How to Use Negative Keywords to Cut Wasted Ad Spend

If your search terms report is full of “cheap,” “free,” and random job seekers, you do not have a traffic problem — you have a filtering problem. That is the real starting point for how to use negative keywords well. Deloitte’s 2021 digital marketing strategy guidance explicitly recommends negative keywords to stop ads showing on phrases such as “free” and “cheap,” while Skai’s 2024 guide argues they reduce wasted spend, improve click-through rate, and sharpen campaign performance. Most teams still treat negatives as a cleanup chore at the end of the week. That misses the point. Negative keywords are one of the fastest controls you have for tightening query intent, protecting budget, and improving the economics of the entire account.

Paid search gets traffic quickly. Deloitte makes that plain: paid search is a short-term strategy that can rank on the first page straight away. That speed is useful, but it also means waste accumulates quickly when your filters are weak. A campaign can generate impressions, clicks, and even early conversions while quietly filling the top of the funnel with traffic that was never going to become revenue. We have seen this pattern repeatedly in SaaS and lead generation accounts: teams focus on bids, creatives, and landing page design first, while leaving query filtering loose. The result is predictable. Search volume rises, CTR drifts, CPC edges up, and nobody can quite explain why efficiency keeps slipping.

This article argues for a different view. Negative keywords are not housekeeping. They are a control system for intent. Used well, they improve the fit between search query, ad copy, and landing page, which makes every other optimization effort work harder. Used badly, they block good traffic, flatten learning, and starve campaigns that need room to discover converting demand. That trade-off matters, especially in accounts using automation and broad targeting. We will show exactly how to decide what to exclude, which match type to use, where to structure exclusions, and when to wait for more data instead of cutting too early.

If you also care about the economics behind traffic quality, our guides on measuring ROAS properly and the reporting metrics that actually map to growth pair naturally with the process below. Negative keywords do not sit in a silo. They shape what kind of demand reaches your funnel in the first place.

Why wasted spend starts in search intent

Search waste does not begin with a bad bid. It begins when the platform matches your ad to a query that never had a realistic chance of producing value. Search Engine Land’s 2026 article on negative keyword strategy makes the mechanism clear: misalignment between query, ad, and landing page wastes budget, drags down CTR and Quality Score, and pushes CPCs up. That is the chain most teams underestimate. One irrelevant click is not just one bad click. It can weaken the performance signals the account depends on.

Why does one irrelevant query hurt more than one wasted click?

Because Google Ads does not evaluate clicks in isolation. It reads patterns. If your ads repeatedly show for searches that the audience does not actually want, CTR falls. Lower CTR feeds weaker relevance signals. Weaker relevance often translates into poorer Quality Score, and that tends to raise your cost to win the next useful click. The damage compounds.

Consider a simple example. A SaaS team spends $12,000/month on non-brand search. Their average CPC is $8, so they buy 1,500 clicks. If 18% of those clicks come from low-intent queries such as “free software,” “jobs,” and “template,” that is 270 wasted clicks, or $2,160/month in direct wasted spend. But the larger problem appears in the performance layer:

• Good-intent queries average 5.4% CTR
• Low-intent queries average 1.8% CTR
• Blended account CTR falls from 5.4% to roughly 4.7%
• If the weaker relevance signal pushes average CPC from $8.00 to $8.60, the same budget buys 1,395 clicks instead of 1,500

That is the part people miss. Poor filtering does not just waste $2,160. It can also reduce the number of valuable clicks you can buy next month.

What happens when query, ad, and landing page do not match?

Think of it as a misalignment chain. The query sets the user’s expectation. The ad either confirms or muddies that expectation. The landing page then either fulfills it or breaks the promise. If the searcher types “free CRM tool,” sees an ad for a premium B2B platform, then lands on a demo request page, the whole chain fails before your product even gets judged on merit.

This is why negative keywords matter even when your copy and landing pages are strong. You can have excellent messaging and a high-converting page, then still bleed performance because the wrong audience keeps arriving. That is also why negative keyword work complements page testing. If you are running experiments on landing page variants, cleaner query intent gives you cleaner test inputs. Otherwise you risk blaming the page for problems caused upstream in search matching. That is exactly why teams doing structured testing on acquisition pages usually see better signal when they tighten traffic quality first.

The contrarian point: more traffic can make you less efficient

Many advertisers still treat volume as proof that a campaign is healthy. It is not. More impressions and more clicks can be a sign that your filters are too loose, not that your targeting is improving. Paid search scales quickly, as Deloitte 2021 notes, which is precisely why waste can grow faster than teams expect.

The edge case is early-stage exploration. If you are launching a new product category or entering a new geography, you may want broader query exposure at first to learn what buyers actually search for. In that case, do not lock the account down on day one. But even then, you should isolate exploratory campaigns and budget them deliberately. Loose matching across the whole account is not learning. It is drift.

That sets up the next question: if query intent matters this much, what do negative keywords actually do inside Google Ads?

