless — has matured since it first launched. But we still reach for Next.js for headless Shopify builds because our team knows it better and it works well with Storefront API queries through a thin GraphQL layer.

How We Actually Decide on Client Projects

When a client comes to us with an existing Shopify store asking if they should go headless, here is the actual conversation we have:

What specific problem is Shopify not solving? If the answer is vague or aesthetic, we steer them toward a theme upgrade instead.
What is your monthly revenue and gross margin? If revenue is under ₹50L/year, a full rebuild rarely pays back inside 24 months.
Do you have a designer with e-commerce experience on staff? If not, the custom path will almost certainly regress your conversion rate.
Can you commit to a 2-year horizon of continuous development? Headless commerce is not a project; it is an ongoing product.

Most founders fail at least two of those four questions. We tell them honestly that a headless rebuild is the wrong move and recommend a Shopify theme investment or a Shopify Plus upgrade instead. That is harder to sell in a pitch deck but it is the advice that makes our clients actually succeed.

Thinking about going headless? Let us stress-test the idea with you. → Book a Free Tech Strategy Call

How to Build a Shopify Store in India: Cost, Setup, and Launch Checklist

Ravi Rai — Thu, 23 Apr 2026 19:19:34 +0000

Most 'how to build a Shopify store' articles floating around the Indian search results are written by SEO farms who have never actually launched a store. This one is the opposite — it is the step-by-step process we use on every Shopify engagement with Indian founders, with real rupee costs, the exact settings we configure, and the mistakes we see new store owners make before the first real order comes in.

Step 1: Decide If Shopify Is Actually Right For You

Shopify is not universally the right choice even though it is frequently the default. Before you pay for a plan, be honest about these three questions:

Is your catalog straightforward (physical products with standard variants like size and color), or do you have something unusual like configurable furniture, subscription boxes, or booking-based services? Shopify handles the former perfectly and the latter only with expensive app stacks.
Do you have the budget for a monthly plan plus a few essential apps, or are you extremely cost-sensitive for the first 6 months? WooCommerce has lower ongoing costs but needs a developer on call.
Do you plan to run aggressive email and retargeting marketing? Shopify's ecosystem of Klaviyo, Gorgias, Recharge, Shop Pay, and Aftership genuinely gives you a head start compared to any other platform.

If you answered yes, yes, and yes, keep reading. If you hesitated on any of them, read our WordPress vs Shopify comparison before committing to a plan.

Step 2: Pick the Right Plan (INR Pricing)

Shopify pricing changes with the USD-INR rate and occasional regional promos, but as of April 2026 the standard Indian pricing looks like this:

Shopify Starter (social-only selling): ₹399/mo — skip this unless you are genuinely only selling through Instagram DMs
Basic Shopify: ₹1,994/mo (~$24) — correct starting plan for most founders shipping under ₹25L/year
Shopify: ₹7,447/mo — worth upgrading to when you cross roughly 100 orders a month or need better transaction fees
Advanced Shopify: ₹30,164/mo — only makes sense past ₹1 crore annual revenue or when you need the advanced reporting
Shopify Plus: starts around ₹1.6L+/mo — enterprise tier, unless you are doing ₹5+ crore a year, ignore it

The free 3-day trial followed by the $1/month-for-three-months promo is still running in most regions — grab it. Even if your store is not ready, lock in the promo and build during the promo period.

Step 3: Register Your Indian Business (Required)

To accept payments in INR via Shopify Payments or Razorpay, you must have a legal Indian business entity. In practice that means one of:

Sole proprietorship with GSTIN — fastest to set up (₹3,000 to ₹8,000 via a CA, ~1 week), fine for solo founders doing under ₹20L/year
Private Limited company — more overhead (₹8,000 to ₹15,000 + ongoing compliance), but required for serious fundraising or hiring full-time staff
LLP — middle ground, mostly used for small partnerships

You also need a current account (a regular business bank account) in the business name, a GSTIN once you cross ₹20L turnover (₹10L for certain states), and an Import Export Code if you plan to ship internationally.

