55 of 216 handoffs arrive intact.

What do the DTC handoffs that actually work have in common?

55 of the 216 rated ad and landing page pairs keep the ad's promise intact, a base rate of 25.5%, and I read all 55 handoff notes end to end to find what they share (The Handoff, 243 teardowns, August 2026). One trait is in every one. A second is in 43. Four more run between 5 and 24.

Every pair was captured by hand, the creative as it ran in feed and the page behind the click. 216 of the 243 published entries carry a rating: 55 Continuous, 110 Partial, 51 Break. This is the other half of the six ways the 161 failures fail. There is no trait field in the data, so no count here is a tally. Each came from reading the note.

Does a clean handoff always land on the exact thing the ad sold?

Almost. 43 of the 55 land on the advertised item's own page, or on a collection holding only the drop the ad announced, so nothing has to be searched for. The other 12 are mostly built campaign pages, 8 of them, plus three homepages and one shop grid from ads that asked for nothing.

BRUNT Workwear is the shape of it. The note opens: one boot, one name, one page. The ad's light grey boots are in every tile of the gallery, and the button reads GET IT BEFORE IT'S GONE at $164.99.

A collection qualifies only when it is not really a shop. Bandolier ran a toile phone case and landed where, in the note's words, the collection holds three products and all three are pieces the ad set up.

36 of the 43 are plain product page landings; the other 7 are collections I judged to hold only the collab, the call I would argue with first. It is also where the base rate bites: 115 of the 216 pairs land on a product page and 79, 68.7%, fail anyway.

What is the one thing all 55 clean handoffs do?

The money never gets worse at the click. In all 55 the page either raises no offer the ad did not raise, or pays out at least as much as the ad promised. Not one quotes a smaller discount than the ad did, and in 15 neither side raises money at all.

Everyday Dose is the version where the two numbers do not even match and it holds anyway. The ad buttons out on Lower Your Cortisol, 35% Off; the page bar says UP TO 62% OFF and the 60 serving option shows $49 struck from $132. The note: same product, same frame, and the deal is larger on arrival.

Graza's largest word is 35% Off and the page prints no percentage at all, which the note calls the ad arguing with itself rather than the page failing, since $74 to $49 is 33.8%.

The check against the failures is clean here, and it passes. The exact inverse, the offer on the page is not the offer in the ad, runs to 23 entries with no Continuous. Body Candy sells fifty percent off with one code, prints forty percent off with another, and breaks.

How much of a clean handoff is just recognition?

More of it than I expected. In 24 of the 55 the exact frame from the creative is the hero or first gallery tile. In 16 the colour or shade the ad showed is already selected on arrival. Neither costs anything: both assets exist before the ad ships.

Bluebella ran one studio frame of a model in a black mesh basque, and that is the first image in the page gallery, same pose, same crop, same garment. Fenty Beauty is preselection at full strength: PINK BLINK is not just stocked but preselected, its swatch ringed and the dropdown reading Pink Blink, hot pink. Animal Pak's flavour selector already reads Watermelon Lime, so nobody re-picks what the ad sold them.

Rows one and five are the only clean checks: the inverse of each, 23 offer mismatches and 13 wrong-variant landings, holds no Continuous at all.

What happens to the ad's proof number after the click?

In 17 of the 55 it is answered with a matching number. A review count, rating, percentage or seal named in the creative comes back as a real figure at or above the fold, instead of generic reassurance. In several it goes up on arrival.

Bare Necessities ran a black card: the name, four and a half stars, 1,000+ Reviews. The page resolves it at 4.5 based on 1,129 reviews with 85 percent recommending. AIRWAAV turned a testimonial into figures: 4.6 from 416 reviews, a Clinicians' Choice badge citing 339 clinicians, $119.98 struck to $89.99.

It need not be social proof. Manta Sleep's ad reads 15 Sounds, and the page carries an accordion reading 15 non-looping sleep sounds. Honoured, not restated.

This is the trait I most want to call a cause and cannot. The failure side has a mode that covers it, the page drops the argument that earned the click, 43 entries, but it does not separate proof from audience or occasion. The pattern breakdown itemises it.

Can an ad with no offer at all still hand off cleanly?

Yes, and it is the only case where a broad destination is honest rather than lazy. 5 of the 55 are ads making no product promise, no price and no call to action. That is below the reporting floor: too few to quote a percentage from.

All three clean homepage landings sit here. Chubbies ran boys sprinting into the water with no price and no product callout, and the homepage hero is the adult half of that joke under BIG LAKE ENERGY. The note: because the ad made no conversion ask, a homepage that leads with swim is the honest place for it to land rather than a dodge.

rhode ran a portrait with one word on it and no button. Its note is blunt: a homepage would normally be the weak destination, but here it is the only one the ad could honestly resolve to. Gymshark's note says what such a rating is worth: Continuous is a low bar when the ad promises this little.

The rule is narrow. Send bare-domain traffic only from creative naming no product, no price and no rival. Everything else that tried failed: the homepage carries 15 pairs, 3 Continuous and 8 breaks.

Which plausible explanations for a clean handoff do not hold?

Six of the seven I tested. The right product page, an offer in the creative, a deeper discount, video over static, a modal free fold and a friendlier category all fail to predict a clean handoff. Only the campaign objective predicts hard, and it predicts failure.

Row two is backwards from what everybody buys, and two mechanisms could produce it that the data cannot separate: an offer gives the page something specific to contradict, and the campaign that prints a discount is also the one that points at a sale grid.

The modal row needs care, because its inverse holds hard: three clean pairs name an overlay and delivered underneath it, but of the 32 entries gating a discount behind an email capture, 31 fail.

What should you copy from the 55 on Monday?

Five things, in the order the counts justify, starting with the only trait all 55 share. First the limit: 55 of 216 is a 25.5% base rate, so a trait becomes a cause only once it is rarer among the 161 failures, and for three of the six it cannot be shown.

Nobody recorded whether failing pages reused the ad's artwork, answered a proof number or asked for nothing, because a teardown documents what went wrong. Copy those three because they are cheap, not because they are proven.

1. Check the money on both sides before the ad goes live. No clean pair quotes a smaller number than the creative did, and all 23 pairs where the offer changed failed.
2. Point every creative that names one object at that object's own URL. 43 of the 55 do, and routing owns 24 of the 51 outright breaks. A collection only works if it holds nothing else.
3. Wire the advertised variant into the link. 16 of 55 arrive with the shade or colourway selected, and all 13 pairs that opened on the wrong one failed.
4. Make the ad's own frame the page's first image. 24 of 55 do. The asset exists and is approved, and it makes the click recognition rather than reading.
5. Answer the ad's number with a number. 17 of 55 do. Bare Necessities advertised 1,000+ Reviews and the page paid it back at 1,129.

One item is deliberately absent: make the page better. Not one of the 55 notes describes a design decision. They describe an object, a price, a photograph, a swatch and a number, each of which survived the click or did not. Every figure comes from the same 216 pairs, open to read with their captures attached, and the free store audit runs the read on yours.

Common questions

Five questions, answered with counts from the same 216 rated pairs, 55 of them Continuous.