The ad makes the argument.
The page makes the offer.

Who actually makes the promotion, the ad or the page?

Across 243 published teardowns, 65 entries meet a percentage discount that exists only on the landing page, against 26 whose creative printed one (The Handoff, August 2026). Two and a half times as many shoppers are surprised by a discount at the click as are told about one before it. Promotion, in consumer advertising as it actually runs, is something the destination does.

Every pair was captured first-hand: the creative as it ran in feed, and the page it clicked through to, opened signed out from a US IP and read on first view before anything was dismissed. A discount you can only see after closing a modal is not the same object as one printed in a banner. There is no offer field in the underlying data, so offer types were mined from four prose fields, hook, summary, handoff.note and read, then read by hand. Hook is verbatim ad copy, so it is ad side without argument. For the rest the sentence decides: the creative reads is ad side, the page opens with is page side.

216 of the 243 are rated pairs; the other 27 carry no rating and sit outside every rate below. Continuous, Partial and Break mean what they mean where the boundary is written down, and the six ways a handoff fails is the companion piece. Any rate resting on fewer than eight rated entries is reported as too few.

What do consumer ads actually offer in the creative?

49 of the 243 ads carry any offer at all inside the frame. 194 carry none. Inside those 49 the ranked list is short and top heavy: a percentage discount at 26 entries, a named sale event at 12, a free gift at 11, a dollar saving or struck price at 11, a bundle or kit at 10, then a tail that never reaches double figures.

An entry can carry several types, so the column sums past 49. Grüns runs the loudest version of the commonest one: "61% off + 4 free gifts on Grüns right now, and here's why everyone's grabbing it." Burrow prints "WAREHOUSE SALE above UP TO 70% OFF", which is a discount and no product.

Two types have a count of one. Free shipping is printed inside an ad frame exactly once, in a Grüns creative closing "GET UP TO 50% OFF + FREE SHIPPING" across the foot. A prize draw appears once, Ridge's "6th Annual Summer Sweepstakes". Free shipping is ordinary furniture on the destinations here and effectively absent from the creative.

How deep are the discounts brands print in ads?

25 of the 26 percentage ads print a fixed number, and the median is 35%. The range runs from 10% to 70%. 11 of the 25 print 40% or more, which makes the deep end the largest band in the distribution rather than the outlier most people assume it is.

The floor is Oura at 10%, gated on ID.me verification for students, teachers, health workers and first responders, which is a discount doing identity work rather than price work. The ceiling is Burrow at up to 70%, a warehouse sale headline with no product in the frame.

Cash figures are rarer and more specific. OSEA prints "SAVE $92", Birddogs "Get $20 Off Your First Order." Anchor pairs are the other cash form: Laura Geller at $178 struck to $60, Graza at $74 struck to $49, David Protein turning $3.25 into $2.60 per bar. Savings and anchor pairs are kept apart, because folding one into the other invents a discount the ad never stated as one.

Read the distribution as a ceiling. Where an ad printed up to N percent, the N was taken, and one ad is excluded: Carpe ran a discount of one percent per landed BMX trick, which has no fixed figure and no honest place in a median.

What promotion does the landing page introduce on its own?

65 entries, 26.7% of the published library, arrive at a page carrying a percentage discount the ad never mentioned. 52 of the 65 are rated Partial or Break, an 80.0% failure rate, the same rate carried by the ads that printed a discount themselves. The surprise is not free.

The shape is consistent. Avocado Green Mattress meets the click with "SAVE UP TO 20%", plus a Labor Day Mattress Sale bar and a Get $100 Off. At Dagne Dover, "Three discounts the ad never mentioned crowd the $295 price". Bums & Roses opens an "end of summer sale at up to 40% off for 48 hours" the creative said nothing about.

None of this is incompetence. The banner is owned by merchandising and runs on the promotional calendar; the ad is owned by paid social and runs on the campaign calendar. The visitor is the only person who sees both artefacts inside four seconds of each other.

