Website Optimization Tips

Our frontend builds crossed 20 minutes. The obvious move was to rip out Webpack for something faster. We didn't. Before touching a single line of config, we profiled the entire pipeline. Webpack's ProgressPlugin turned up something we didn't expect — one plugin, Terser, was eating 13+ minutes of every build. Webpack wasn't the bottleneck. JavaScript minification was. 🔍 That one number changed the whole plan. We kept Webpack. Replaced Babel with SWC. Swapped Terser for esbuild. Moved CSS optimisation to LightningCSS. Parallelised compilation across cores with thread-loader. ⚙️ Fixed a hidden bottleneck too — AWS CodeArtifact tokens were re-authenticating every 12 hours and quietly busting our Docker cache. Restructured the Dockerfile into layers that actually cache. 🐳 20 minutes to 2 minutes. Cold build ~90% faster. Incremental ~95% faster. Across 20 production apps, 600+ PRs a month, 5,000+ CI runs. ⚡ No migration. No four-week rewrite. No new bundler to debug for the next year. The real lesson wasn't about tooling. It was that we almost skipped measurement and went straight to migration — the industry default. Glad we didn't. 💡 Full breakdown — the Speed Measure Plugin data, the pnpm switch, the exact Dockerfile changes — link in the first comment. 👇 Really proud of what our engineering team pulled off here. Worth the read. 🫡 Zepto Zepto Tech #ZeptoTechXPress #ZeptoEngineering

𝟭𝟬 𝗥𝘂𝗹𝗲𝘀 𝗳𝗼𝗿 𝗛𝗶𝗴𝗵-𝗖𝗼𝗻𝘃𝗲𝗿𝘁𝗶𝗻𝗴 𝗪𝗲𝗯𝘀𝗶𝘁𝗲𝘀 🔥 Most product sites don’t convert. Here’s how to fix it: 𝟭/ 𝗨𝗻𝗱𝗲𝗿𝘀𝘁𝗮𝗻𝗱 𝘆𝗼𝘂𝗿 𝗯𝘂𝘆𝗲𝗿 Before designing, talk to real users. Figure out what they want, what stops them, and what triggers action. → Talk to 5 signups: “What made you try it?” → Exit survey: “What’s stopping you?” → Watch session recordings → Skim support chats → Bonus: Buy someone coffee for quick feedback ✅ Example: Users say: “I just want to send invoices and get paid.” → Don’t write: “Smart billing software” → Say: “Send your next invoice in under 60 seconds.” 𝟮/ 𝗡𝗮𝗶𝗹 𝘆𝗼𝘂𝗿 𝗵𝗼𝗺𝗲𝗽𝗮𝗴𝗲 𝘀𝘁𝗿𝘂𝗰𝘁𝘂𝗿𝗲 Your layout needs: → Headline: pain point → Subheadline: curiosity → CTA: single action → Visual: product in action → Body: benefits > features ✅ Example: → “Hiring is broken.” → “Our AI recruiter finds top 3 candidates in 24h.” → “Try it free” → Demo video → “Save 10+ hours/week on screening” 𝟯/ 𝗠𝗮𝗸𝗲 𝘃𝗮𝗹𝘂𝗲 𝗰𝗹𝗲𝗮𝗿 𝗮𝗯𝗼𝘃𝗲 𝘁𝗵𝗲 𝗳𝗼𝗹𝗱 Most people won’t scroll. → What is this? → Who’s it for? → Why does it matter? → What should I do next? ✅ Example: → Don’t say: “AI-powered web builder” → Say: “Launch your landing page in 60 seconds” 𝟰/ 𝗙𝗼𝗰𝘂𝘀 𝗼𝗻 𝗼𝘂𝘁𝗰𝗼𝗺𝗲𝘀, 𝗻𝗼𝘁 𝗳𝗲𝗮𝘁𝘂𝗿𝗲𝘀 People don’t want “real-time sync.” They want fewer meetings, faster work. ✅ Example: → Don't say: “Real-time collaboration” → Say: “No more back-and-forth emails. Edit together live.” 𝟱/ 𝗔𝗱𝗱 𝗽𝗿𝗼𝗼𝗳, 𝗲𝗮𝗿𝗹𝘆 Trust builds conversion. → Logos → Quotes → Counters → Screenshots → Case studies ✅ Example: → “Trusted by 4,000+ teams at Meta, Notion, and Vercel” 𝟲/ 𝗥𝗲𝗺𝗼𝘃𝗲 𝗱𝗶𝘀𝘁𝗿𝗮𝗰𝘁𝗶𝗼𝗻𝘀 Stick to one goal and cut everything else. → No blog links → No footer clutter → No secondary CTAs ✅ Example: If your goal is “Try for free,” everything should lead there. 𝟳/ 𝗨𝘀𝗲 𝗯𝗲𝘁𝘁𝗲𝗿 𝗖𝗧𝗔 𝗹𝗮𝗻𝗴𝘂𝗮𝗴𝗲 Avoid vague buttons. Make CTAs feel easy + specific. ✅ Example: → Don't say: “Start now” → Say: “Try for free” 𝟴/ 𝗗𝗲𝘀𝗶𝗴𝗻 𝗺𝗼𝗯𝗶𝗹𝗲-𝗳𝗶𝗿𝘀𝘁 60%+ of traffic is mobile. If it’s clunky, it’s broken. → Large tap targets → Sticky CTAs → Short scroll → Preview breakpoints ✅ Example: → Desktop: CTA beside video → Mobile: CTA pinned bottom → Preview with Lovable 𝟵/ 𝗧𝗲𝘀𝘁 𝗺𝗼𝗿𝗲 𝘁𝗵𝗮𝗻 𝘆𝗼𝘂 𝘄𝗮𝗻𝘁 𝘁𝗼 One version is only one guess. ✅ Example: → “The fastest invoicing tool for freelancers” vs. → “Send your next invoice in under 60 seconds” → Ship both with Lovable 𝟭𝟬/ 𝗗𝗼𝗻’𝘁 𝘀𝘁𝗼𝗽 𝗮𝘁 𝘁𝗵𝗲 𝗖𝗧𝗔 Conversion isn’t the goal. The activation flow right after is. → Pre-fill content → Show a 60s walkthrough → Highlight one key action ✅ Example: User signs up → edits sample invoice → sends in 1 click LFG

