You pay for clicks and see traffic in the reports, but orders are slow to come. Your first reaction may be to change the ads or increase the budget. Before you do, find out where people drop off and whether the reports describe correctly what is happening in your store.
Why do you get clicks but no sales? You may be attracting visitors with no intention to buy, you may have an unclear offer, or there may be an obstacle in checkout. Sometimes the orders exist but are measured wrongly; other times the period analysed is too short. Check measurement and whether purchasing works first, then traffic, the offer and returning visitors. Increase the budget when the data and the margin support the decision, after fixing the blockages you identified.
The practical priority: check a test order, compare the store's orders with the reports, and look for errors on mobile and at payment. Only then analyse the traffic segments and test the offer. An error that prevents payment takes priority over changing the colour of a button.
The path from ad to order
Look at buying as a sequence of steps:
impression click real visit product view add to cart checkout started order confirmed.
An impression means the ad was shown. A click indicates an interaction, but does not prove the page loaded. A session is a group of interactions measured on the site. The same person can have several sessions and can click the ad several times.
That is why clicks and sessions need not be identical. Closing the page before it loads, how measurement is implemented, or several clicks within the same visit can produce differences. Google explains these situations in the documentation on clicks and sessions. Measured sessions do not necessarily represent all real visits either.
In Google Analytics 4, abbreviated GA4, actions are recorded as events. For online retail, the usual names include view_item, add_to_cart, begin_checkout and purchase. These must be implemented and verified; simply installing the tool does not confirm that purchases are fully measured. See the ecommerce events recommended by Google.
A confirmed order does not automatically mean money received either. Payment can fail, a parcel can be refused and a product can be returned. Keep recorded orders, paid orders and revenue remaining after cancellations and returns separate.
1. Are you measuring orders and conversions correctly?
The symptom and possible explanations
The store has orders, but the campaigns report zero purchases. Or the ads report more purchases than the orders that exist. An event may be missing, it may be sent twice, or there may be attribution differences: the rules by which a tool gives an ad credit for a purchase.
Don't immediately assume a platform is wrong. Two platforms can claim the same order, and the data can use different periods, time zones or definitions.
What to check and how to interpret it
Pick a completed period and export the order ID, date, value and status from the store. Compare it with the purchases in GA4 and the conversion actions in the ad accounts. Check the currency and the treatment of shipping, taxes and discounts before you compare revenue.
Then place a test order, using a safe method offered by the store. Note the ID and watch whether a single purchase appears, with the correct value. In GA4, DebugView helps verify events and parameters. Check the standard reporting separately after the data is processed. A direct test order validates collection, but it does not prove that a sale is attributed to an ad.
Reload the confirmation page and check whether the implementation resends the purchase without justification. For web flows, GA4 uses the transaction ID to deduplicate purchases. It must be unique for each order, not empty or reused, according to the transaction ID documentation.
When the differences are concentrated on a particular day or payment method, check the timing of technical changes. If they appear in every period, review the definitions and the coverage of the measurement. Ask for a reconciliation table with the observed difference, the possible explanation, the evidence available and the person who continues the check. That lets you track the investigation without making random changes.
What to do next
If the order exists in the store but the event is missing in the test, investigate the implementation together with the developer. If the test works, compare attribution, time windows and collection conditions before declaring an error.
Also test how measurement behaves after consent options are accepted or refused, without forcing collection. Don't ask for perfect equality between sources with different coverage.
A mistake to avoid: adding up the purchases reported by Google Ads and Meta and presenting them as unique orders. The basis for commercial reconciliation remains the store's order records.
2. Do you have enough data to draw a conclusion?
The symptom and possible explanations
You have a few dozen visits without an order, or results swing sharply from one day to the next. The explanation can be the wrong traffic, but also a low volume of observations or a product with a long decision time.
The absence of orders in a small sample doesn't prove on its own that the ad cannot work. Still, uncertainty doesn't justify unlimited spending.
