Nothing Up Our Sleeves: The Unexciting Procedure Every Price Must Survive

The Editorial Staff has been asked, more than once, how a price ends up on this website. The question is usually posed casually, in the tone one reserves for small talk. The honest answer is a procedure, and procedures are not small talk. We have therefore written it down.

What follows is the whole thing. Not a summary of the whole thing, and not merely the reassuring parts of it. Every threshold quoted below is a constant in our own software, and every one of them is doing exactly one job: making it harder for something to get published than to be quietly discarded.

The Provenance of a Number

Before anything can be judged, it must be measured against something that cannot be staged.

A list price cannot be that thing. A list price is a number a human typed into a field, and it can be typed again on a Tuesday afternoon for reasons having nothing to do with the product. A rolling average of what a product has actually sold for, hour after hour, is considerably harder to arrange. So our price data comes from a continuous Amazon price-tracking feed, and the reference figure we care about is the ninety-day average. Our software re-sweeps every product on the watchlist every four hours and recomputes everything described below from scratch. Nothing is cached and then trusted. There is no stored verdict at all, only a stored price history and a formula that runs again.

We note, without enthusiasm, that this makes our own past claims perishable. A deal described accurately in the morning may be inaccurate by evening. The procedure exists so that the website notices before you do.

Ten Percent, and Not a Rhetorical Ten Percent

A product enters the site when its live price sits at least ten percent below its ninety-day average. Ten percent is not a mood. It is a comparison performed on two numbers, and a price that misses it by a fraction does not get in on the strength of being interesting.

The reverse gate matters more, and almost nobody builds one. When a price recovers to within three percent of its ninety-day average, the deal is pulled automatically. It does not wait for a person to notice. It does not linger for a few days because the traffic was good. The retirement rule runs on the same sweep as the entry rule, which means our deals page routinely removes an item while the retailer is still advertising it as a sale.

Deals arriving from secondary channels, such as community feeds or editor submissions, do not skip any of this. They enter as unpublished drafts and stay unpublished until the same price verification passes against our own tracked data.

The Arithmetic, Disassembled

Every Deal Score is a sum rather than an opinion. It begins at two points, awarded solely for the existence of verified tracked price data, and then adds four components.

Depth below the ninety-day average carries up to five points and is the core of the entire scale. Ten percent below the average earns two points; twenty-five percent or more maxes the factor out. A sale priced at or above the average earns zero here, and the score card states that in plain language rather than quietly omitting the line.

Depth below list price carries up to two points. It is capped low on purpose. This is the discount every other site leads with, and we count it the way one counts a rumour: recorded, and weighted lightly.

Proximity to the recent tracked low carries up to one point. Matching that low earns the full point; landing within five percent of it earns half. If the product was cheaper a fortnight ago, the score is required to admit as much.

Absolute dollars saved carries up to one point, because percentages flatter inexpensive things. A hundred and fifty dollars or more of real savings earns the full point, fifty dollars earns half, twenty dollars earns a quarter. A dramatic percentage off a phone case does not get to outrank a modest percentage off a laptop.

The resulting figure is clamped to a one-to-ten scale and labelled without generosity. Eight and a half or above is Exceptional. Seven is a Strong Deal. Five and a half is Solid. Four is Modest. Below that the label is Weak, rendered in grey, because we see no reason to shout in a colour we do not mean.

What Does Not Receive a Score

If a product has no ninety-day average, no thirty-day average, and no usable list price, our scoring function returns nothing whatsoever. Not a low score. Not an estimate with a caveat attached. Nothing.

Several staff members have observed that this is the most important rule on the site, and it reduces to four words: no reference, no verdict. An estimate dressed as a measurement is worse than an absence, because an absence is at least honest about itself.

The Gate That Rejects Our Own Work

Past the pricing logic sits a structural gate built on the assumption that we will eventually make a mistake, and designed to catch us doing it.

It refuses a listing whose stated list price is below the price we are advertising, that being the arithmetic signature of a fabricated discount. It refuses a product identifier that does not match the exact Amazon format. It refuses an Amazon link lacking our affiliate tag, on the grounds that a broken link is a broken promise regardless of which party it disadvantages. It refuses a description containing raw markup, or exceeding four hundred characters, or bearing the unmistakable residue of unedited machine output. It flags for human attention any price so far below list that a data error is the likelier explanation than a bargain.

Nothing that fails is discarded silently, and nothing that fails is published anyway. It is held as a paused draft for a person to examine.

That same gate is pointed at our writing. An article is rejected outright if the body runs short, if it carries fewer than two section headings, or if it contains stray heading levels or unedited artefacts. This article was submitted to it before publication. Had it failed, you would be looking at a blank space where it should be.

The Shelves We Vet for Something Other Than Price

Two sections of the site list products on merit rather than discount, which demands a different species of scepticism entirely.

For natural-fibre goods we read the full listing, meaning title, feature bullets, and description together, because that is precisely where listings contradict themselves. A single mention of a synthetic fibre anywhere in that text disqualifies the product. A declared composition percentage attached to a synthetic term disqualifies it. An unambiguous hundred-percent natural claim must be present, and the listing must carry enough body text to corroborate it. The product must also clear a reputation floor of four stars and one hundred reviews. All four conditions hold, or the answer is no.

We do not accept bamboo fabric at all, and the reasoning is instructive: retail bamboo textile is almost always viscose or rayon, which is to say a chemically regenerated fibre wearing a leaf costume.

For supplements the bar is raised deliberately, because the category is fake-heavy. Snake-oil vocabulary is disqualifying. An independent-testing signal is mandatory. Dosage must be disclosed, which means an undisclosed proprietary blend fails on that basis alone. The reputation floor rises to four point two stars and two hundred reviews. No middle verdict is available: the status is verified or rejected, and a product we are unable to positively verify is simply not posted.

What the Procedure Does Not Claim

It does not forecast. A ten out of ten describes where today's price sits relative to its own recorded history, and history is not a promise about next month.

It does not evaluate whether the product is good, or whether you want it. A flawlessly scored deal on something you do not need remains a purchase you did not need to make, and no quantity of arithmetic on our end alters that.

It measures one retailer's price history closely rather than every retailer's shallowly, which is a genuine limitation, and we prefer stating it to implying otherwise.

The staff finds the entire apparatus faintly absurd. This much machinery, arrayed against the question of whether a laptop is cheap this week. We built it anyway, and we intend to keep it.

Shop Current Deals

We track live prices against each product's 90-day Amazon average, so you can tell a real discount from a banner. Browse today's deals →

The deals page shows only what currently clears the ten-percent threshold. Anything that has since recovered has already removed itself, without being asked to. The complete scoring formula is published on our methodology page.

— The BuyGetRewards Editorial Staff