Lesson 3 of 8
Perfect Manual Sourcing
Working a retailer's site by hand: which pages to start from, and how to tell a real discount from a permanent price.
This is the original way to do online arbitrage: open a retailer's website and work through it by hand. It is slow, it needs no software beyond a free app, and it is worth learning even if you intend to automate, because it is the only way to understand what a tool is doing for you.
The loop
The example uses Argos, but the method is identical anywhere: Walmart, Target and the rest in the US; Argos, Tesco and the rest in the UK.
- Go to the clearance or sale section. This is where the price gaps are, because a retailer clearing stock is pricing to move it rather than to make a margin.
- Narrow to a category. Pick one you will get to know. The example uses Toys. Working one category repeatedly builds a feel for normal prices that browsing everything never does.
- Pick a product and copy its title.
- Search that title on Amazon and find the matching listing.
- Check the numbers: selling price, fees, sales rank, price history, and whether Amazon is on the listing.
A worked example
The lesson shows the arithmetic on a real item. The product costs £15 at the retailer and sells for £49 on Amazon. Entering the cost gives a profit of £18.51, which is over 100 percent return on investment. Amazon is not selling it.
Everything so far says buy. The lesson then declines it, and the reason is the important part of the whole lesson: the product has no sales rank.
No sales rank means no evidence that it sells at all. The margin is excellent and the demand is unknown, so the money could sit in Amazon's warehouse indefinitely, paying storage fees. A high return on something that does not sell is not a good deal; it is capital removed from the business.
This is the discipline the whole course rests on. Margin qualifies a product. Demand decides it.
The faster method
Copying titles into Amazon is tedious and it is also unreliable, because a title match is not an identification. Different sizes, colours, multipacks and editions share almost identical titles, and attaching your stock to the wrong listing causes returns and account damage.
The better technique, and the practical point of the lesson: use the Amazon Seller app to scan the barcode on your monitor. Open the app, tap the camera, and point your phone at the barcode shown on the retailer's product page. The app reads it off the screen exactly as it would off a box.
That gives you an exact identification rather than a guess, plus the sales rank, the fee calculation and your profit, in one action.
The trap the example shows
The second example is worth studying. A product costs £26.99 at the retailer. On Amazon, one matching item has a sales rank around 70,000 and sells near £60, which looks strong. Scrolling further down the results reveals another item selling at exactly £26.99, the same price as the retailer.
The lesson is that the first match is not necessarily your product. Amazon returns several similar listings, and it is easy to compare the retailer's price against the wrong one and calculate a profit that does not exist. Check that the item you are looking at is genuinely the item you would be buying: same size, same quantity, same edition, same colour.
This is exactly the failure mode a barcode scan avoids.
Extensions that speed up the check
Browser extensions such as Keepa, SellerAmp SAS and BuyBotPro overlay the history and the profit calculation directly on the Amazon page, which removes the switching between tabs. Keepa in particular is worth installing before your first session, because the price and rank chart is the thing you should be checking on every candidate.
What manual sourcing is for
Realistically it is slow. Working a clearance category by hand might produce a handful of buyable products in an hour, and that is the ceiling.
Its value is in the learning. After a few sessions you know what normal prices look like in your category, which retailers discount usefully, and what a deal that fails looks like as well as one that works. That judgement is what you will use to evaluate a software-generated list later, and it does not come from any other exercise.
The next lesson is about where to find candidates once the clearance pages have been picked over.