What to check and how to interpret it
Analyse the relevant sessions, not just the number of days since launch. Separate paid traffic from other sources and compare periods with similar days of the week, promotions and availability. Check the history of the interval between interacting with the ad and buying. Google describes the impact of conversion delay on evaluating campaigns.
Hypothetical example — conversion rate and acquisition cost. In a completed 30-day period, a paid segment has 1,000 measured sessions and 20 confirmed orders associated with those sessions. Each order belongs to a distinct session. The segment's advertising spend is 1,500 RON.
- Session conversion rate = sessions with at least one order / total sessions × 100 = 20 / 1,000 × 100 = 2%.
- Advertising cost per order = advertising spend / orders = 1,500 / 20 = 75 RON.
- If the 20 orders come from 15 new customers and five existing customers, the media cost per new customer = 1,500 / 15 = 100 RON.
The last calculation allocates the segment's entire spend to new customers. It is not the full acquisition cost: for that you must define and include the relevant creative, management and other acquisition costs. None of the results in the example is a universal target.
What to do next
Before the test, decide how much you can afford to spend, when you will review the data and which signal would make you stop the test. A confirmed payment error demands immediate action; variation of a few orders calls for caution in interpretation.
Keep a table with the launch date, budget changes, promotions and stock problems. When the result changes, this timeline helps you form verifiable hypotheses. A drop that appears as the popular variant sells out calls for a different response than one that appears after an audience change. Coincidence in time doesn't prove the cause, but it shows you where to start investigating.
A mistake to avoid: comparing the order rate per session with a platform's conversion rate per click. The denominators differ, so the percentages answer different questions.
3. Are the ads bringing the right people?
The symptom and possible explanations
Visits increase, but few people reach relevant products or add to cart. You may be attracting informational searches, people from areas where you don't deliver, or curiosity triggered by an image. An unclear page can produce the same symptom, so the traffic is not yet the diagnosis.
What to check and how to interpret it
In Google Ads, inspect the search terms available: the phrases people actually used. They can differ from the keywords chosen in the campaign. Group the searches by intent: buying, comparing, informing, or a need unrelated to the offer.
A store that sells office chairs can receive visits for “office chair repair”. The term relates to the product but calls for a different service. Google's search terms report also has visibility limits; some queries are omitted for privacy reasons. Don't treat the visible list as the complete inventory of searches.
Check locations, devices and entry pages. In Meta, compare ads and placements, meaning the places where the ads appear, using the breakdowns available in the account. Look at the actions after the visit, not just the cost of the click. Avoid firm conclusions about segments with very few observations.
If a segment brings relevant product views and carts but loses people at payment, you have reason to investigate checkout. If the need expressed is incompatible with the offer, the relevance problem is clearer.
What to do next
Exclude clearly unsuitable searches, correct the locations and adapt the message to the person who can buy. For analysing campaigns and search terms, you can consult BlueAds Google Ads services.
Keep a log of exclusions and the reason for each. That way you can revisit a decision if you accidentally removed a relevant commercial need.
A mistake to avoid: assuming a cheap click is automatically valuable, or that a narrower audience is always better.
4. Are you optimising the campaign for the right result?
The symptom and possible explanations
The report shows many “results”, yet orders don't grow. The result being tracked may be the visit, the interaction or the add-to-cart. These actions can help analysis, but they have a different value from an order.
What to check and how to interpret it
Ask for the exact definition of the result in the report. In Google Ads, check the goals each campaign uses and the conversion actions included. Typically, primary actions can be used for bidding when that goal is selected; secondary ones are for observation. There is an exception for custom goals, explained in the documentation on primary and secondary actions.
In Meta, check with the campaign administrator which result delivery is actually optimising for and which event confirms the purchase. Don't infer the configuration from the campaign name alone. A campaign named “October Sales” doesn't prove that the settings and the event are correct.
If ads are evaluated by visits, don't interpret traffic growth as proof of commercial efficiency. If they are evaluated by purchases but the purchase signal is wrong, correct the measurement first.
What to do next
Keep the intermediate actions for diagnosis and define explicitly the commercial result you are tracking. If you temporarily choose an intermediate signal, decide why, for how long and how you will verify its link with orders. A lack of data doesn't turn every easy-to-get action into a good goal.
