The screens and buttons. A click, a dropdown, or a filter here is the start of the story — not the whole story.
Ten company types, one pattern
Companies in the same kind of business often pick similar stacks because they have similar jobs to do. These are typical recipes, not a company’s private internals — and not the only way to build the product.
Every example below uses the same three layers: front-end (what you see), service layer (back-end — the messenger), and database (also back-end — the storage). A click never edits the live database by itself. The browser sends an HTTPS request, the service decides what is allowed, the database runs a CRUD command, and a packaged response comes back so the page can update.
Receives the HTTPS call, checks who you are, talks to the database, and packages JSON (or similar) to send back.
CRUD is the four things a database can do: Create, Read, Update, Delete. Every click below ends in one or more of those.
01 of 10
A blog
Most common click: pick a category from the dropdown and load those posts.
A typical stack: PHP theme (WordPress) · PHP · MySQL
You open the Category menu and click Travel. The browser does not already have every post. It sends an HTTPS request asking for Travel posts only.
PHP (WordPress) receives GET /posts?tag=travel. It checks the request is valid, then asks MySQL for posts tagged Travel — not the whole archive.
It then packages those rows as a page (or JSON) to send back.
CRUD is how a database is allowed to change (or not change) data. This click is a Read: MySQL finds matching posts and hands them back. Nothing is written.
02 of 10
Shopping website
Most common click: pick a size from the dropdown and add the item to the cart.
A typical stack: Liquid or React storefront · Node.js or Ruby · MySQL + Redis
You choose size 9 and click Add to cart. The browser sends a POST so the server can remember that line in your cart.
The store’s service (Shopify, WooCommerce, or a Node/Ruby app) receives the POST. It checks stock for size 9, then tells the database to add a cart line.
Redis often holds the cart for speed. MySQL still stores the product and, later, the order.
Create means add a new row. A cart line for navy sneakers, size 9, is written. The service then reads the updated cart so the badge can say “1”.
03 of 10
Social networking
Most common click: connect with a suggested person.
A typical stack: React · Python or Java · PostgreSQL + Neo4j + Redis
Weekend hiker. You both know Anna.
You click Connect on Sofia. The React page sends an HTTPS POST. A graph database will later help suggest the next person.
Python receives the POST. It checks you are logged in, that Sofia exists, and that you are not already connected. Then it writes the link and may ping Redis so Sofia sees it quickly.
Create a connection: PostgreSQL stores the record; Neo4j stores “you — Sofia” as a graph edge so later suggestions stay smart. CRUD here means add that relationship, not rewrite Sofia’s whole profile.
05 of 10
Video streaming service
Most common click: pick a genre, then press play.
A typical stack: React · Java · Cassandra + object storage for the video files
You filter Comedy, then press Play on The Open Road. Two asks: the catalog, then a stream plus “remember my place.”
Java (or similar) checks your subscription, finds a stream URL, and starts Continue Watching. The heavy video file itself usually lives in object storage, not in the database table.
Read the title and a stream pointer. Update continue-watching as you play. CRUD does not store the movie as a giant row — it stores facts about the movie, while files sit in cloud storage.
06 of 10
Integration software
Most common click: pick apps from a dropdown and turn the automation on.
A typical stack: React · Node.js · PostgreSQL
You pick Gmail from the app menu, then Turn on. The UI is a flowchart. The click still becomes one HTTPS call.
Node.js marks the scenario as on, stores your Gmail/Slack tokens safely, and starts listening for new mail. Later, Gmail’s servers will call this service — a second company’s HTTPS call, not yours.
Update flips enabled from false to true on an existing row. CRUD is not only “add something new” — most toggles are Updates. Each later run may Create a history row.
07 of 10
Room reservation
Example click: choose a room, dates, and guest count to check availability.