What negative keywords actually do

At a practical level, negative keywords tell Google Ads where not to show your ad. Karooya’s 2025 guide describes them as a way to block unwanted, irrelevant, or low-value searches that eat budget without converting. Skai’s 2024 guide makes the same point and adds the business outcome: fewer irrelevant queries, better CTR, and stronger ad performance. That is the operational definition we recommend using.

How do negative keywords work in Google Ads?

They work by excluding terms or patterns from eligibility. If a user’s query matches your negative logic, your ad does not enter the auction. This is why negatives affect more than cost. They reshape the pool of searches your campaign can win.

Skai gives a straightforward example: a furniture retailer might exclude “cheap” to avoid searches such as “cheap modern sofa” or “cheap dining chairs.” The retailer is not just avoiding a few bad clicks. It is protecting the campaign from users whose intent conflicts with a premium offer.

The same logic applies in SaaS. If you sell enterprise analytics software with contracts starting at $15,000/year, then terms like free, open source, course, jobs, and salary often belong on your exclusion radar. Not always, but often. A premium product should not keep paying for people whose search intent announces price resistance or a completely different objective.

What is the difference between blocking traffic and improving relevance?

Blocking traffic sounds defensive. Improving relevance sounds strategic. In practice, they are the same mechanism viewed from different sides.

When you exclude a bad query, you do two things at once:

• You stop paying for a click that was unlikely to convert.
• You improve the average fit between the remaining queries, your ads, and your landing pages.

That second effect matters. Better fit usually means higher CTR, stronger Quality Score signals, and cleaner performance reporting. Karooya 2025 explicitly says negatives can improve ad relevance, increase CTR, help improve Quality Score, lower CPC, and improve ROAS.

A simple before-and-after traffic example

Consider a campaign targeting “project management software” with a monthly budget of $20,000.

Before exclusions:

• 2,500 clicks at $8.00 CPC
• 22% of clicks come from terms containing free, template, jobs, or training
• Conversion rate on good-intent queries: 4.8%
• Conversion rate on low-intent queries: 0.4%

Math:

• Good-intent clicks: 1,950 → 93.6 conversions
• Low-intent clicks: 550 → 2.2 conversions
• Total conversions: 95.8
• Cost per conversion: $208.77

After exclusions remove half of that low-intent waste and average CPC falls to $7.60 due to stronger relevance:

• Budget still $20,000
• Clicks increase to 2,631
• Low-intent share drops to 11%
• Good-intent clicks: 2,342 → 112.4 conversions
• Low-intent clicks: 289 → 1.2 conversions
• Total conversions: 113.6
• Cost per conversion: $176.06

That is a 15.7% reduction in CPA without changing the offer, the creative, or the landing page.

The edge case: not every “free” search is bad

This is where teams get clumsy. A query like “free trial CRM” may be perfect if your motion includes a trial. A query like “free dashboard template” may even convert later through content-driven nurture if your economics support that path. The point is not to block obvious cheap intent mechanically. The point is to ask whether the query aligns with your offer and buying path.

That leads directly to the next operational choice: the match type you use determines how sharply or broadly you exclude demand.

Use the right match type

Negative keywords are not one blunt instrument. Search Engine Land’s 2026 guidance draws a clean distinction: use negative exact match for one strict long-tail variation, negative phrase match for related query groups, and negative broad match for words you want to eliminate entirely, such as free, cheap, or dangerous. If you skip this distinction, you will either under-filter waste or overblock valuable searches.

When should you use negative exact match?

Use negative exact match when a specific search term has proven unhelpful, but nearby variants still deserve room to perform. This is the most conservative exclusion type.

Example: you sell B2B call tracking software. The query [call tracking jobs] spends $96 over 90 days with zero conversions. But other searches containing “call tracking” perform well, including “call tracking software” and “call tracking for agencies.” In that case, negative exact lets you remove the bad term without damaging the theme.

Decision rule:

• Use exact when the waste is specific.
• Use it when the base keyword is still commercially valuable.
• Prefer it when account learning is still in progress.

When should you use negative phrase match?

Use negative phrase match when a modifier consistently signals low value across multiple variants. Think jobs, salary, reviews, complaints, or competitor names if you have decided not to buy that traffic.

Example: a cybersecurity SaaS campaign sees the following queries over 60 days:

• “cybersecurity software jobs” — 18 clicks, $126, 0 conversions
• “cybersecurity analyst jobs” — 11 clicks, $71, 0 conversions
• “cybersecurity careers software company” — 7 clicks, $44, 0 conversions

Instead of adding three separate exact negatives, you can use the phrase "jobs" if you are confident employment intent is never valuable for that campaign. Phrase match cuts faster and keeps the structure simpler.

When should you use negative broad match?

Use negative broad match for universal exclusions that should almost never appear anywhere in the account. Search Engine Land 2026 specifically cites examples like free, cheap, and dangerous, while Deloitte 2021 also recommends filtering phrases such as free and cheap.

This is the strongest setting. It suits terms that conflict with your offer at a structural level.

