If you run paid campaigns long enough, you know the feeling. You open your dashboards, glance at performance, and something feels off. Spend is climbing, results are not. You can tell money is leaking somewhere, but it is buried under layers of ad sets, keywords, and charts.

That leak is what we call a money pit - budget that quietly disappears into ads or keywords that technically “run” but do not move real outcomes.

At third i, we wanted to stop finding these leaks at the end of the month. So we built two analyzers - one for creatives and one for keywords - that behave like always-on sentries for your ad accounts. Instead of you digging for problems manually, they scan patterns in the background and surface the ones that actually deserve attention.

Here is how we designed them.

What we mean by “money pits” in paid campaigns

Not every underperformer is a money pit. Some ads are still in learning, some keywords are experimental, some tests need time.

A money pit is different:

• It keeps getting a budget.

• It does not contribute meaningfully to revenue or core goals.

• It often hides behind “average” looking numbers until you look a layer deeper.

For creatives, that might be an ad that still gets impressions but has clearly fallen flat with your audience. For keywords, it might be a term that attracts clicks but rarely leads to conversions. In both cases, the account slowly bleeds, and no single alert screams loud enough on its own.

The analyzers we built are designed to isolate exactly those patterns.

How our Creative Analyzer finds “money pit” ads

We treat creative analysis as a pipeline, not a gut call. Every active ad goes through a series of checks that answer three simple questions:

1. Is this ad still worth showing at all?

2. If yes, is the format doing its job?

3. If yes, is the problem actually somewhere after the click?

If an ad fails one of those gates, it gets flagged with a clear label and a suggested action, instead of a vague “bad performance” tag.

Phase 1: Catching obvious drains

The first pass looks for creatives that are clearly wasting budget and should be reviewed immediately.

The system looks for patterns such as:

• Engagement drops off while the same people keep seeing the ad again and again.

• Platform quality signals that suggest you are paying more than usual to deliver this particular ad.

• Impressions being pushed into an audience that has already seen the message too many times.

When it finds those patterns, it marks the creative as an urgent candidate for pausing or refreshing. The goal is simple: stop pouring money into assets that the platform and your audience have already rejected.

Phase 2: Understanding format-specific issues

If an ad passes the first gate, the analyzer looks deeper at how the format behaves. A video failing in the first two seconds is very different from a static image that people scroll past without noticing.

Some of the patterns we look at:

• For video, where viewers tend to drop off - right at the start, in the middle, or only near the end.

• For video, whether people are staying long enough to hear or see the main message.

• For static creatives, whether people even pause long enough to click, despite reasonable delivery costs.

Instead of treating all “low performance” creatives the same, the system tries to pinpoint where the attention is leaking: at the hook, in the story, or at the visual level. That gives your creative team something specific to fix: open stronger, simplify the story, adjust the layout, and so on.

Phase 3: Separating ad problems from funnel problems

Finally, the analyzer checks what happens after the click.

Sometimes an ad is doing exactly what it should: people notice it, click it, and land on your page. The drop-off happens later - on the landing page, in the form, or in the checkout.

When the system sees that pattern, it does not punish the creative. It flags the issue as a funnel or post-click problem. You still get an alert, but the message is closer to:

• “This ad is pulling its weight. Look at the page or flow it sends traffic to.”

This simple distinction prevents a classic mistake: killing a strong creative because the real issue lives further down the path.

How our Keyword Analyzer digs into wasted spend

Keywords are trickier than creatives. A single term can underperform for many reasons - wrong intent, poor match type, weak ad, misaligned landing page - and applying the wrong fix can make things worse.

Our Keyword Analyzer is built around one idea: do not just flag “bad” keywords. Explain why they are not pulling their weight.

Step 1: Spotting clear drains

First, the system looks for terms that behave like pure drains - they regularly consume budget over time while rarely leading to valuable outcomes.

It also checks for keywords that the platform has quietly made more expensive, for example when relevance and experience signals are weak. In those cases, every click costs more than it should, even if headline metrics do not look extreme at a glance.

These terms are your classic money pits on search: they tick along in the background, never outright crashing performance, but quietly dragging your blended costs up.

Step 2: Identifying the root cause

Next, the analyzer looks at how each keyword fits into the bigger picture, instead of treating them all the same. It asks questions like:

• Is this a strong keyword that simply is not getting enough exposure yet?

• Is this a mismatch between what people search for and what your ad promises?

• Is this a situation where people click, but do not find what they expected after the click?

By comparing intent signals, engagement, and what happens post click, the system tries to slot each term into one of a few “why” buckets:

• Budget or exposure issue - a solid term held back by limited bids or share.

• Relevance issue - what the searcher wants and what you offer do not line up.

• Funnel issue - the keyword and ad do their job, but the experience afterward breaks.

The recommendation that comes out of this is very different depending on the bucket. You might increase bids, rewrite copy, rework the landing page, or in some cases retire the keyword entirely.

How we keep the AI from adding noise

The hardest part of building these analyzers was not getting them to spot patterns. It was making sure they were quiet most of the time, and loud only when it truly matters.

To keep them useful rather than overwhelming, we follow a few simple design rules.

Wait for real signals, not random spikes

The system does not pounce on every small fluctuation. It waits until it has seen enough consistent behavior over time and enough data points before calling something a problem.

That means you are not getting alerts every time a campaign has a weird day. You hear about issues when they look like real trends instead of random noise.

Measure against your own account, not generic benchmarks

“Good” and “bad” performance is different for every account, industry, and stage of growth.

So instead of using static global benchmarks, the analyzers learn what “normal” looks like for your setup - typical click-through rates, usual costs, usual conversion behavior - and flag items that deviate from that baseline in a meaningful way.

This makes the diagnosis more personal and makes the suggestions feel less like canned advice and more like context-aware guidance.

Show the highest-impact leaks first

In any decent-sized account, there are always many things you could tweak. The question is: which ones actually move the needle?

When the system sees multiple issues on the same creative or keyword, it surfaces the most important one first, the one that explains the largest part of the waste. It also tries to focus your attention on a short list of items where fixing them will have outsized impact.

The outcome should feel less like “yet another dashboard” and more like a focused list of repair jobs.

Why this matters for marketers, not just engineers

All of this logic lives under the hood. Most users never see the internal rules. What they feel is the effect:

• Fewer surprise weeks where spend looks fine but profit is squeezed.

• Clearer explanations of why specific ads or keywords are dragging you down.

• Concrete ideas on whether to pause, fix, or scale different parts of the account.

You still own the strategy. The AI just takes the first pass through the numbers, highlights the quiet leaks, and gives you a head start on where to look.

By the time you sit down with your morning coffee, the analyzers have already gone through your accounts and circled the likely money pits. Once those are found and handled, everything else - scaling, testing, creative experiments - becomes a lot easier.

If you want to see what they uncover in your own data, plug in your accounts and let them run in the background for a few days. The patterns they surface are usually the ones that are hardest to spot by hand.