Ad Spend Going Out, No Conversions: A Diagnostic Order

In the dashboard everything looks like it is working: impressions are there, clicks are there, the budget burns down to nothing every day. At the end of the month the only thing missing is sales. Most businesses facing this picture reach for the same two reflexes — raise the budget, or change the keywords. Neither is a diagnosis; both are guesses.

In Turkish-language sources the usual answer to this problem is a list: nine reasons, fifteen causes, five critical mistakes. Most of the items are correct. The problem is not the items, it is the order. What nobody writes down is which check comes first, what result on one step lets you move to the next, and what evidence proves each step. Arguing about bid strategy on an account where measurement is broken is like planning a diet on a scale that is out of calibration.

The sequence below was built for exactly that gap. There are six steps, they run in order, and each one has a written definition of "clean". If a step is not clean you do not move to the next one, because the data you would see there is already wrong.

If clicks are the only thing growing, the problem is not the budget

An ad account has four quantities: impressions, clicks, conversions and revenue. The four form a chain, and when any link in the chain breaks the next link stays empty. The sentence "budget goes out but no conversions come in" does not tell you where the chain broke; it only tells you the last link is empty.

Raising the budget does nothing except increase the volume arriving in front of the broken link. If the break is in measurement, more clicks means more invisible conversions. If the break is on the landing page, more clicks means more visitors bouncing back. In both cases the invoice grows and the picture does not change.

The right question is: where does the chain break? The way to find out is not to line up every possible cause side by side, but to start at the top link and work down. The order is not arbitrary, because each step produces the data for the next one. If measurement is broken, the search term analysis gives you the wrong answer; if the search terms are irrelevant, evaluating the landing page is pointless; if the landing page does not work, there is no correct signal for the bidding algorithm to learn from.

Step 0 — write down what counts as a conversion

There is one job to do before the diagnosis, and it is not technical: write, in a single sentence, which user action counts as a conversion. Without that sentence every other measurement becomes contested, because the parties use the same word for different things.

In practice three separate definitions circulate at once. The ad platform may be counting the "form submitted" event. The sales team counts "the person who picked up the phone and asked about price". The business owner counts only "the person who paid". All three numbers are correct on their own terms and none of them agree. Surprisingly often, the "no conversions" complaint is born from those three definitions colliding.

Here is the work: put the conversion candidates into a table — form, phone call, WhatsApp, appointment request, add to cart, sale. Next to each, write its value to the business and where it is recorded. Then declare one of them the primary conversion and leave the rest secondary. The primary conversion is what the bidding algorithm will chase; defining more than one primary conversion means handing it a target with no direction.

The test for this step is simple: the person running the ads, the person making the sale and the person paying the invoice can all state the same definition in the same sentence. If they cannot, the problem is not in the account, it is in the definition. When you are choosing which metric actually ties back to a business outcome, our article on customer data metrics for small businesses gives you a starting list.

Step 1 — measurement verification: is the conversion actually being recorded

This step comes first because when it is broken it corrupts the data for every other step. If an account shows close to zero conversions there are two possibilities: conversions genuinely are not happening, or they are happening and not being recorded. They are entirely different problems, and confusing them costs months.

The check is manual and takes half an hour. Go to your own site from outside, like a real visitor; if you can, arrive by clicking the ad. Fill in the form, tap the phone link, do whichever action you are measuring. Then confirm that this single event shows up in the dashboard. The record can land with a delay, which is why you need to look twice on the same day.

The common breaking points are these: after the form is submitted the visitor is not redirected to a thank-you page, so a page-based setup never fires; the tag is loaded on only part of the site; the measurement chain is silently cut when cookie consent is declined; the same conversion is counted separately by both the ad platform and analytics and so doubles; the payment or booking step moves to a different domain and the session breaks.

The evidence for this step is: you see with your own eyes that your test conversion appears in the dashboard as a single record, under the correct conversion name, with a reasonable delay. If you cannot see it, stop here. For reading where a measurement chain breaks, the checking logic in our article on tracking traffic and conversions in GA4 applies here as well.

Step 2 — search term data: which query is the money going to

If measurement is clean, the second question is: on which search queries is your ad being shown and clicked? Your keyword list shows what you wrote; the search terms report shows what the user actually typed. The gap between the two is the most visible source of wasted spend.