Step 4: Domain, Logo, and Core Identity

Before you install a single app, nail down:

A .in or .com domain (we recommend buying both if available and 301-redirecting one to the other) — typically ₹800 to ₹1,500/year
A minimum-viable logo — a clean wordmark is enough for launch. Do not spend weeks on this. Canva or a ₹5,000 Fiverr designer gets you to launch. Upgrade later.
A brand color palette of 2 to 3 colors, written down, used consistently on every page, ad, and email
A clear one-line brand promise — what do you sell and to whom? Write it in 10 words or fewer, pin it to the theme header

Step 5: Theme Selection and Customization

Start with Dawn, Shopify's free flagship theme. It is fast, well-maintained, and handles 80% of standard store needs. Only buy a premium theme when Dawn has a specific limitation you cannot work around — and even then, a theme by Out of the Sandbox, Switch, or Archetype is more reliable than anything on the Shopify Theme Store top-sellers list.

Theme cost: free for Dawn, $200-400 for a solid premium theme (one-time). If a developer is configuring and customizing for you, budget ₹30,000 to ₹1L depending on complexity.

Step 6: Essential Apps (Budget ~₹5,000-15,000/mo Total)

After 40+ Shopify client builds, this is the lean stack we default to:

Klaviyo — email and SMS marketing. Free up to 250 contacts, then around ₹1,500-6,000/mo. Non-negotiable.
Judge.me — product reviews. Free tier is generous; paid is ₹1,200/mo for branded emails.
Shopify Inbox — free, bundled chat widget. Start here before paying for Gorgias.
SEO Manager or Plug in SEO — ₹800-1,500/mo. Automates schema, meta, and redirect management.
PageFly or Shogun — page builder for landing pages. ₹2,000-4,000/mo. Only if you plan to run ad campaigns to dedicated landers.
Aftership — shipping tracking pages. Free tier is enough for under 50 shipments/mo.

What to NOT install at launch: upsell apps, bundle apps, cross-sell apps, currency converters (Shopify handles this natively now), and any 'boost sales' app with more than 10 features. These slow your store, clutter your admin, and solve problems you do not yet have.

Step 7: Payments, Shipping, Taxes

Shopify Payments is available in India now (as of late 2024) and is usually the best default for Indian stores. Transaction fees are around 2% for Basic plan customers. If you can use Shopify Payments, you avoid the extra 2% transaction fee on third-party gateways.

For shipping, integrate Shiprocket, Delhivery, or Shyplite directly via their official Shopify apps. This gets you automated label printing, rate shopping, and tracking updates. Manual shipping does not scale past about 10 orders a week.

For GST, you need to configure 18% default tax rates and capture GSTIN on B2B orders. Shopify's built-in tax engine handles the basics; for anything complex, install the GST India app.

The 30-Point Pre-Launch Checklist

Store Setup

Store name, logo, and favicon uploaded (favicon is easy to forget)
Currency set to INR
Timezone set to India Standard Time
Legal business name, address, and GSTIN in the store details
All checkout email templates reviewed and branded
Test customer account created end-to-end
All product descriptions, images, and alt text complete
Collections organized (do not launch with 50 uncategorized products)
At least 3 collection pages with hand-picked featured products
About Us, Contact, FAQ, Shipping Policy, Return Policy, Privacy Policy, Terms of Service, Refund Policy pages written and linked in footer

Technical

Domain connected, SSL verified, www and non-www both redirect correctly
Google Analytics 4 and Google Tag Manager installed
Meta Pixel installed and test events verified in Events Manager
Google Merchant Center connected (for Shopping Ads later)
Sitemap.xml submitted to Google Search Console
All homepage meta title and description set (Shopify defaults are generic)
Structured data verified with Google's Rich Results Test on at least one product page
PageSpeed Insights score above 60 on mobile for the homepage and a popular product page

Payments and Fulfillment

Shopify Payments (or Razorpay) fully verified with bank details confirmed via test INR ₹1 order
At least one successful real checkout with a real card, refunded
Shiprocket/Delhivery account connected, carrier rates shown correctly at checkout
Test order printed and label generated end-to-end
Abandoned cart email flow active (built-in Shopify recovery, or Klaviyo)
Order confirmation email tested
Shipping confirmation email tested