🔮 UX Metrics and KPIs Cheatsheet (Figma) (https://lnkd.in/en9MK4MD), a helpful reference sheet for UX metrics, with formulas and examples — for brand score, desirability, loyalty, satisfaction, sentiment, success, usefulness and many others. Neatly put together in one single place by fine folks at Helio Glare. To me personally, measuring UX success is focused around just a few key attributes — how successful users are in completing their key tasks, how many errors users experience along the way and how quickly users get through onboarding to first meaningful success. The context of the project will of course request specific, custom metrics — e.g. search quality score, or brand score, or engagement score or loyalty — but UX metrics are all around delivering value to users through their successes. Here are some examples: 1. Top tasks success > 80% (for critical tasks) 2. Time to complete top tasks < Xs (for critical tasks) 3. Time to first success < 90s (for onboarding) 4. Time to candidates < 120s (nav + filtering in eCommerce) 5. Time to top candidate < 120s (for feature comparison) 6. Time to hit the limit of a free tier < 7d (for upgrades) 7. Presets/templates usage > 80% per user (to boost efficiency) 8. Filters used per session > 5 per user (quality of filtering) 9. Feature adoption rate > 30% (usage of a new feature per user) 10. Feature retention rate > 40% (after 90 days) 11. Time to pricing quote < 2 weeks (for B2B systems) 12. Application processing time < 2 weeks (online banking) 13. Default settings correction < 10% (quality of defaults) 14. Relevance of top 100 search queries > 80% (for top 5 results) 15. Service desk inquiries < 35/week (poor design → more inquiries) 16. Form input accuracy ≈ 100% (user input in forms) 17. Frequency of errors < 3/visit (mistaps, double-clicks) 18. Password recovery frequency < 5% per user (for auth) 19. Fake email addresses < 5% (newsletters) 20. Helpdesk follow-up rate < 4% (quality of service desk replies) 21. “Turn-around” score < 1 week (frustrated users -> happy users) 22. Environmental impact < 0.3g/page request (sustainability) 23. Frustration score < 10% (AUS + SUS/SUPR-Q) 24. System Usability Scale > 75 (usability) 25. Accessible Usability Scale (AUS) > 75 (accessibility) Each team works with 3–4 design KPIs that reflect the impact of their work. Search team works with search quality score, onboarding team works with time to success, authentication team works with password recovery rate. What gets measured, gets better. And it gives you the data you need to monitor and visualize the impact of your design work. Once it becomes a second nature of your process, not only will you have an easier time for getting buy-in, but also build enough trust to boost UX in a company with low UX maturity. [continues in comments ↓] #ux #design