When you open the report, sort the terms into three buckets. Bucket one: queries that carry purchase intent and describe your service. Bucket two: queries on the topic but with informational intent — "what is", "how to", "free", "can I do it myself". Bucket three: queries that are entirely unrelated, or describe a product, city or segment you do not serve.

Bucket three goes straight onto the negative list. Bucket two requires a decision: those queries will not produce sales today but they do describe future demand, and meeting them on the content side rather than in ads is usually cheaper. Bucket one is where the budget should actually go, and in most accounts its share of total spend comes out smaller than expected.

The evidence for this step is a ratio: how much of the last thirty days' spend landed in bucket one? Calculate it once and write it down. If the number is low, the problem is not in the bid strategy — look at match types and the negative list. If the ratio is reasonable, move to the next step.

Step 3 — the landing page: is the promise from the ad on the first screen

If the right money is going to the right query and there are still no conversions, it is time to look at what happens after the click. There is one question here: is the promise in the ad copy met on the first screen of the page that opens?

The practical way to measure that is to open the page on a phone and look without scrolling. Three things need to be on that first screen: the name of the service from the ad, in the same words; one sentence saying who that service is for; and a single call to action. If the ad says "evening appointments" and the page never mentions it, the visitor bounces before working out that they came to the right place.

The second common mistake is pointing ads at the home page. Because the home page has to describe every service at once, it describes none of them well enough; it leaves the visitor hunting through the menu for what they came for.

The third is form friction: a form with ten fields does not get completed by as many people as one with three. Ask the sales team which field is genuinely necessary; every field justified with "we can always ask later" can come out of the form.

The evidence for this step: show the page on a phone for five seconds to someone who has never seen it, ask "what does this page sell and what is the next step", and get the right answer. If you do not, the problem is on the page and no bid adjustment will fix it.

Step 4 — bid strategy and the learning period: is there enough data

If the previous three steps are clean, the bidding side can be examined. The real issue here is not the names of the strategies but whether there is enough data for the algorithm to learn from. Conversion-based automated bidding tries to extract a pattern from past conversions; where there is no steady flow of conversions, there is no pattern to extract.

There is no single correct threshold; the threshold depends on the account's own volume. But the direction is clear: in a campaign that struggles to find a few conversions a week, a strategy targeting cost per acquisition behaves erratically. On accounts like that, accumulating data with a simpler method first is a defensible order of operations.

The second issue is the frequency of changes. After every significant change the campaign enters a relearning phase, and performance fluctuates during it. An account where bids, budgets and targeting are touched every other day stays permanently in the learning phase. Batching changes and applying them on a weekly rhythm produces a more readable picture than one-off interventions.

The evidence for this step: when you open the change history for the last thirty days, every change has a written reason and a measurable expectation next to it. If what you see is a pile of unexplained changes, some of the performance fluctuation is being produced by the account itself. We went into where measurement falls short in our article on ad measurement, consent, GCLID and modelling, and into how much can defensibly be handed over to automation in our article on PMax and AI Max.

Step 5 — budget and campaign type: which line item should be switched off

The last step is the one most often postponed: looking at how the budget is divided. On small-budget accounts the most common pattern is money split across a large number of campaigns. A budget split across five campaigns can leave all five below the learning threshold.

The question to ask here is: which of these campaigns, if it did not exist, would leave the end-of-month business result unchanged? Read the answer from the data, not from a guess. Keeping searches for your brand name and searches for your service in separate campaigns makes that reading easier; when both sit in one campaign, the good numbers from brand traffic mask the bad numbers on the service side.

The evidence for this step: you can write one sentence next to every active campaign explaining why it exists. A campaign you cannot write that sentence for is a candidate to be switched off.

The evidence for each step: what you look at before saying "clean"

Diagnostic work with no written output is work nobody remembers. The table below collects the evidence for all six steps in one place; filling in each row and dating it stops you from having the same argument again next month.

Once you have filled the table in, you hold a diagnostic report. Filling in the same table again in later months and putting the two versions side by side tells you more than tracking individual metrics does.