For social campaigns, managing Facebook Ads through BlueAds can be the starting point for a conversation about goals, measurement and the relevance of messages.
For services, separate the submitted enquiry from the qualified enquiry and from the paying client. A tap on the phone number doesn't confirm a conversation, and a conversation doesn't confirm a sale.
A mistake to avoid: presenting a low cost per add-to-cart as a low cost of acquiring a customer.
5. Do the ad and the landing page promise the same thing?
The symptom and possible explanations
An ad catches attention, but the visitor quickly leaves the page or searches for a long time for the promised product. They may land on the homepage, find a different price or discover that the variant in the image is unavailable.
What to check and how to interpret it
Put the ad and the landing page side by side. Compare the product, variant, price, discount conditions, delivery and availability. Open the final address on a phone and watch for any redirects.
If you promote a set, the page must clarify what the set includes. If the ad shows the price of the cheapest variant, the visitor must be able to identify that variant and the price conditions. A general banner doesn't compensate for a product being out of stock.
In the entry-pages report of your analytics tool, compare the paths for the same offer. A category page can suit a broad need, while a product page suits a search for an exact model. Don't automatically send all ads to the same destination.
What to do next
First correct the verifiable mismatches: wrong link, expired discount, unavailable variant or hidden conditions. Then test the clarity of the first screen: can the person understand what they get and what the next step is?
For catalogues, check with the person responsible the source of the product data and when it was last updated. An offer changed in the store can remain inconsistent if other materials use old information.
Don't change the price, the destination and the message at the same time if you want to find out which change helps. Note the intervention and compare results under conditions as close as possible.
A mistake to avoid: using a spectacular promise to get more clicks and leaving the essential explanations until the checkout.
6. Is the offer compelling enough?
The symptom and possible explanations
People browse the products but don't find enough reason to buy. Price can be an explanation, but slow delivery, insufficient information or a simpler alternative can matter too.
What to check and how to interpret it
Compare the complete offer with real alternatives: final price, shipping, availability, what's in the package and the services included. Use the questions your support team receives. “Does it fit my model?” shows a different need from “Why is it more expensive?”.
Hypothetical example — margin and advertising spend. You analyse 100 orders kept by customers in a completed month, after returns are recorded. For simplicity, all commercial values exclude VAT, and the orders have the same average economics:
- Average net order value: 250 RON.
- Average product cost: 150 RON.
- Other variable costs, including shipping paid and processing: 25 RON.
- Contribution before advertising = 250 − 150 − 25 = 75 RON/order.
- Contribution margin before advertising = 75 / 250 × 100 = 30%.
If advertising costs 6,000 RON for this scope, the advertising cost per order is 6,000 / 100 = 60 RON. That leaves 15 RON per order before fixed costs and other expenses not included.
ROAS is the ratio between the revenue attributed to advertising and the advertising spend. In the example, 25,000 / 6,000 = 4.17. The simplified limit that consumes the entire contribution is 250 / 75 = 3.33, with no money left for fixed costs or profit. It is not a recommended ROAS for your business.
What to do next
Calculate the economics by category and test a relevant benefit: better explanations, a useful bundle or clearer delivery terms. A discount can increase orders and reduce total contribution; track both results.
A mistake to avoid: choosing the budget based on gross revenue or on ROAS, without checking the margin, returns and which costs are included.
7. Does the site work well on a phone?
The symptom and possible explanations
Phone traffic is significant, but orders appear mostly on desktop. There may be a usability, loading or payment problem. There may also be different buying behaviour, so the difference between devices is a clue, not a verdict.
What to check and how to interpret it
Use a real phone, without administrator login, and a normal mobile connection. Open the page used in the ad, select the variant, add the product to the cart and try to complete the purchase. Repeat on the main device and browser combinations seen in the data.
Look for concrete obstacles: a banner covers the button, the menu won't close, the size selector doesn't respond, the keyboard hides the field or the page shifts when an image loads. Also check navigation from the in-app browser of the app the visitor comes from.