Everyone's freaking out about GEO, LLMO, and AEO. After 7 months of running tests across tons of sites… I can tell you this: It's all built on SEO fundamentals. The same principles that rank you on Google also get you cited in ChatGPT, Claude, and Perplexity. So before you buy into shiny new tactics that promise “AI visibility”…here's what actually moves the needle: 1. Trust Signals AI tools pull from review platforms to assess business credibility and expertise. Build trust signals in the right places: - Local businesses: prioritize Google Business Profile reviews and responses - SaaS companies: maintain strong G2 and Capterra profiles  - Ecommerce: focus on Trustpilot or industry-specific review platforms - Respond to reviews professionally and keep profiles updated 2. Document Structure LLMs love well-structured documents. Instead of optimizing just for human readers, structure content for AI platforms too: - Add company context throughout documents. Instead of "our latest update," write "Acme Corp's Q4 2024 update" - Use clear headings and comprehensive sections that can stand alone - Include key facts in multiple formats (inline text, bulleted lists, data tables) 3. Link Building for Relevance Quality and topical relevance matter more than quantity for AI visibility. Focus your link building efforts: - Target industry-relevant sites where your brand mention makes logical sense - Pursue guest posts and collaborations within your industry - Don't ignore nofollow links from high-authority sites in your niche - Seek brand mentions even without direct links. (the mention itself carries weight) Avoid completely unrelated sites. 4. Topical Authority Still Rules LLMs are trained on the same web content that Google indexes. The more deep, high-quality content you publish around your niche, the more AI systems recognize you as the go-to source, the more you get mentioned. Take out the trash. Delete random blog posts about topics unrelated to your business. They're actually hurting your AI visibility. 5. Be everywhere LLMs crawl Repurpose your content across Reddit, Medium, LinkedIn, and YouTube. These platforms get crawled heavily by AI, and showing up on them regularly builds brand visibility. LLMs love patterns. The more places they see you, the more they assume you’re an authority. 6. Technical setup - Use HTML-driven pages - Add schema markup - Clean site architecture (no page more than 3 clicks from homepage) - Ensure your critical content loads server-side (most AI crawlers don't render JavaScript) 7. Traditional Search Feeds AI Most AI tools use Bing or Google's index for real-time data. Better search rankings directly improve AI visibility.

A very easy way to improve your Amazon ads efficiency by at least 10% Let’s say you’re spending ₹4–5 lakhs/month on Amazon ads. Your ACoS looks okay. Conversion rate seems fine. But your gut tells you—you’re still wasting some money on irrelevant traffic You’re not wrong At Atomberg, we had found that some of our Amazon spend was going toward search terms that had no business seeing our ads: - “cheap fan” -“rechargeable fan” - “usb fan under 1000” None of these users were in-market for a ₹3,000+ BLDC ceiling fan. But we were still showing up. And paying for those clicks. And it’s not just us. I’ve seen 6–7 brands' Amazon ad accounts across categories over the last few years—same problem, every single time The fix? N-gram analysis Takes less than an hour. You don’t need to be a performance marketing expert. But the results compound What’s N-gram analysis? It’s breaking down every search term into its word components—1-grams, 2-grams, 3-grams—and then identifying patterns that consistently drive waste… or conversion. Example: “cheap rechargeable fan for hostel room” turns into: 1-grams: cheap, rechargeable, fan, hostel, room 2-grams: rechargeable fan, hostel room 3-grams: fan for hostel, etc. When you do this across all your search terms, you start seeing the real picture. Why this matters more than just checking your search term report: Search terms ≠ keywords a) One keyword can trigger 100s of different queries. Some convert. Most don’t. You need to find the patterns. b) Waste is diluted across low-volume terms. Maybe “rechargeable fan for hostel” spent ₹300. You ignore it. But what if 12 other queries with “rechargeable” spent ₹6,000 in total with zero conversions? c) Long-tail is infinite. N-grams are finite. You can’t negate every bad search. But you can block the core terms—“cheap”, “usb”, “mini”—once and be done with it. d) It helps you scale campaigns too. You can find goldmine phrases like “white ceiling fan”, “silent BLDC fan”, “fan for living room”—with 5x+ ROAS. Those became exact match campaigns What you should do: a) Pull last 3 months of search term data b) Break them into unigrams, bigrams, trigrams c) Create a pivot with spend, orders, ROAS by N-gram d) Negate high-spend, low-conversion N-grams (e.g., “cheap”, “rechargeable”) e) Boost high-ROAS ones (e.g., “bldc”, “ceiling fan white”) f) Add exact match campaigns g) Rinse and repeat monthly Try it. Guaranteed to improve efficiency at whatever scale you are operating If you want to read an expanded version of the post, link is in the first comment

Core Web Vitals are alive & working on them can improve your conversion rate. Check out this case study from the engineering team at Nuvemshop, looking into and improving LCP with the images on their site. ✅ Removed CSS transitions for first-position sections so they render immediately. ✅ Ditched lazy loading for images at the top of the viewport. ✅ Added explicit fetchpriority="high" signals to the actual LCP candidates. The impact is nice to see: 📈 LCP pass rates jumped from 57% to 96%. ✅ CWV pass rate jumped from 48% to 72%. 🛒 Mobile conversion rates increased by 8.9%.