That gap has its own failure mode: the offer on the page is not the offer in the ad, 23 pairs, 8 of them outright breaks. Aroma Retail is the blunt version, where "the advertised 30% is nowhere on screen and the only number that is says something smaller".

Why do email gates and spin-to-win fail so completely?

The two commonest things a destination does with a promotion are the two worst outcomes in the library. A discount held behind an email capture appears in 32 entries and 31 are Partial or Break. A discount turned into a game appears in 12 and all 12 fail, with no Continuous rating anywhere in the group.

The gate is always the same object in the same place. At Cuts, an "Anniversary Sale Up To 50% Off email capture covers the fold". At Bescher, "$20 off your first order covers the left half of the fold". At Fellow, a "Get 10% off" panel "opens over the product title, the price" and the buy button. The first decision a paid visitor makes is whether to hand over an address to learn the price of the thing they already wanted.

The games add a step to a decision already made. Jambys drops a "scratch card modal" that "dims the entire fold". Beam Organics opens on "Try your luck, scratch below to see what you win". Grüns lands its apology creative on a nine-card flip game, so the conceit, the limited edition flavour and the product all wait behind a round of pick a card. Twelve attempts, zero Continuous.

The quiet member of the group is a subscription preselected in the buy box, 8 entries and 7 failures: biom opens on "AutoShip and save 19%" at $45.22, Plant People with "Buy Once demoted underneath". The discount is real; the commitment behind it is what the ad never raised.

Does putting the offer in the ad improve the handoff?

It does the opposite. 38 of the 46 rated ads that carry an offer in the creative are Partial or Break, 82.6%, against 123 of 170, 72.4%, for the ads that carry none. Ten points in the wrong direction, on the tactic most brands reach for when paid performance slips.

This is a correlation on 216 pairs, not a controlled test, and the obvious confound is that brands discount hardest when they are already struggling. But the mechanism is visible in the entries. An ad without an offer makes one promise. An ad with an offer makes two, and the second is owned by a different calendar than the page.

Depth does not rescue it. 20 of the 25 rated percentage ads fail even though 11 advertise 40% or more. Hiya prints the code SUMMERSUN and "the page pays out only half of that: $17.50 with $35 struck through". Bonobos promises 20% off a first order and it "arrives on the page as a quantity promo". A bigger number on a page that contradicts it only makes the contradiction louder.

The two ad side types that do best are the two most specific: a dollar saving or struck price fails 6 of 11, a bundle or kit 6 of 10. Both name an exact object at an exact price, a promise a product page keeps without anyone coordinating anything. The worst is a free gift, 10 of 11: the easiest thing to drop from a page and the hardest to notice missing.

What should you change first?

Fix in the order the money leaves, and start with the cheapest change nobody makes. 43 of the 44 entries that gate or gamify a discount on arrival are rated Partial or Break. That is a targeting rule in a popup tool, not a redesign.

1. Suppress welcome modals, scratch cards and spin wheels for paid arrivals. One Continuous rating across 44 gated or gamified entries. A visitor who clicked the ad has converted once; charging them an email address to see the price is charging twice.
2. Decide which side of the click carries the offer, then hold it. 65 entries surprise the visitor with a page-only discount. Printing it in the creative is fine. Running no discount is fine. Discovering it one beat after the decision is what costs the rating.
3. Audit every live creative that prints a number or a code against what the destination shows today. 23 pairs run an offer on the page that is not the offer in the ad, 8 of them breaks.
4. If the offer is a bundle or a struck price, land on the exact SKU that carries it. The two best performing ad side offers in the set, 6 of 11 and 6 of 10.
5. Do not preselect a subscription the ad never raised. 7 of 8 entries whose buy box opens on a recurring plan fail.

One thing that is not on the list: going deeper. The median advertised discount here is already 35% and the ads carrying one fail more often than the ads carrying none. What a deeper cut has to earn back is a separate and unforgiving calculation. The same read on your own funnel is what the free store audit does.

Common questions

Five questions the data answers directly, from the same 243 published teardowns and the 216 rated pairs inside them.