How do I know if my ad creative is the problem or my landing page is?

Split the funnel into four segments and read them separately: click-through rate, landing-page-to-add-to-cart rate, add-to-cart-to-order rate, and repeat purchase rate. Click-through is the only segment nothing on your site can touch, so it isolates the ad side: creative plus targeting. A healthy click-through means the ad earned attention from the audience it was shown to. The next segment, landing page to add-to-cart, is shared between the two: an ad that over-promises shows up as a drop there even when the page itself is fine. A fall further down, at checkout or after delivery, points at the site, the price or the product rather than the video.

What does a low add-to-cart to order rate mean?

It means people wanted the product and did not finish buying it. The usual causes: shipping cost or delivery date appearing for the first time at checkout; a payment method your buyers expect is missing; checkout forcing account creation; a tax and duties line landing as a surprise; or the item being sellable but delayed (pre-order or back-order), so the ship date revealed at checkout kills an order that add-to-cart happily accepted. Add-to-cart is a strong intent signal, so a collapse at this step is almost always a site or operations problem rather than an advertising one.

Why is my ROAS low when my CTR is good?

Because return on ad spend is a blended number covering four different jobs: earning the click, holding the visitor on the page, converting the cart into an order, and getting a second purchase. A strong click-through rate with weak ROAS means the ad is doing its job and something after the click is not. Segment the funnel to see which step drops, and check for the traps that flatter or punish blended numbers: attribution lag on recent days, and mixing product lines with different margins into one average.

What is a good add-to-cart rate for ecommerce ads?

There is no single number, and using a published index built on a different kind of business is the most common way to misread your own account. Benchmarks for apparel are usually dominated by low average-order-value fast fashion, where browsing behaviour, price sensitivity and cart abandonment look nothing like a higher-ticket brand selling at several hundred dollars per order. Build your benchmark from your own account history within your average-order-value band, then compare product lines against that.

Your Creative Didn't Fail, Your Landing Page Did: Read the Funnel in Four Segments

A product line prints a weak blended ROAS. The room reaches the same conclusion within about ninety seconds: the creative isn't working, let's remake it. Another production cycle gets booked, another two weeks disappear, and the new videos land on exactly the same number.

That decision is often wrong, and it is the most expensive possible way to be wrong. Remaking creative is the slowest and costliest response available to you, so it should be the last thing you reach for, not the first. Before you spend anything, split the funnel into four segments and read them one at a time.

The four segments and the question each one answers

This is the diagnostic half of a pair, so it is worth naming the division of labour up front. ROAS is lying to you answers whether the money worked at all, by putting a break-even cost per acquisition next to the blended return. This post starts one step later, when you already know the number is bad, and answers a different question: which of four things broke.

A blended return number collapses four separate jobs into one figure. Pull them apart and each segment answers a question the blend cannot:

Click-through rate. Did the ad earn attention and the click? This is the only segment nothing on your site can touch, so it isolates the ad side: the creative and the audience it was pointed at.
Landing page to add-to-cart. Did the page hold the person the ad delivered? This is the handoff, and it is the one shared segment: an ad that over-promises shows up as a drop here even when the page itself is fine. So read it as a question about the gap between what the ad claimed and what the page shows, not as a verdict on either one alone.
Add-to-cart to order. Did checkout let a buyer finish? Add-to-cart is a strong intent signal, so this segment is about mechanics, not persuasion.
Repeat purchase. Did the product and the delivery hold up after the money changed hands? This one decides whether your acceptable acquisition cost is actually acceptable.

The principle underneath: a blended number tells you that something is wrong; only a segmented funnel tells you where. Every hour you spend arguing about a blended figure is an hour spent not knowing which of four different businesses is broken.

The misdiagnosis that eats a production cycle

Here is the shape that catches teams, described in directions rather than invented numbers, because the directions are the whole point.

Two product lines, both with weak blended ROAS. Segmented, they look nothing alike.

Line A: click-through below account average, landing-page-to-add-to-cart below average, checkout normal. That is a genuine creative and targeting problem. The ad is pulling the wrong people, or it is not pulling anyone.

Line B: click-through normal, landing-page-to-add-to-cart at or above account average, add-to-cart-to-order far below. Same weak blended number, opposite diagnosis. The creative brought qualified traffic. The page did its job well enough that more visitors than usual added the item. Then the purchase failed to complete.

One cause practitioners run into repeatedly for exactly this shape: the item was sellable but delayed, in a pre-order or back-order state. Add-to-cart still fired normally, so intent kept registering; the ship date, disclosed at cart or checkout, is what killed the order. (The uglier variant: inventory running out between add-to-cart and checkout.) Paid traffic was being sent, at full cost, to people who wanted the product and would not accept the wait. Nothing in the blended ROAS says that. The four-segment read says it in one glance, and it costs nothing to check.