Dear CEO, if your organic traffic chart looks like this, don’t freak out - Google basically wiped out everyone’s clicks this year. I’ve spent 10+ years helping SaaS companies grow through search (Google + now LLMs), and I’ve never seen a shift this dramatic. This year hit almost everyone - even companies with world-class teams. And before you assume something’s wrong internally, let me say this clearly: - It’s not your CMO. - It’s not your Head of Content. - It’s not your SEO Manager. The game changed - fast. People don’t search the same way anymore. And Google isn’t the only place they look. Google changed too: - From links to answers: Google answers queries directly, pushing organic results down. - UGC takeover: Reddit, forums, communities, social posts outrank blogs. - SERP volatility: AI cards, experiments, constant reshuffles. So if you’re still measuring success by traffic, you’re using a 2019 metric in a 2025 landscape. And no - you won’t “get that traffic back.” You don’t need to. Traffic doesn’t pay your bills. Conversions do. Across dozens of GSC accounts I’ve reviewed this year, the pattern is identical: - Traffic down. - Conversion rates steady or rising. - Revenue from organic still growing. One example from the screenshot below: This SaaS lost 77% of clicks YoY - and still had its most profitable year ever with 3x more organic conversions. Because the right people still find you - even if the “what is…” fluff disappears. What matters now: - Branded demand - Homepage & direct visits - Pricing-page hits from LLMs - High-intent queries - Entity-level visibility across all search surfaces Traffic ≠ visibility. Visibility ≠ revenue. But visibility + trust = conversions. So, dear CEO, if organic is still driving revenue, stop obsessing over TOFU traffic drops. The companies winning now don’t chase traffic - they chase impact.

Just add a search bar. How hard can it be? You add an input, call an API, query the DB - done. Until real users start typing… and your entire system collapses under “simple” expectations. Here’s why 👇 Debouncing Users don’t type instantly - they type like j -> ja -> jav -> java. If you hit your backend on every keystroke, that’s 10+ calls per user. Debounce API calls (wait 300-500ms after typing stops) to avoid race conditions and server overload. Fuzzy Matching People misspell everything. “javscript” !=  “javascript”. Use fuzzy matching (e.g. Levenshtein distance) to find similar words, not identical ones. Without it, your search engine fails 20% of the time - even though the answer exists. Tokenization & Stop Words Queries like “how to implement binary search in array” need to be broken into useful tokens - [implement, binary, search, array]. Remove filler words (“how”, “to”, “in”) so your index focuses on meaning, not noise. Ranking & Relevance Search isn’t about matching - it’s about ordering. You score results by: Frequency (how often term appears) Position (title > body) Recency or popularity That’s how “React Hooks” returns a tutorial - not a random article mentioning React once. Caching & Consistency Search queries repeat - cache them smartly. But cache keys must include filters, sort options, and user tiers. Otherwise, you’ll accidentally serve premium content to free users. Also, keep DB and index in sync via event streams - search that lags behind the database breaks trust. The Lesson Search isn’t a text box. It’s debouncing, fuzzy logic, ranking, caching, and synchronization - all pretending to be one simple input.