Marketing Foundations

Welcome email flow written and scheduled (3 emails minimum)
Instagram Shopping connected and at least 20 products tagged
At least one friends-and-family discount code ready to share
Post-purchase upsell flow set (optional but high-ROI)
UTM conventions documented so paid ads can be tracked cleanly

Launch Day and the First 30 Days

Quiet launches beat loud ones. Share the store with 20 friends and family before you announce publicly — their feedback catches broken flows that you missed. Then do a small paid ads test (₹10,000 on Meta is enough to learn what creative works). Only after you have your first 10 organic sales should you scale ad spend.

In the first 30 days, resist the urge to add new features or redesign pages. Instead, obsess over the three metrics that matter: add-to-cart rate, checkout completion rate, and average order value. Every change you make should be in service of one of those.

Want us to build your Shopify store end-to-end? → Get a Free Scoping Call

Next.js 16 in Production: What We Learned Shipping Real Client Projects

Ravi Rai — Thu, 23 Apr 2026 19:19:33 +0000

Next.js 16 landed late last year with a set of changes that looked, on paper, like mostly incremental improvements. Six months and five production client deployments later, we have a clearer picture of what actually shifted and where you still get tripped up. This is not a release notes recap — plenty of those exist. This is the honest production view from a small agency that has shipped it.

Turbopack as the Default: Real Impact, Real Edges

The loudest Next.js 16 change is Turbopack becoming the default bundler for both dev and build. The developer experience improvement is real — builds for our largest client project dropped from 2 minutes 40 seconds on webpack to 21 seconds on Turbopack. Dev startup is effectively instant.

The rough edges we hit in production:

Static export (output: 'export') works on Turbopack now, but one of our CI deploys produced an out/ directory that was missing the sitemap.xml route — we had to switch to the webpack build for that one project as a workaround
A couple of less-maintained plugins in our stack were not Turbopack-compatible and silently produced stale bundles. The error was not obvious — pages just did not reflect code changes until we figured out the culprit
Build output differs slightly between webpack and Turbopack in ways that break some third-party bundle analyzers

Our rule of thumb: stick with Turbopack for new projects. If you are migrating an existing Next.js 13 or 14 app with a long plugin history, test the Turbopack build against every major route before trusting it in CI.

App Router is Now Genuinely the Default

It took three major versions but the App Router now feels like the right way to build. The mental model clicks once you accept that:

Server components are the default and usually the right choice
Client components are an escape hatch for interactivity, not a starting point
Loading, error, and layout boundaries are first-class — not bolted on
Parallel and intercepting routes solve real UI patterns that were genuinely hard before (modals as routes, dashboards with multiple concurrent loading states)

The App Router stopped feeling like a new way to build React apps and started feeling like React finally grew up.

The Caching Story Is Still The Trap

If there is one place where Next.js 16 has burned us in production, it is caching. The defaults changed in 15 and were tuned again in 16 — and every client project we took from staging to production had at least one 'why is this page serving stale data' incident.

What We Do Now On Every Project

Read the revalidation behavior of every fetch before shipping — do not assume the default matches what you want
Use { cache: 'no-store' } explicitly for user-specific data instead of relying on the default
Tag-based revalidation (revalidateTag) for anything content-driven like blog posts or product catalogs — so our CMS can trigger a revalidation webhook on publish
Avoid unstable_cache for now unless you have a concrete reason — the footgun-to-value ratio is still too high

Server Actions: Great for Forms, Risky for Everything Else

Server Actions feel like the future for form submissions. One function, no API route boilerplate, progressive enhancement works out of the box, and the DX for client-to-server data flow is excellent. We use them on every contact form and every admin mutation in client projects.

Where we stopped using them: public-facing mutations that need to scale, or anywhere we might want a third-party client (mobile app, partner integration) to hit the same endpoint later. For those, we still write proper API route handlers — partly because Server Actions are implicitly tied to the Next.js runtime, and partly because they do not compose well with rate limiting and request tracing in our stack.