A typical stack: React or Vue.js app with interactive calendar / timeline grids (FullCalendar-style) and SVG floor-plan dashboards · Node.js (TypeScript) or Python (FastAPI) reservation API: time-range math, time-zone conversion, booking CRUD; Twilio SMS + SendGrid email; MQTT / webhook integration with smart locks and door tablets · PostgreSQL with native tsrange booking ranges, PostGIS room maps and exclusion constraints that forbid double-booking + Redis 5-minute session locks
You set Austin, dates, and 2 guests, then Search. The front-end asks for listings that are actually free those nights.
The Node.js or FastAPI reservation service validates the dates, converts time zones, and asks PostgreSQL which rooms have no overlapping booking. Redis supports short-lived booking locks when you reserve a room.
Search is a Read across listings and calendars. Booking later would be a Create (reservation) plus an Update (those nights become unavailable). Same CRUD letters, different moment.
08 of 10
CRM
Most common click: filter the customer list — here, everyone in Texas.
A typical stack: React · Java or Node · PostgreSQL
You open State and choose Texas. The grid should show only Texas customers — that list lives in the database, not in the browser.
The CRM service checks you are allowed to see these contacts, then asks PostgreSQL for state = TX. Permissions matter here: a sales rep may not see every row in the company.
Read with a filter: do not download every customer. Saving this view later would Create a saved-filter row. Editing Jordan’s phone would be an Update.
09 of 10
Banking
Most common click: pick an account from the dropdown and view recent transactions.
A typical stack: React · Java or .NET · SQL Server or DB2 (core ledger is often older than the app)
You choose Checking ••44. The app asks for the last 30 days. Banks treat this as a Read of a ledger they cannot casually rewrite.
Java or .NET checks it is really you, then often calls an older core system (the legacy factor) for the official ledger — not only the app’s own database.
A statement view is a Read. Banks almost never Delete money rows — they Create a reversing transaction instead. That is why “just change a line in the database” is the wrong mental model here.
10 of 10
Design collaboration
Most common click: open a file from the list — or pick Share from the dropdown.
A typical stack: React + canvas · Node.js · PostgreSQL + object storage
You click Open. The canvas is the front-end. Layers, comments, and permissions still travel as HTTPS to a service and a database.
Node.js checks you may open this file, loads the document, and points the browser at images in object storage. Live cursors later ride a WebSocket — still the same service talking to storage.
Opening is a Read. Drawing a new rectangle would Update the document (or Create a change in version history). Share → Can view would Create a permission row.
Same three layers every time. The company type changes the example click and the recipe of tools — not the pattern. That is why a feature request is rarely “change a few lines of live code”: it has to be true in the UI, allowed in the service, and stored with the right CRUD command.
Named recipes — nicknames, not products
These are just pre-packaged sets of tools that work well together. Same three layers. Not a company. Not the only way.
The 10 types above are kinds of product. MERN, LAMP, and Serverless are nicknames for a bundle of tools. They answer different questions:
- “What is this company?” → blog, shopping, CRM
- “What is MERN?” → a pre-packaged recipe that still has the same three layers
What the letters are
Mongo + Express + React + Node
- Front-end
- React
- Service layer
- Node + Express
- Database
- MongoDB
What the letters are
Linux + Apache + MySQL + PHP
- Front-end
- PHP pages
- Service layer
- PHP + Apache on Linux
- Database
- MySQL
What the name means
Cloud functions, no always-on server to manage
- Front-end
- Still a UI
- Service layer
- Functions that wake up
- Database
- Still a database
A blog might use LAMP (WordPress). A social network might look more like MERN. Serverless can show up in any of them.
04 of 10
Social media
Most common click: like a post. (Networking is “who you know.” Social media is “what you broadcast.”)
A typical stack: React · Python or Java · Cassandra + Redis
Front-end
You tap the heart. The app sends a POST immediately so the count can jump, even while the rest of the feed stays put.
Service layer
The service records that you liked post 88 (so you cannot like it twice) and bumps the public count. Redis often holds the live number; Cassandra holds the huge like history.
Database layer
Two CRUD commands: Create your like row, Update the count from 12,408 to 12,409. Unlike a blog Read, this write has to stay fast for millions of taps.