Use PageSpeed Insights for technical clues. Core Web Vitals measure loading, response to interactions and visual stability, according to Google's explanation. Data from real users and lab tests answer different questions. A good score doesn't prove the form works and doesn't guarantee conversions.
What to do next
Send the developer the steps to reproduce, the device, the browser and a screenshot that doesn't expose personal data. Fix first the obstacles that block buying, then those that make it harder. Test again after the change.
If you need work on a store built on this platform, you can see BlueAds services for Shopify sites. Choosing the platform doesn't replace verifying the concrete implementation.
A mistake to avoid: rebuilding the whole site based on a score, or declaring the mobile experience good after checking only the homepage.
8. Does the product page answer questions and inspire trust?
The symptom and possible explanations
Visitors look at the product but hesitate to add it to the cart. An essential piece of information about compatibility, dimensions, use or delivery may be missing. It may also be unclear who answers if the product isn't suitable.
What to check and how to interpret it
Pick the products that receive paid traffic and read the pages as a new buyer. Do the photos show the real product, the details and the proportions? Does the description explain who it's suitable for? Are the available variants and the stock clear?
Compare the questions from messages, calls and return requests with the information displayed. If people repeatedly ask about installation, a visible explanation can be more useful than another paragraph about “quality”. If the product needs a separate accessory, say so before purchase.
Check the contact details, the merchant's identity, the delivery and return terms and how reviews are presented. Use authentic reviews and describe how they are verified correctly. Don't invent ratings and don't add symbols that suggest non-existent certifications.
What to do next
Rewrite the information in the order of the decision: what the product is, what problem it solves, whether it fits and what buying involves. Add useful photos and answers to documented objections. Put the important information next to the choice and purchase area, without cluttering the screen.
First test a group of products with enough traffic and keep a record of the changes. If add-to-carts rise, also track the orders customers keep: a description that creates wrong expectations can shift the problem to returns.
A mistake to avoid: compensating for missing information with superlatives, artificial countdown timers or stock claims that don't reflect the real situation.
9. Is checkout simple and functional?
The symptom and possible explanations
People add to cart, but few complete the purchase. There may be unexpected costs, a mandatory account, unnecessary fields, a missing preferred payment method or an error. Some people also use the cart for comparison, so not every abandonment is automatically a fault.
What to check and how to interpret it
Compare the steps using the same unit of measure. Don't divide the raw number of add-to-cart events by the number of people who buy: the same person can add products repeatedly.
Hypothetical example — a blockage between cart and order. Over a 14-day interval, you track a closed path, with the steps in order and within the same session. Out of 200 sessions with an add-to-cart, 100 reach checkout initiation and 20 reach purchase. Each session is counted once at each step.
- Cart checkout started = 100 / 200 × 100 = 50%.
- Checkout started purchase = 20 / 100 × 100 = 20%.
- Cart purchase = 20 / 200 × 100 = 10%.
If the last two comparable weeks had different results, the difference justifies an investigation. These percentages don't establish the cause on their own and are not industry standards. Returning and buying in a different session are not included in this path.
Check the payment processor's logs, failed orders and complaints received. Test addresses in the areas where you deliver, valid coupons, payment and the return to the store after authorisation. Don't test only cash on delivery if you also promote card payment.
What to do next
Show the total cost before the last step, remove requirements with no operational justification and explain errors next to the affected fields. Evaluate guest checkout and payment methods against your platform, audience and costs.
A mistake to avoid: trying to recover with ads some carts abandoned because of a payment error that is still active.
10. What happens to people who don't buy on the first visit?
The symptom and possible explanations
The products receive interest, but buying rarely happens in the first session. The visitor may compare, may wait for payday or may need another person's approval. Sometimes they don't come back because they haven't found an answer, not because they forgot the store.
What to check and how to interpret it
Analyse the time to purchase, the sources of return visits and the questions asked before ordering. In your email tool, check the existing messages, the associated orders, unsubscribes and the exclusion of people who have already bought.
