Face serums: the more expensive the product, the better discounts work
I measured 3,669 price changes across 421 face serums and compared each one against the products that held their price. Overall, price and rank move almost one for one in both directions. Split by price level, the picture inverts what most people assume: in the cheapest third of the category, rank barely responds to a discount. In the most expensive third, it responds more than the discount itself.
In short
- Premium serums (€28–90): a 10 percent discount improved rank by about 14 percent. More than the discount itself.
- Budget serums (€5–17): the same discount improved rank by about 6 percent — less than it costs in margin.
- The two ranges do not overlap. This is a demonstrated difference between the segments, not a pattern I am guessing at.
- Discounts keep working the deeper they go. There is no ceiling in this category: a cut of 25 percent or more improved rank by 31 percent.
- Competitors answer within a day. Three in four answer on the same day.
Price cuts
A 10% cut improved rank by roughly 7 to 11 percent.
Price increases
A 10% increase cost roughly 7 to 10 percent of rank.
The same discount, three different outcomes
For price cuts the reaction rises steadily with the price level: 0.58 in the budget group, 0.94 in the middle, 1.42 at the premium end. The budget and premium ranges do not touch, which makes this the first category in this series where a difference between price segments is demonstrated rather than merely visible.
Read as a decision: in the budget third, a discount buys less visibility than it gives away in margin. At the premium end the same discount buys more. The cheap end of this category is where price competition works least — which is the opposite of how discount budgets are usually allocated.
Increases behave more evenly, between 0.67 and 1.08 across the three groups, with no demonstrated difference between them. Raising a price costs roughly what it earns, wherever the product sits.
The full picture
01 · 1 · Does the size of the discount matter?
Yes, and unusually: deeper keeps paying.
Rank after price changes of different sizes
All six sizes produced a demonstrable effect, which is rare — in most categories the small and the very deep changes disappear into noise. Here the curve simply keeps going: 16 percent better rank for a small discount, 22 percent for a moderate one, 31 percent for a quarter off or more. On the increase side the cost grows the same way, from 17 percent worse for a small rise to 37 percent for a large one.
What this rules out is a discount ceiling. In other categories a cut beyond roughly a quarter buys nothing extra. In serums it still does — which does not make deep discounting profitable, only effective. Whether it pays depends on margin, and that is a calculation per product, not per category.
02 · 2 · Will competitors follow?
Within a day, almost always.
Time until the first competitor answers a price cut
Across 6,246 price changes from 495 products, three quarters of all moves were answered on the same day and 70 percent of the remaining answers arrived the next. This is a category run by automated repricing.
The practical consequence: a price advantage in serums lasts hours, not weeks. Discounts here are campaigns with an end date. And increases are slightly more likely to be followed than cuts, so raising a price is a less lonely move than it feels.
03 · 3 · Where could a new product enter?
€11.19 to €14.11 — the only price range with proven demand and a low barrier.
Competition and demand by price range
Only two of twelve price ranges clear that baseline, and the corridor at €11.19 to €14.11 is the stronger of them: 33 products, of which almost half rank in the top third, a typical review count of 77 against 229 one range above, and 9 of 16 recent entrants winning or building. Prices there hold for about six weeks, so a newcomer should budget for regular repricing.
Two honest qualifications. First, part of the demand evidence in that range rests on one established product carrying 10,037 reviews against a band median of 77 — its success comes from brand equity, not from the price position. Second, that corridor sits in the budget tier, which this report shows to be the least price-responsive part of the category. Entering there means competing on product and visibility, and holding price rather than discounting into the market.
Worth knowing about this category as a whole: of 697 priced products, 436 launched within the last two years, 91 of them have already reached the top third and only 10 disappeared. Serums take newcomers.
04 · 4 · What the market looks like underneath
Restless, and full of products that are not really serums.
Of 7,100 price changes found, 1,257 had another change from the same product running in the same window — nearly one in five moves happens inside its own turbulence. Another 647 coincided with a burst of new reviews and 327 with a stockout; both move rank on their own, so they are excluded. 313 events were dropped because the product was already sliding before the price came down, which is a seller reacting to a problem rather than causing an effect.
Two more numbers describe the category. The observed price changes run from 76 percent off to more than tripling, so this is not a market with settled prices. And 164 of the 861 products listed under serums do not sit in that branch of Amazon's own category tree — 854 of their events are excluded from every figure here.
Data appendix
861 products on amazon.de, taken from the category listing and then narrowed: no multipacks or gift sets, no ampoules, no eye or body products, no face oils, no creams or masks that were listed under serums, nothing below €6 or above €90, container sizes between 15 and 75 ml where stated, and at most eight listings per brand so that a single seller's catalogue cannot dominate the control groups. Products that Amazon does not place in the serum branch are flagged automatically and excluded from all figures. The product list is available on request for methodological review.
| Step | Events | Removed because |
|---|---|---|
| Price changes found | 7,100 | at least 5% and €0.30, held for 7 days |
| Gaps in the price history | −224 | the change cannot be located precisely |
| Too little ranking data | −442 | less than 70% of days covered |
| Overlapping price change | −1,257 | another change from the same product in the window |
| Deal or coupon markers | −0 | none recorded in this set |
| Stockout signals | −327 | rank moves for availability, not price |
| Burst of new reviews | −647 | more than 10% and at least 15 new reviews |
| Listing too young | −373 | less than 60 days of history |
| Too few control products | −93 | fewer than five comparable products held their price |
| Passed every check | 4,563 | an event can fail several checks, so the removals sum to more than the difference |
| Used for the figures | 3,669 | from 421 products, after removing 313 with a pre-existing trend and 854 from other category branches |
price change ≥ 5% and ≥ €0.30, held ≥ 7 days · comparison: 14 days before, days 3–14 after · control products: same category, price moved less than 2%, at least 5 of them · rank coverage ≥ 70% of days · listing older than 60 days · review burst = more than 10% and at least 15 new reviews · ranges are 90%, from a bootstrap clustered by product
Review history was available for every product in this set, so the review-burst check applied to all events. Observed price changes: −75.7% to +343.7%.
Limits — I measure the ranking, not units sold, in the 14 days after a price change. Results hold only for the price range actually observed. Every figure is for the category as a whole or for one of its three price groups, never for an individual product. Where a range touches the level at which nothing happens, I say so instead of quoting a number.
- 861 products
- 7,100 price changes found
- 4,563 passed every check
- 3,669 used for the figures
- 421 products in the fit set
- ref 379aa815
The Monthly
Each weekly report measures a single category. The Monthly is where they get compared, and it carries the quarterly report when one is published.