Your ads are getting clicks. The store is not getting orders. Before anyone asks for another round of creative, open the ad and the page it sends people to, side by side.
Does the product match? Is the advertised offer easy to find? Can someone on a phone buy the thing without dismissing three pop-ups or discovering an unexpected shipping charge?
To tell whether your ads or landing page need work, measure each step separately: who sees the ad, who pays attention, who clicks through, who actually reaches the page, and who completes the next action. Strong ad engagement plus weak on-site conversion is a reason to investigate the destination. It is not, by itself, proof that the page is at fault.
A successful campaign needs the ad, audience, offer, website, and measurement to work together. The useful question is where the evidence tells you to intervene first.
What tells us an ad is doing its part?
Start with the job you hired the campaign to do. Video views can be a sensible awareness objective. A sales campaign needs evidence of qualified shopping activity and, eventually, profitable purchases. Cheap attention is not a substitute.
These are the signals we would bring into a performance review. Compare them with the same account’s history and similar audiences, placements, formats, and objectives. A cold prospect and a returning customer are not a fair head-to-head comparison.
| Signal | What to look for | What it does not establish |
|---|---|---|
| Audience and delivery | Relevant geography, placement mix, prospecting versus retargeting, and actual search terms for search campaigns | A targeting setting does not prove every impression reached a likely buyer |
| Hook rate | Whether the opening earns a brief view; explicitly name the view threshold and denominator | Stopping a scroll does not show purchase intent |
| Video retention | Whether viewers stay long enough to understand the product, offer, and next step | A watchable video can still attract the wrong shoppers |
| Outbound click-through rate | Clicks leaving the platform divided by impressions, where that metric is available | Clicks do not guarantee the destination loaded or the visitor wanted to buy |
| Cost per relevant click or landing-page view | Cost of getting a potential customer to the destination, alongside downstream behavior | A cheap visit is not necessarily a valuable visit |
| CPA, purchase value, and lead quality | Commercial outcomes using consistent attribution and a realistic margin target | Platform attribution alone does not prove incremental revenue |
Be especially careful with “CTR.” All-click CTR can include interactions that never take someone to your website. Use the click type that fits the question and keep it consistent across comparisons.
Hook rate also needs a definition. For a Meta video report, you might calculate three-second video plays divided by impressions. Label that exact calculation; do not present it as a universal platform standard. A TikTok report may use two-second views instead. TikTok’s video metric definitions distinguish ordinary views from focused views, which can include interactions. Those numbers are not interchangeable.
A strong opening followed by a sharp drop before the product appears suggests a different creative problem from a weak opening. And a sensational promise can lift CTR while making the eventual sale harder. The creative still owns the expectations it creates.
Follow the click before judging the landing page
First, check that people arrive. Compare outbound clicks with measured landing-page views and site sessions over the same period. A new gap warrants investigation: slow loading, redirects, broken URLs, accidental clicks, consent choices, or missing tracking can all contribute.
Do not treat the difference as an exact count of people who abandoned the page. Platforms and analytics use different counting rules. One person can click more than once, and some visits cannot be measured.
Then follow a consistent on-site funnel. For ecommerce, that might be landing session, product view, add to cart, checkout, purchase. For lead generation, use page visit, form start, completed submission, and qualified lead. Define whether each count represents events, sessions, or people; repeated events can make a funnel misleading.
- People arrive but rarely explore or add to cart: investigate message match, product clarity, price, trust, mobile usability, and visitor intent.
- People add to cart but do not finish: inspect shipping charges, delivery dates, account requirements, payment failures, and checkout usability.
- Forms start but do not submit: test validation errors, required fields, and the mobile keyboard experience.
- Submissions look healthy but sales says the leads are poor: revisit targeting, the promise, qualification, and follow-up. A shorter form is not automatically the fix.
Google’s GA4 engagement definition is broader than buying: an engaged session can qualify through time, a key event, or multiple page views. Use bounce rate and engagement as clues, not a pass/fail grade for the page.
When does the evidence point more strongly to the website?
A visible defect is a good place to start. If the advertised size is sold out, the discount fails, or the submit button is covered on mobile, fix that. You do not need a statistical debate to repair a broken purchase path.
For less obvious problems, look for patterns. Did conversion fall across several otherwise stable campaigns after a website release? Does one destination underperform other relevant pages for comparable traffic? Is the loss concentrated on a device where you can reproduce a usability problem?
Those findings strengthen a website hypothesis. They still need context. Branded search visitors may already know what they want; cold social visitors may be meeting the product for the first time. Comparing their raw conversion rates can make a perfectly usable page look broken.
Google advises keeping the ad and destination aligned, including the offer and requested action. On search campaigns, Quality Score components distinguish expected CTR, ad relevance, and landing-page experience. Treat those as diagnostic context, not proof that a campaign is profitable.
The same clicks can produce very different acquisition costs
Here is a hypothetical example, not a BPM client result. Two periods each spend $1,000, produce 50,000 impressions and 1,000 outbound clicks, and record 800 landing sessions. Outbound CTR stays at 2%; cost per outbound click stays at $1.
In the first period, 32 of those sessions include a purchase: a 4% purchasing-session rate and $31.25 in ad spend per purchasing session. In the second, only eight do: 1% and $125. Assume one order per purchasing session for this simplified example.
The ad’s click metrics look unchanged. The business result does not. Before ordering new ads, inspect what changed after arrival, but also check whether the audience, placement mix, or intent of those clicks changed. Equal click counts do not mean equal traffic quality.
The reverse matters too. If better-qualified creative attracts fewer clicks but more purchases, a lower CTR can be a good trade. Evaluate the campaign against what your margins can support, not whichever metric makes the report look best.
Use this worksheet to choose the next test
Put one row per campaign and landing-page combination in a shared document. Record the date range, objective, audience, device, offer, and attribution settings, then fill in:
- Impressions, the explicitly defined hook rate, retention, outbound clicks, and spend.
- Measured arrivals, relevant product or form actions, completed outcomes, and qualified leads or net order value.
- The first meaningful drop-off compared with an appropriate baseline.
- The suspected cause, an alternative explanation, and evidence that would contradict your preferred explanation.
- One proposed change, an owner, a budget, a measurement window, and a success metric.
Verify tracking against actual orders or received leads before interpreting an abrupt collapse. Allow for conversion delay. Twenty visits and no orders is not enough to declare a page broken; the amount of evidence needed depends on the underlying conversion rate and the size of the change you want to detect.
Where traffic allows, randomly split comparable visitors between the existing page and a focused revision while keeping the offer and ads stable. Judge completed outcomes and quality, not just button clicks. A simple before-and-after comparison is weaker because promotions, seasonality, and audience mix can move at the same time.
Agree on the fix, the owner, and the decision
The media team can diagnose a checkout problem without having permission to change the checkout. That makes ownership part of the campaign plan. Decide who can edit the page, who approves the offer, who checks tracking, and when the team will review the result.
If a material conversion problem cannot be fixed yet, discuss limiting spend rather than buying more traffic into it. Creative tests can continue where they answer a useful question, but they should not conceal an unresolved site issue.
If performance has faded on a previously successful ad, our creative-fatigue diagnostic helps narrow that question. For the broader campaign, BPM’s paid social team and Shopify website work address connected parts of the customer experience. Bring the ad, the destination, and the numbers to the same conversation. That is how you get to a useful next step.
