Most sites calling themselves price comparison engines work from a stored catalogue. Prices are collected on some schedule, saved, and served to you later. It's cheap to run and it produces a familiar experience: you click a ₹18,499 price and land on a page showing ₹21,999.
We built 7Compare the other way round, which is harder and occasionally slower, and this is what it actually does.
Step 1: a live parallel scan
When you search, your query goes out to nine store adapters simultaneously — Amazon, Flipkart, Myntra, Meesho, AJIO, Croma, Snapdeal, Nykaa and Decathlon.
Each adapter is a piece of code that knows how to read one store's listings: where the title sits, where the price sits, how that store formats an MRP, how it marks an out-of-stock item. They run in parallel rather than one after another, because nine sequential requests would take long enough that nobody would wait.
What comes back is what exists right now: titles, current prices, stated MRPs, images, and links. Nothing is served from a catalogue that was assembled last week.
The honest trade-off: this is slower than reading from a database, and stores occasionally rate-limit or return nothing. When a store is unreachable we say so rather than showing you a stale price we saved earlier — an old price presented as current is worse than an absent one, because it makes a decision for you on bad information.
Step 2: matching identical products
This is the genuinely hard part, and the part that makes the difference between a useful comparison and a misleading one.
Consider what the same product looks like across four stores:
- "Sony WH-1000XM5 Wireless Bluetooth Headphones (Black)"
- "SONY WH1000XM5 Noise Cancelling Headphone Black"
- "Sony WH 1000 XM5 Over Ear Headphones with Mic"
- "Sony WH-1000XM4 Wireless Headphones (Black)"
Three of those are the same product. The fourth is a different, older model, and to a naive text-similarity check it looks more similar to the first than the third one does.
The engine extracts model codes and variant markers from these messy titles and normalises them, so WH-1000XM5, WH1000XM5 and WH 1000 XM5 collapse to the same identity — while XM4 stays firmly separate. It applies the same logic to the variant dimensions that matter:
- Storage: 128GB never matches 256GB.
- Size: a 43-inch TV never matches a 50-inch one; a 1-ton AC never matches 1.5-ton.
- Tier markers: "Pro", "Plus", "Max" and "Lite" are treated as identity, not decoration.
- Quantity: a pack of two is not the same listing as a single unit.
And a design decision worth stating plainly: when there isn't enough evidence to be confident two listings are the same product, we show them separately rather than grouping them. A wrong match is far more damaging than a missing one — it tells you a store is cheaper for a product it doesn't actually sell, and you find out at checkout. Under-grouping is a mild inconvenience; over-grouping is a lie.
Step 3: ranking, and the honesty checks
Within a matched group, prices are sorted low to high and the cheapest store is highlighted. That's the simple part.
The less simple part is discounts. Each store states its own MRP, and as covered elsewhere on this blog, stated MRPs are sometimes fiction — particularly on unbranded goods. So discount percentages are computed against the store's own claim rather than being taken at face value, which lets you see when a large percentage is doing more work than the actual price is.
Results are cached for about 30 minutes. That's a deliberate compromise: repeat searches for popular products return instantly instead of hammering the stores, and half an hour is short enough that prices rarely move meaningfully within it. After that window, the next search re-fetches.
Step 4: the background scanning
Separately from your searches, the engine works continuously through popular categories — groceries, fashion, electronics, home, kids — checking prices across stores and recording what it finds.
That's what fills the live catalogue and the hot deals feed, refreshed roughly every ten minutes. Nothing there is manually curated: if a store drops a price meaningfully against its own stated MRP, it appears. Nobody at 7Compare decides which deals are worth showing, which is why the feed sometimes contains genuinely random things.
The same background process powers price tracking. Anything you track is re-checked automatically across stores, and a real drop triggers a browser notification — even if the site isn't open. If you decline notification permission, the alert simply waits for your next visit.
Why matching is harder than it looks
It's worth spending a little longer on the matching problem, because it's the part people assume is easy and it's where comparison sites usually fail quietly.