Deployment: Vercel Is The Easy Path, And That's Fine

We have shipped Next.js 16 to Vercel, Netlify, AWS (via OpenNext), and self-hosted Node on a VPS. Our takeaways:

Vercel is the frictionless path. Deploy-per-branch, preview URLs, image optimization, edge functions — everything works without configuration. For most client work, this is the correct default.
Netlify is fine for static export with output: 'export'. If you need SSR, their Next.js runtime has improved dramatically in the last year but still lags Vercel on edge cases.
OpenNext on AWS Lambda is viable if you have compliance reasons to stay on AWS, but budget 2 to 3 extra days for the setup compared to Vercel
Self-hosted Node with a reverse proxy is simple, but you are now responsible for image optimization, request coalescing, and a build pipeline. Only worth it for specific reasons.

When We Still Do Not Reach for Next.js

Next.js is our default for most web work. But we do not use it for:

Simple static brochure sites — plain HTML or Astro is lighter and cheaper to host
Internal tools with mostly-interactive dashboards — Remix or plain React + Vite is often a better fit
CMS-driven content sites where the client needs to edit pages — WordPress or a headless CMS + a simpler framework usually wins
Anything where SEO does not matter and latency is dominated by API calls — a single-page React app is less overhead

Should You Upgrade If You're On 14 or 15?

If you are on 14, upgrade when you next have a two-week window of bandwidth. The migration is straightforward and the performance gains are worth it. If you are on 15, the 16 upgrade is smaller and you can probably do it in an afternoon on a well-tested codebase.

If you are still on the Pages Router on 13, this is a good moment to plan the App Router migration. The ecosystem has stabilized, the tooling for incremental migration works, and you will be dragging the Pages Router further into legacy territory every month you wait.

Building with Next.js and need an experienced team? → Talk to Us About Your Project

WordPress vs Shopify for E-commerce in 2026: An Honest Comparison

Ravi Rai — Thu, 23 Apr 2026 19:03:53 +0000

Nearly every e-commerce client we have worked with in the last three years has asked the same question at the start of the project: WordPress with WooCommerce, or Shopify? The answer depends less on which platform is 'better' and more on the specifics of your catalog, your margins, your team, and how much custom behavior you need. This is the honest version, written by a team that ships on both.

The Core Tradeoff in One Sentence

WordPress + WooCommerce gives you maximum flexibility at the cost of ownership and maintenance. Shopify gives you a fully managed platform at the cost of customization limits and a monthly revenue share. Everything else flows from that tradeoff.

Where WordPress + WooCommerce Wins

You already have content, a brand, or an existing WordPress site — WooCommerce plugs into that without a rebuild
Your catalog is unusual: made-to-order, configurable products with dependencies, B2B pricing tiers, rental models, subscription boxes with custom billing logic
You need deep SEO control — WordPress gives you direct access to every URL, meta tag, schema output, and page-level redirect rule
You want to avoid platform revenue share — you pay for hosting and a dev once, not 2% of every sale forever
You have international tax complexity that pre-built Shopify apps cannot handle cleanly

WooCommerce powers roughly 28% of all e-commerce sites globally as of early 2026. It is battle-tested at every scale, from a single product landing page to catalogs of 100,000 SKUs. The ecosystem of plugins and developers is deeper than any other platform.

Where Shopify Wins

You are launching a standard DTC store and want to be selling in 2 weeks, not 2 months
You do not have an in-house developer and do not want one — Shopify handles hosting, security patches, PCI compliance, and SSL automatically
You rely heavily on the modern marketing app stack: Klaviyo, Gorgias, Recharge, Zendesk, Aftership all have native Shopify integrations
You want Shopify Payments and Shop Pay — one-tap checkout genuinely moves conversion rates up 2 to 4 percentage points on mobile
You need a point-of-sale system that syncs with your online inventory — Shopify POS is the clear winner

The Real Cost Comparison (Year One)