Optimizing your site for SEO / LLMS? It's CRUCIAL that user-agents / bots are actually able to see the rendered output. Not all tech stacks are suitable out of the box - JS tech stacks are notorious for: 1. Not being easy to render even for Googlebot which can render JS but isn't overly efficient at it 2. Not being suitable for LLMs to crawl / ingest content Want to know if your site is? 1. Visit a page on your website, hit F12 and open DevTools 2. Press CTRL + SHIFT + P 3. Type Javascript and click Disable Javascript 4. Click NETWORK in the header menu of devtools 5. Ensure Preserve Log and Disable Cache are set 6. Now on your website URL append ?test and hit enter 7. On your network tab, click DOC (shown with red box in image) 8. Where it says NAME you should see ?test, click it 9. Click on RESPONSE tab (shown with red box in image) This will show you the OUTPUT of what an LLM would see. Google CAN render Javascript but, not all Javascript tech stacks are the same - if you want to see the RAW TEXT OUTPUT, do the following: In the bottom half of DevTools you will see menu items "Console, Issues, Coverage" etc. click on CONSOLE (selected by default). Click the CLEAR icon (circle with line through it) Copy and paste this command into console and hit ENTER: const text = document.body.innerText; for (let i = 0; i < text.length; i += 1000) {  console.log(text.slice(i, i + 1000)); } This will show you the RAW, NON JS output of your URL. THIS is the content that LLMS will see, and potentially Google/other search engines if there are any issues with SSR. IF YOU ARE USING JS based tech stacks you SHOULD consider SSG (static site generation) or pre-render. Effectively you want to MAXIMISE the output of RAW TEXT/HTML in the dom post JS processing before that document output is served. The amount of SITE OWNERS who have these fancy tech stacks are often oblivious to the risks/issues with using JS for a web app / site framework. Here are some solutions based on common platforms: React Next.js – [SSG + SSR + ISR] Most popular for React, supports hybrid static rendering and server-side rendering Gatsby Fully static, React-based. Uses GraphQL for data and pre-renders everything at build time Astro Partial hydration. You can use React components but ship less JS by default Vue Nuxt Offers static site generation (nuxt generate) as well as server-side rendering VitePress Lightweight Vue-based SSG, ideal for docs and content-driven sites VuePress Vue-powered static site generator, originally designed for documentation Svelte SvelteKit Can be configured for full SSG, SSR, or hybrid deployment. Excellent for fast, SEO-friendly output Elder.js Performance-focused SSG for Svelte apps (less common now but still useful) Angular Scully The go-to SSG for Angular. Pre-renders Angular apps into static HTML Analog.js Newer project aiming to bring Next.js-like DX to Angular with SSR and SSG #seo #seo2025 #aiseo

SEO plan 2025 A – Audit Your Website: Begin with a comprehensive SEO audit. Use tools like Screaming Frog or Ahrefs to identify broken links, duplicate content, and technical errors. B – Build Backlinks: Quality backlinks remain crucial. Focus on guest posting, digital PR, and creating link-worthy content. C – Core Web Vitals: Optimize for Google’s Core Web Vitals (LCP, FID, CLS) to enhance user experience and improve rankings. D – Data-Driven Decisions: Use Google Analytics and Search Console to track performance and guide your SEO strategies. E – E-A-T Compliance: Establish Expertise, Authoritativeness, and Trustworthiness in your niche, especially for YMYL (Your Money Your Life) websites. F – Fresh Content: Regularly update or add new content. Google rewards websites that stay current and relevant. G – Google Business Profile: For local SEO, optimize and maintain an accurate Google Business Profile listing. H – Headings Optimization: Use H1, H2, H3 tags properly to structure content for both users and search engines. I – Internal Linking: Build a logical internal link structure to guide users and distribute link equity. J – JavaScript SEO: Ensure content rendered via JavaScript is crawlable and indexable by search engines. K – Keyword Research: Use modern tools like Semrush or Ubersuggest to identify long-tail and intent-driven keywords. L – Link Structure: Maintain clean and SEO-friendly URLs with proper slugs and no unnecessary parameters. M – Mobile Optimization: Ensure your website is mobile-responsive, as mobile-first indexing is now the standard. N – Niche Authority: Focus on creating depth in your content to become an authority in your niche. O – On-Page SEO: Optimize titles, meta descriptions, images (alt tags), and content around target keywords. P – Page Speed: Use tools like Google PageSpeed Insights to identify and fix slow-loading pages. Q – Quality Content: Always prioritize content that provides real value to users over keyword-stuffed articles. R – Responsive Design: Adapt your site design for all screen sizes and devices. S – Schema Markup: Implement structured data to enhance search listings with rich snippets. T – Technical SEO: Fix crawl errors, sitemaps, robots.txt, canonical tags, and other backend elements. U – User Experience (UX): A seamless UX improves dwell time, reduces bounce rate, and supports SEO. V – Voice Search Optimization: Target conversational queries and FAQs for better visibility in voice results. W – Web Security (HTTPS): Secure your site with SSL – it's a ranking factor and builds trust. X – XML Sitemap: Keep your XML sitemap updated and submit it to Google Search Console. Y – YouTube SEO: If you use videos, optimize titles, descriptions, and tags for better visibility on YouTube and Google. Z – Zero-Click Searches: Optimize for featured snippets, People Also Ask, and knowledge panels. #seoexpert #seo #topratedseoexpert #seotips #expartagency