Shampoo: price only works at the top
I measured 6,713 price changes across 465 shampoos on amazon.de and compared each one against the products that held their price. Both directions move rank far less than the price does: a 10 percent discount bought 3 to 6 percent of rank, a 10 percent increase cost 2 to 5. The average hides the real finding — premium shampoos react clearly, the mid segment barely at all. Together with hair oils, this is the weak end of the series.
In short
- A 10 percent discount improved rank by 3 to 6 percent — less than half of what the cut gives away.
- A 10 percent increase cost 2 to 5 percent of rank. The market punishes increases no harder than it rewards cuts.
- Discounts only clearly work in premium (20–41 €): premium responds at 0.68, the mid segment at 0.16, and the two ranges do not overlap.
- Raising a budget price showed no measurable rank cost within the price moves I observed (5–13 €).
- Almost every move sees competitor movement, almost none is an answer: 98–99 percent background activity, but only 5 to 9 percent above each rival's own repricing rhythm.
Price cuts
A 10% cut typically improves the sales rank by 3–6%.
Price increases
A 10% increase typically costs 2–5% of rank.
Strong at the top, dead in the middle
Each dot is the measured rank response of one price segment; the whiskers show the 90 percent confidence band. A value of 1.0 would mean the rank moves one for one with the price. Everything in this category sits well below that line.
The segments split. Premium shampoos (20–41 €) respond to cuts at 0.68 [0.53–0.90]. Mid-priced shampoos (13–20 €) respond at 0.16 [0.03–0.50]. Those two bands do not overlap, so the gap is a finding, not noise. In the mid segment, a 10 percent discount buys roughly 1.5 percent of rank — paid margin with almost nothing back.
Budget behaves differently again. Cuts work a little (0.35 [0.19–0.53]), while price increases showed no measurable rank cost (0.14 [−0.01–0.32], not demonstrable). Within the price ranges I actually observed, raising a budget price did not visibly hurt rank.
The full picture
01 · 1 · Does discount depth matter?
Deeper does more, but each point buys less.
Rank after price changes of different sizes
The discount side climbs slowly and stays shallow: 12 percent better rank for a small cut (5–15 percent), 14 percent for a moderate one, 20 percent for a quarter off or more. The direction is consistent — deeper tends to do more — but neighbouring depth classes overlap, so no single step is demonstrable in this data.
What clearly falls is efficiency. Tripling the depth from around 10 percent to around 30 percent moved the average factor from 0.88 to 0.80 — each extra discount point bought less rank than the point before it.
The increase side has a ceiling, and it mirrors the pattern from hair oils: a rise of 15 to 25 percent cost 15 percent of rank, a rise above 25 percent cost 14 percent. Beyond roughly 15 percent, going further costs nothing extra in visibility.
02 · 2 · Will competitors follow?
Everything moves, almost nothing answers.
Time until the first competitor answers a price cut
Shampoo reprices constantly. Within 14 days of any price move, 98 to 99 percent of moves saw some other price change in the category. That number sounds dramatic and means little: with this much background activity, something always moves.
So I compare against chance. Each competitor's expected activity is its own repricing rate from the 365 days before the trigger. Against that baseline, cuts were answered at 1.05 times chance and increases at 1.09 times — barely above background, and even closer to chance than in hair oils.
When answers come, they come fast: the median first reaction lands on day 1, and about 61 percent of moves see a same-day change elsewhere — as automated as eye creams, and the opposite of hair oils, where fewer than half landed on the same day. Same-day moves are excluded as ambiguous because I cannot tell who moved first, so the reaction figures are a conservative lower bound.
03 · 3 · Where could a new shampoo enter?
One band clears the bar: €7.84–€9.43.
Competition and demand by price range
93 of the 410 judgeable products entered the category within the last 24 months. 12 of them are winning, 9 already exited.
One band clears the demand baseline convincingly: €7.84–€9.43. It holds 14 products with a median of 366 reviews, 43 percent of them rank in the category's top third, and all three recent entrants there are winning or building. Prices in this band hold for a median of 83 days, so a newcomer should plan for regular repricing.
The rest of the shelf is either busy or defended. The most crowded band (€13.66–€16.43, 41 products) stays below the demand baseline, and the €19.77–€23.79 band carries a median of 1,320 reviews — that ground belongs to established brands. The cheapest bands hold too few products to judge at all.
04 · 4 · What the market looks like underneath
A mature shelf that reprices around the clock.
465 products from 321 brands made the curated set: single bottles between 200 and 400 ml, no multipacks, no sets, no men's or kids' lines. The median price is €16.73.
It is a mature, price-stable market with heavy repricing traffic and high review walls in the mid and upper bands — median reviews per price band run from 320 to about 1,320. Of the 6,713 price changes found, 1,071 overlapped with another change from the same product, the highest share in the series so far: this shelf never sits still.
A market where prices move constantly gives a single discount little room to stand out. That is consistent with the weak elasticity I measured, though the measurement itself does not test this mechanism.
Data appendix
465 products on amazon.de, taken from the Shampoos listing and then narrowed: single bottles between 200 and 400 ml — the volume band alone removed 949 products where bottle size would have masqueraded as price positioning — no multipacks or gift sets, no men's or kids' lines, no solid shampoos, no treatments or accessories, and a brand cap so that a single catalogue cannot dominate the control groups. Category membership is taken from Amazon's own listing rather than from keywords in the title. Products that Amazon does not place in this branch are flagged automatically and excluded from all figures: 55 such products carried 0 of their events into the fit set and appeared 0 times among 215,984 control observations. 13 further products carried no category data from the source and contributed zero events. The product list is available on request for methodological review.
| Step | Events | Removed because |
|---|---|---|
| Price changes found | 6,713 | at least 5% and €0.30, held for 7 days |
| Gaps in the price history | −136 | the change cannot be located precisely |
| Too little ranking data | −416 | less than 70% of days covered |
| Overlapping price change | −1,071 | another change from the same product in the window |
| Deal or coupon markers | −0 | none recorded in this set |
| Stockout signals | −250 | rank moves for availability, not price |
| Burst of new reviews | −344 | more than 10% and at least 15 new reviews |
| Listing too young | −144 | less than 60 days of history |
| Too few control products | −141 | fewer than five comparable products held their price |
| Passed every check | 4,729 | an event can fail several checks, so the removals sum to more than the difference |
| Used for the figures | 4,282 | from 200 products, after removing 373 with a pre-existing trend and 79 from other category branches (5 overlap) |
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
Observed price changes: −99% to +6,004%. The extreme upper end comes from listings returning to a nominal price after being unavailable; the figures above are read only from the range where changes are dense.
Limits
- 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. Competitor reaction and advertising interplay are not modelled.
- 465 products
- 6,713 price changes found
- 4,729 passed every check
- 4,282 used for the figures
- 200 products in the fit set
- ref 562eccb2
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.