Retail listing titles are written by whoever uploaded the product — sometimes the brand, usually a seller, occasionally an automated feed with its own conventions. There's no shared identifier you can rely on across stores. The same product can carry different names, different capitalisation, different word order, extra marketing words, and specifications listed in a different sequence.
Meanwhile, genuinely different products can look almost identical in text. A single character often separates a 3-star air conditioner from a 5-star one, or one storage variant from another. Colour names differ across stores for the same shade. Bundle listings include an accessory in the title in a way that reads like a description rather than a difference.
The approach that works is layered: extract anything resembling a model code, normalise the punctuation and spacing, identify variant markers separately from the model, and require agreement on all of them before grouping. Where the evidence is thin — no model code, generic title, unbranded product — the safe answer is to group nothing.
The result is that comparison quality varies by category, and it's honest to say so. For electronics with model numbers, it's reliable. For branded fashion with product codes, good. For unbranded goods, there is no comparison to make, and any site claiming otherwise is grouping things that aren't the same.
What we deliberately don't do
We don't take payments. You always check out on the store's own site, at the store's own price. We never see or ask for card details, and there's no 7Compare wallet, checkout or account holding your money.
We don't require an account. Tracking and preferences live in your browser's own storage. That means your watchlist is tied to that browser — a different phone has its own list — which is the trade-off for not collecting anything about you.
We don't let commissions reorder results. This is the one that matters most, so it's worth being specific. 7Compare earns affiliate commissions when a purchase follows a click through to a store. Those commissions differ by store. The lowest price is highlighted regardless, and the sort order is price, not payout.
That isn't primarily an ethical stance — it's a practical one. A comparison site that quietly promoted whoever paid best would be detectable within a few weeks by anyone who checked the stores directly, and the entire value of the site is that you don't have to check. Rigging the results destroys the only thing being sold.
The limitations, stated honestly
Scrapers are fragile. Stores redesign their pages, and adapters break. We write them with multiple fallback patterns, but expect the occasional store to return nothing on a given day.
Unbranded goods can't really be matched. Two ₹350 kurtis from two sellers are genuinely different products, so there's no honest way to group them. In those categories the engine shows you listings rather than a comparison, and no tool can do better — it's a property of the market, not of the software.
We only see what stores publish. Prices visible after login, member-only pricing and personalised offers are invisible to us. Bank offers, which frequently matter as much as the base price, depend on your card and have to be checked at the store.
Coverage isn't total. Nine stores is a lot of the Indian market, not all of it. A product sold exclusively somewhere else won't appear.
Why it's free
Two revenue sources, both worth naming plainly. Affiliate commissions paid by stores on referred purchases — at no extra cost to you, since you pay the store's normal price. And advertising on the site.
That's the entire business model. It only works if the number we show you is the number you actually pay, which is why the engineering effort goes into live fetching and conservative matching rather than into a bigger catalogue that would be cheaper to run and quietly wrong.
If you find a case where a highlighted price doesn't match what the store shows, that's a bug worth knowing about — usually a broken adapter or a matching error — and it's the kind of thing that gets fixed rather than tolerated.
How to get the most out of it
Three habits make the engine considerably more useful, and they take no extra time.
Search the model number, not the product name. "WH-1000XM5" produces a far cleaner comparison than "sony headphones", because the matching has something unambiguous to work with. For anything with a model code — electronics, appliances, branded goods — this single change improves the result more than anything else you can do.
Include the variant in the query. Adding the storage, the size or the capacity narrows the results to what you're actually buying and removes the risk of eyeballing a cheaper price that belongs to a different configuration.
Track the things you're planning, not the things you're browsing. Two or three tracked products produce useful alerts. Twenty produce noise you'll start ignoring within a week, which defeats the purpose entirely.
And one thing worth doing outside the tool: once you've got the shortlist down to the two cheapest stores, open both listings and read the return terms and the seller rating before choosing. The engine gets you to the shortlist in seconds; that last thirty seconds is still yours, and it's the part that prevents the purchases people actually regret.
7Compare compares live prices across every store that stocks a product, keeps the price history, and tells you whether today is a good day to buy.
Compare a price