Note what the two lines have in common. They would both have been "fixed" by the same reflex, and in Line B's case that reflex burns a production cycle on a checkout bug. Creative takes the blame for the site's failure constantly, because creative is the visible part of the machine and the only part everyone in the room feels qualified to have an opinion about.

The diagnostic table

One scoping note first: the four segments hold on any funnel, but the levers differ. On your own site, every row below is yours to fix. Selling through TikTok Shop's in-app checkout, the checkout mechanics are not yours to touch, so your levers concentrate in the creative, the product anchor, the price, and the product-page fields.

Read the segments in order. The first one that is meaningfully below your own baseline is where you work, and you stop there until it is fixed. Two calibration rules so "below baseline" means something: a gap worth acting on is large, a third or more off your own trailing baseline rather than a few percent; and a segment is only readable once it has real volume under it (dozens of clicks or carts, not a handful). Below that, wait rather than diagnose.

Segment that is off	What it usually means	Check first
Click-through rate	Hook or targeting mismatch	The first three seconds; whether the opening frame states a reason to keep watching; whether the audience is the one the product is for
Landing page to add-to-cart	Page, price, or a mismatch with what the ad promised	Does the page repeat the ad's exact claim above the fold; mobile layout and load time; price shock versus the expectation the ad set; whether the ad sends people to the product page instead of a collection or homepage
Add-to-cart to order	Checkout mechanics or inventory state	Ship-date and stock state (a sellable pre-order kills orders at the ship-date reveal); shipping cost or delivery date appearing for the first time at checkout; missing payment methods; forced account creation; duties and tax surprises
Repeat purchase	Product or fulfilment	Delivery time versus what was promised; the unboxing and first-use experience; quality against the expectation the ad created; whether any post-purchase flow exists at all

Run the table per product line, not per account. An account-level funnel averages a healthy line and a broken one into a mediocre middle, which is the same mistake as the blended ROAS one level up.

Benchmark against your own average-order-value band

Segmenting only helps if "below average" means something real, and this is where a second silent error creeps in.

Published category indexes are usually built from whoever contributes the most data, and in apparel that means low average-order-value fast fashion. Browsing patterns, cart abandonment, price sensitivity and repeat cadence in that band look nothing like a higher-ticket business, several hundred dollars per order, where a buyer takes days to decide, compares more, abandons more carts, and returns less often but spends far more when they do.

Grading a high average-order-value account against a fast-fashion index is self-deception in both directions. You will condemn a healthy add-to-cart rate as broken, or excuse a real problem because the index says you are fine.

The working discipline: before quoting any benchmark, confirm what kind of business it was measured on. In practice the most useful benchmark is your own account, over a long enough window, segmented by product line, within one average-order-value band. Your own history is the only source that is guaranteed to be measuring your business.

Two more traps that distort any segment read:

Attribution lag. Ad platforms that credit an order back to the day the ad was seen will always under-report the most recent day or two. Any decision made on single-day data is a decision made on a number that has not finished arriving. Read multi-day windows, always.
Margin blindness. The same ROAS is profit on a high-margin line and a loss on a low-margin one, which is why the segment read pairs with a per-line break-even cost per acquisition. That is the subject of the companion piece, ROAS is lying to you.

What to do before you touch the creative

An order of operations that saves most of the wasted production cycles:

Pull the four segments per product line, over a window long enough to survive attribution lag.
Find the first segment below your own baseline. Work only on that one.
If the break is after the click, fix the site. Stock state, shipping reveal, checkout friction, page-to-ad promise. These are same-day fixes with no production cost.
Re-read the funnel after the fix before drawing any conclusion about the creative. You cannot evaluate a video that was pointed at a broken checkout.
Change one variable at a time. Swapping the character and the format in the same test means you learn nothing when the result moves.

When it really is the creative

Sometimes it is, and the four-segment read tells you that too: click-through is genuinely below your own baseline while everything after the click is normal. That is a clean verdict, and now you know what you are buying with the production budget.

That is also the point at which volume matters more than polish. You are not looking for one perfect video, you are looking for the hook that earns the click, and that takes several distinct attempts rather than one expensive one. Winners also decay on a schedule, which is a separate problem from this one and is covered in TikTok creative fatigue.

Riffkit exists for that step: you start from a short video whose structure already earned attention, and it rebuilds the formula (the hook, the pacing, the emotional beats) around your product, with new footage and no filming. A verdict of "the creative is the problem" stops meaning a two-week production cycle and starts meaning an afternoon of variations, which is also what makes step 4 above affordable to repeat.

The short version

A blended number proves something is wrong. Four segments prove where. Click-through is the only segment your site cannot touch, so it isolates the ad side; landing page to add-to-cart is shared, because an over-promising ad breaks there too; and everything past the cart judges your checkout and your product. Benchmark inside your own average-order-value band or you are grading yourself against a different business. And check the stock status before you book the shoot.