The common sequencing mistake: changing keywords while measurement is broken

The most expensive mistake is not doing a step wrong, it is doing the steps in the wrong order. On an account where measurement is broken, changing keywords, raising the budget or trying a bid strategy teaches you nothing, because you cannot see the result. When you cannot measure whether a change worked, all you are left with is a feeling.

The second form of the same mistake is making several changes on the same day. When the keyword list, the bid strategy and the landing page all change at once, you will not know which one fixed it even if the picture improves. An unknown cause does not produce a repeatable method.

The third form is impatience with time. The verdict "this campaign is not working", delivered before enough data has accumulated, is usually an early verdict; the criterion is not the number of days but whether enough conversions have built up to decide on.

What comes out of the process changes too when the order is right: instead of "the ads are not working", you are left with something you can act on, like "measurement breaks here" or "this much of the spend is going to bucket two".

How to ask your agency for this diagnosis

If an agency manages the account you do not have to run these six steps yourself; but you do have to ask for the output in writing. The request fits in one sentence: "Could you share, in writing, the current state and the evidence for each of these six steps?"

A good answer looks like this: a screenshot of the test conversion and its date, the bucket-level breakdown of the last thirty days of search terms, the mobile first screen of the landing page, an itemised change history with reasons, and a reason to exist for each campaign. A bad answer is a one-line performance summary and the phrase "we optimised it". We built the wider version of this distinction in our article on whether your SEO agency is doing GEO, written for auditing an existing supplier; the logic there works the same way on the advertising side.

The second question is commercial: where does this diagnostic work sit in your fee model? In a relationship priced as a percentage of spend, a diagnosis that recommends cutting the budget carries no incentive for the supplier. That is not an accusation, it is the natural consequence of the model — and it is a solvable topic when it is discussed up front. We covered how the models differ and what to look for in the contract in our article on ad agency fee models and account ownership.

If you want to see the order in which we run this diagnosis and which items are in scope, the scope list on our ad management service can be a starting point. If you would like to fill in the table for your own account together, you can get in touch.

Frequently Asked Questions

If the ad budget is being spent but there are no conversions at all, what do you check first?

The first thing to check is not the campaign settings but measurement itself. A situation where conversions genuinely are not happening and one where they are happening but not being recorded look identical in the dashboard, yet they are completely different problems. Verification is manual: go to the site from outside, fill in the form, and confirm that this single event appears in the dashboard under the correct name as a single record. Until you have confirmed that, no change on the keyword, bid or budget side produces a readable result.

How many days should you wait for results, and when should you intervene?

The criterion is not the number of days but whether enough data has accumulated to decide on. On an account getting a few conversions a week, a week of data says nothing; on an account getting dozens of conversions a day, the same period can be meaningful. The practical approach is to tie changes to a weekly rhythm and to write down the expectation for each change in advance. An account intervened in every other day stays permanently in the relearning phase, and some of the fluctuation is produced by the account itself.

Is traffic that clicks but never calls fake click traffic?

Usually it is not; but before claiming that, you have to eliminate the measurement and relevance side. Three possibilities are checked before the suspicion of click fraud: the conversion record may have broken, the ad may be showing on irrelevant search terms, the landing page may not be meeting the promise in the ad. If all three are clean and there is a regular pattern in the clicks, then examining traffic sources and the distribution over time becomes meaningful. Reversing that order turns a solvable problem into an unsolvable complaint.

Is it right to pause a campaign while the conversion count is low?

The decision to pause is right if it is taken after the diagnostic sequence is complete. A campaign paused before measurement has been verified carries the risk of shutting down a channel that actually works but is invisible. If the sequence is complete and you cannot write in one sentence why the campaign exists, pausing is a reasonable decision. Concentrating the budget in a small number of campaigns is generally the more defensible choice, especially on small budgets, because it stops every campaign from sitting below the learning threshold.

If the agency is not doing this diagnosis, what questions should I ask?

Ask for the current state and the evidence for each of the six steps in writing: the dated record of the test conversion, the intent-level distribution of search terms, the mobile first screen of the landing page, an itemised change history with reasons, a reason to exist for each campaign, and who approved the conversion definition. Also ask how the fee model relates to this diagnosis: where pricing runs on a percentage of spend, a diagnosis that recommends cutting the budget carries a weak incentive for the supplier, and that is a topic to discuss in advance.