Remarketing means advertising communication with people who have previously interacted with the business. Use it where it is appropriate and permitted, with eligible audiences and signals. For Google products, the EU user consent policy explains the consent obligations for ad personalisation. Turning on a feature doesn't automatically solve the requirements that apply to your data.
What to do next
Prepare messages that answer a real hesitation: a buying guide, compatibility, delivery or help with ordering. Use email recovery only for the contacts and situations where you have the right to communicate this way; check the applicable basis before activating. Respect withdrawal of consent and avoid repetitive pressure.
A BlueAds study on email marketing for a lighting fixtures store describes newsletters and automations, including cart recovery. It is a public example about a business with existing traffic and subscriber base. It doesn't establish what revenue you will get and doesn't demonstrate on its own the incremental effect of each message.
For services, check the enquiry path separately: did it reach the responsible person, was it picked up, qualified and followed through to a decision? A relevant enquiry lost between the form and the sales team isn't fixed just by more traffic.
A mistake to avoid: attributing all return orders to the last ad and assuming that, without that ad, they wouldn't have happened.
Diagnostic table: from symptom to next check
Start with the row for which you have concrete evidence. If several symptoms appear at once, first fix whatever prevents an order from being recorded or paid. After the fix, measure the path again before you change the audience. That way you avoid cutting suitable traffic over a problem that was in the store.
Use the table as a working tool. A possible cause becomes a diagnosis only when you have evidence. For each confirmed problem, note the owner, the deadline and how you will verify the fix.
| Symptom | Possible cause | How to check | Action |
|---|---|---|---|
| The store has orders, the report doesn't | Missing event or different attribution | Test order, identifiers, periods | Fix the collection or align the comparison |
| The report shows too many purchases | Duplication or overlap between platforms | Compare unique orders and reported actions | Remove the confirmed duplication |
| Few visits, zero orders | Insufficient sample | Relevant sessions and time to purchase | Continue only in a test with a cost limit |
| Traffic from incompatible needs | Unsuitable terms or messages | Available searches, locations, destinations | Correct relevance and exclusions |
| Many results, few orders | Optimising for an intermediate action | Result definition and campaign goals | Align the commercial goal |
| The promised product can't be found | Inconsistent destination or offer | Compare the ad with the page | Correct the link, message and stock |
| Interest without add-to-carts | Insufficient offer or information | Questions received and commercial alternatives | Clarify the benefit and total cost |
| Loss concentrated on mobile | Usability obstacle | Phone test and device data | Fix the reproducible problem |
| Carts without purchases | Error, unexpected fee or unsuitable payment | Comparable path, logs and tests | Fix the affected step |
| Interested visitors who don't return | Long decision or lack of information | Return history and questions | Offer help and permitted communication |
A practical plan for the first seven days
Day 1 — check that people can buy. Go through an order on a phone and verify payment, confirmation and recording in the store. Reduce or stop traffic to a flow shown not to work until it is fixed. Save the evidence and name the owner.
Day 2 — check measurement. Follow the events of the test order and compare a completed period between the store and the reports. Document the differences explained and those still unresolved. The result of the day is a shared definition of the order and of the revenue reported.
Day 3 — check traffic and goals. Analyse the available searches, locations, destinations and the results being tracked. Correct the clear mismatches. Don't exclude segments just because they have few conversions in a small sample.
Day 4 — check the offer and priority pages. Pick the products that consume budget and answer the essential questions about price, variants, delivery and use. Correct promises not aligned with the ad. Keep the list of interventions.
Day 5 — fix the checkout obstacles. Check the forms, total costs and payment methods. Retest the repaired problems under the same conditions in which they appeared. A “fixed” label without retesting doesn't close the check.
Day 6 — prepare useful follow-up. Review the existing messages, contact eligibility and the exclusion of buyers. Choose a real question to answer, before you automatically add yet another discount.
Day 7 — decide the next test. Consolidate the confirmed problems, their probable cost and the results observed. Choose a priority hypothesis, an owner and a spending limit. Seven days can organise the diagnosis; they don't guarantee enough data for a stable commercial conclusion.
When does it make sense to increase the budget?