Founders routinely underestimate WordPress total cost of ownership and overestimate Shopify's. Here is what a small store looks like in year one on each platform:

Shopify (Basic Plan)

Monthly plan: $32/mo = ~₹32,000/year
Theme (Dawn or premium): $0 to $400 one-time
Essential apps (Klaviyo, SEO Manager, Judge.me): roughly ₹8,000 to ₹15,000/year
Transaction fees: 2% of revenue (if not using Shopify Payments)
Dev costs for setup and customization: ₹50,000 to ₹1.5L
Year one total for a ₹50L revenue store: roughly ₹1.5L to ₹3L

WordPress + WooCommerce

Managed hosting (Kinsta, WP Engine, or solid Indian host): ₹15,000 to ₹60,000/year
Premium theme + builder: ₹5,000 to ₹15,000 one-time
Plugins (Rank Math Pro, WP Rocket, security, backup): ₹15,000 to ₹30,000/year
Payment gateway: 2% to 3% per transaction (Razorpay, Stripe — paid to the gateway, not WooCommerce)
Dev costs for setup, customization, ongoing maintenance: ₹1L to ₹3L in year one
Year one total for a ₹50L revenue store: roughly ₹1.5L to ₹4L

The costs are actually similar in year one. The divergence happens in year two onward. On Shopify you keep paying the same monthly revenue-linked cost forever. On WordPress your ongoing cost drops to mostly hosting and occasional maintenance — usually 30 to 50% less than Shopify at steady state. But if you do not have a reliable developer, the savings on paper get eaten by unfixed bugs, slow sites, and security incidents.

Migration Paths: Escaping When You Outgrow the Choice

Good news: both migrations are well-trodden. We have done them in both directions.

Shopify to WooCommerce: typically 4 to 8 weeks. The hard parts are preserving customer accounts, order history, and SEO via 301 redirects for every product URL.
WooCommerce to Shopify: typically 3 to 6 weeks. The hard part is adapting custom plugin functionality to Shopify apps or Liquid customizations — and accepting that some custom behavior will need to be compromised.
Both: expect to spend roughly ₹2L to ₹5L on a clean migration for a store under 1,000 SKUs. Much more if you have years of content, blog posts, and custom pages to preserve.

Our Honest Default Recommendation

If you are a new founder with a straightforward DTC catalog, start on Shopify. You will be selling faster, you will not need a developer on retainer, and the platform takes responsibility for uptime. Revisit the decision at ₹5 crore annual revenue when the platform fees start mattering.

If you have an existing WordPress site with content traffic, or your catalog has complex configurability, or you are building a multi-brand store, WordPress with WooCommerce is almost certainly the right choice. Budget for a long-term development partner — this is not a 'set it and forget it' platform.

Need help picking between WordPress and Shopify for your store? → Request a Free Platform Consultation

Hire a Freelance Web Developer in India: The Complete 2026 Guide

Ravi Rai — Thu, 23 Apr 2026 19:03:52 +0000

If you are a founder, marketing lead, or product manager trying to get a website or web app built without wasting months and rupees, hiring a good freelance web developer in India is one of the highest-leverage moves you can make. The catch is that the market is noisy: thousands of listings on Upwork, Fiverr, LinkedIn, and Reddit claiming they can ship anything. This guide gives you the honest version of what to actually look for, what to pay, and how to structure the engagement so you get what you need.

Freelancer vs Agency vs In-house: When to Choose Which

Most founders treat this as three separate conversations. It is actually a single question: how much do you want to be involved in the build, and how much risk can you take on schedule and quality?

A solo freelancer gives you the lowest cost and the most direct communication — but if they get sick, go on vacation, or disappear, your project stops.
A traditional agency insulates you from individual risk but adds account managers, project managers, sales overhead, and roughly a 2x to 3x price premium for the same output.
An in-house hire makes sense only when you have continuous product work for the next 12 to 18 months. For one-off builds, it is almost always wrong.

The sweet spot for most small-to-mid projects is what we call a freelance-agency hybrid: a small senior team (usually 2 to 4 people) that works with the directness of a freelancer but has bench capacity to cover illness, specialization, and post-launch support. This is how buildbyRaviRai operates and it is how most of our clients prefer to work — direct conversations with the people actually writing the code, with someone backing them up.

What to Actually Look for When Hiring

Skip the flashy portfolio pages. They are usually designed and built by someone else on the team. Instead, ask for these four things:

A live URL to something they shipped for a real client in the last 12 months — not a template, not a demo on their own domain.
A 15-minute screen share where they walk you through the code of that live URL. If they cannot, they did not actually build it.
References from two past clients, ideally in a similar industry. A one-hour reference call tells you more than a month of chat.
A clear answer to: 'What is one thing you would have built differently on that last project?' A developer who cannot self-critique has not shipped enough.

The best freelancers are not the ones with the cleanest portfolio. They are the ones who can tell you honestly what broke in production and how they fixed it.

— Ravi Rai

Red Flags That Should End the Conversation

Quotes without asking detailed scoping questions. If someone gives you a price in the first email, they are selling templates.
Unrealistic timelines. A proper e-commerce build with custom design cannot be done in 10 days. Anyone who says yes is setting you up for a rewrite.
No contract. A freelancer who does not send a basic SOW or contract before collecting a deposit is one who disappears.
Portfolio pieces that are all the same kind of project. If every project looks like a Shopify theme clone, that is all they can build.
Communication only through WhatsApp. Casual chat is fine — but important decisions, scopes, and invoices need a written paper trail.

Pricing Expectations in the Indian Market (2026 Rates)

Rates in India vary dramatically by skill tier and location. Here is what we see in the market right now:

Junior freelancers (1 to 2 years experience): ₹500 to ₹1,200 per hour, or ₹40,000 to ₹80,000 for a basic brochure website
Mid-level freelancers (3 to 5 years): ₹1,500 to ₹3,000 per hour, or ₹1.5L to ₹3L for a business website with CMS
Senior freelancers and small agency teams (5+ years): ₹3,000 to ₹6,000 per hour, or ₹3L to ₹8L for a production-grade web app
Specialized senior freelancers (Next.js, Shopify Plus, Laravel at scale): ₹5,000 to ₹10,000+ per hour for consulting work

If someone is quoting you well below the junior range for senior work, you are either going to get a junior doing the work in secret, or a template with your logo pasted on. Both outcomes end in a rewrite.

Run a Paid Trial Before Committing

The single best risk-reducer when hiring a freelancer is a small paid trial. Pick a contained piece of work — one page, one feature, one bug fix in their stack — and pay for it at their normal rate. Two weeks and ₹30,000 spent on a trial will save you six months and ₹5L on a bad full engagement.

What you are evaluating in the trial is not just code quality — it is communication, how they handle ambiguity in the brief, and whether their estimate matches reality. A freelancer whose trial runs 3x over the estimate will run 3x over on the real project.

Why Hire a Freelance Developer From India Specifically

The stereotype of 'cheap Indian freelancer' is outdated. Top Indian freelancers now compete directly with US and European rates on quality, and actively choose to stay freelance rather than join Big Tech. The advantages for international clients are:

Overnight delivery for US-based clients — work happens while you sleep, you wake up to updated builds
Strong English communication, especially in Tier-1 cities like Bangalore, Noida, Pune, Hyderabad, Gurgaon
Well-established payment infrastructure (Wise, Stripe, Payoneer) for compliant cross-border payments
Deep talent pool for modern stacks: Next.js, React Native, Shopify, Laravel, WordPress are all mature ecosystems in India

Putting It All Together

Hiring a freelance developer well is fundamentally about treating it like hiring an employee, compressed. Write a real scope document. Take references. Run a paid trial. Sign a contract. Do all of that and you will get 10x the result of whoever is bidding lowest on Upwork this week.

If you want a shortcut to the first conversation, we offer free 30-minute consultations where we will honestly tell you whether your project is a fit for us, a fit for a larger agency, or whether you should hire in-house. No pitch, no sales deck.

Looking to hire a freelance web developer in India? → Get a Free Consultation