An increase is worth testing when orders are credibly measured, purchasing works, traffic is relevant and the contribution remaining after advertising supports the business. You also need stock, shipping capacity and cash for the interval between spend and collection.
Look at results over periods that take into account the buying cycle and returns. One good day, an exceptional promotion or a few large orders don't necessarily describe the pace you can sustain.
Also separate categories with different economics. A product with high volume and low contribution can dominate the report, while a smaller category supports the financial result. Check the availability of the promoted variants and the cost of restocking. Order growth that forces the store to promise unrealistic delivery times can bring cancellations and extra support requests.
Increase in a controlled way, with a ceiling and a date for reassessment. Track the cost of the additional orders, not just the average for the whole period. As you widen distribution, the next orders can have different economics from the first ones. Keep a reserve for variation and don't assume a proportional relationship between budget and revenue.
Fix the problem first if payment fails, the offer isn't available, conversions are duplicated or orders consume more than their contribution. If the data is insufficient, you can fund a limited test, but call it a test, with explicit stopping criteria.
Ask for specialist help when you can't reconcile orders, when the problem spans ads, site and payment processor, or when decisions exceed the financial risk you can tolerate. Ask for verifiable findings, not just a list of settings changed.
Checklist before the next budget increase
- I went through a test order and checked payment and confirmation.
- I know what each conversion in the report means.
- I separated recorded orders from those collected and kept.
- I compare the same periods, segments and units of measure.
- I have taken into account the time needed to buy.
- I checked the relevance of traffic and the campaign goals.
- The ad, page, price and stock are consistent.
- The offer answers the buyer's questions.
- The phone experience and checkout have been tested.
- I know the contribution available for advertising.
- Follow-up communication is relevant and permitted.
- I have defined the ceiling, the owner and the moment of reassessment.
Frequently asked questions
Should I stop all ads until I finish the checks?
Not automatically. If you have confirmed a general payment blockage, temporarily stopping the affected traffic can limit the loss. If the problem concerns one product or destination, the intervention can be local. When you only have uncertainty, set a limited test instead of making a general decision based on assumptions.
Can I analyse a new campaign using the average for the whole store?
The average can provide context, but it includes existing customers, brand traffic and visits with different intents. Compare mainly segments that are close in source, product, device and period. Mention separately if the new campaign aims to attract people who don't yet know the store.
What do I do if I don't have access to the advertising accounts?
Ask for access appropriate to your role and for reports that explain the sources, periods and definitions of the metrics. In the meantime you can check orders, payments and how the site works. A screenshot with the total conversions isn't enough for reconciliation or for a budget decision.
How do I analyse orders paid cash on delivery?
Separate the order placed from the parcel delivered and from the amount collected. Track refusals and logistics costs over sufficiently mature periods. Use the same definition of revenue when comparing campaigns; otherwise a source with many refused orders can look more efficient than it is.
Can I test two page versions if I have little traffic?
You can prepare a test, but reaching a useful conclusion can take a long time. First fix the observable errors and use customers' questions to prioritise. A before-and-after comparison gives clues, but seasonality or different traffic can explain part of the result.
Who is responsible for the lack of sales: the agency or the developer?
Responsibility is set according to the verified cause. The agency can investigate the relevance of traffic and the campaign configuration, the developer how the site works, and the merchant the offer and operations. For shared problems, appoint a coordinator and a closing criterion: for example, the same order appears correctly in the relevant systems.
If I lower the price and orders appear, have I solved the problem?
You have gotten a signal that the offer influences the decision. But check the total contribution, the quality of orders and returns. The discount can attract demand that isn't profitable or can work only temporarily. Also compare alternatives such as clarifying the benefit or improving the delivery terms.
Next step: find out where the order is lost
When you have clicks but no sales, start with a real test order and continue with comparable data. Choose the next intervention based on evidence and the impact on purchasing.
If you need to tie campaign analysis to what happens in the store, talk to the BlueAds team. Prepare the period analysed, the spend, the confirmed orders and the problems you have observed. This information makes the conversation more concrete and helps set priorities.