Example: a premium legal-tech platform starts at €999/month. Over one quarter, searches containing cheap produce:

• 94 clicks
• $564 spend at $6 CPC
• 0 demo requests
• 1:42 average time on site versus 3:58 for high-intent traffic

A broad negative on cheap likely makes sense here because the modifier itself signals a mismatch with a premium pricing model.

Match type comparison you can actually use

The contrarian point: broad negatives are not “better”

Teams often feel more in control when they block aggressively. That feeling is misleading. Negative broad match is powerful, but it also creates the highest risk of overblocking. If you sell a product with a free trial, a broad negative on free could cut a valuable slice of demand. If you offer migration support, a broad negative on template might remove searches that indicate active evaluation rather than low intent.

When uncertainty is high, start tighter. It is easier to widen an exclusion later than to recover the conversions you never saw because you cut too much too soon. Once match type is clear, the next issue is structure: where should those exclusions actually live?

Build negatives at the right level

A good negative keyword list can still perform badly if it sits in the wrong place. Karooya 2025 recommends a layered strategy: account-level negatives for broad exclusions, campaign-level negatives for campaign-specific noise, and ad-group-level negatives for fine adjustment. That is not just neat account architecture. It prevents accidental overlap and protects campaign intent. It also matters more now because Google’s 2025 Performance Max update is rolling out campaign-level negative keywords to all advertisers, which confirms this is no niche tactic. It is a platform-level control.

What should go at account level?

Put universal exclusions at account level. These are terms that should almost never trigger ads anywhere in your account.

Typical examples:

• jobs
• careers
• salary
• free downloads
• support if you are trying to stop existing-customer queries entering acquisition campaigns
• manual pdf if you only want commercial demand

This is the first named framework we recommend: the Layered Negative Structure. It is simple and hard to misuse. Use account-level negatives for universal exclusions, campaign-level negatives for campaign-specific noise, and ad-group-level negatives for tight intent control where themes might overlap.

Worked example:

A SaaS account has three campaign clusters:

• Brand
• Non-brand product keywords
• Competitor terms

At account level, they add jobs, careers, salary, and free pdf because none of those fit any acquisition objective. Over a month, those four exclusions block 310 impressions, 47 clicks, and roughly $423 in spend. Not huge on their own, but they also stop irrelevant demand from contaminating every campaign equally.

What should stay at campaign level?

Use campaign-level negatives when the traffic is unwanted in one campaign but useful in another. This is where structure protects intent.

Example:

• Your brand campaign should not trigger on competitor names.
• Your competitor campaign may intentionally target those same names.
• Your feature campaign for “landing page personalization” may not want pricing queries if those belong in a separate bottom-funnel campaign.

This is especially useful with Performance Max. Google’s 2025 blog post says campaign-level negatives in Performance Max let advertisers exclude specific queries for brand suitability or other strategic reasons. The same update also added a source column in search term insights and a usefulness indicator for search themes. That matters because it gives teams more visibility into where irrelevant demand is coming from and whether a theme is truly incremental.

How do you avoid overblocking in Performance Max?

Treat Performance Max like a campaign that needs rules, not blind trust. Google’s own update signals that query control is becoming more important, not less.

A practical process:

• Start with a short list of non-negotiable exclusions at campaign level.
• Review the source column in search term insights to see whether poor queries stem from keywordless targeting or added search themes.
• Remove or reduce low-utility search themes before adding broader negatives.
• Use campaign-level negatives for clear exclusions that conflict with the offer.

The edge case is new-market exploration. If your Performance Max campaign is trying to discover demand in adjacent categories, heavy exclusions can limit reach too early. In that case, cap budget and watch search term sources closely before escalating exclusions.

Ad-group control still matters in mature accounts

Ad-group negatives are less glamorous, but they still matter where you need tight segmentation. For instance, if one ad group targets enterprise analytics software and another targets marketing analytics software, negatives can prevent overlap so each query sees the most relevant ad and landing page.

This matters even more when you care about page-message fit. If different ad groups map to different landing pages, tight negative structure protects page relevance and improves the quality of any testing programme, including experiments around landing page best practices or AI-assisted page experiences for Google Ads.

Structure alone does not create good negatives, though. You need a repeatable way to find them, which brings us to the search term review habit.

Build a search term review habit

Negative keyword strategy fails when it depends on memory or occasional panic. PPC.co’s 2024 guide recommends using Google’s search term report to identify irrelevant queries and convert them into negative keyword lists. That sounds basic because it is basic. The issue is not knowledge. The issue is discipline.

How often should you review search terms?

The right cadence depends on volume. High-spend accounts need tighter loops than low-volume B2B campaigns.

A sensible baseline:

• 2-3 times per week for high-volume ecommerce or lead gen
• Weekly for most mid-market SaaS accounts
• Biweekly for lower-volume enterprise campaigns with long sales cycles

The key is not frequency alone. It is consistency and thresholds. If your team only opens the report when CPA spikes, you are already late.

What queries should you add immediately?

Some terms deserve instant action because they are structurally misaligned with the offer. PPC.co gives a practical example of a residential HVAC company adding “trucks,” “cars,” and “water heater” when targeting “heater repair.” The principle generalises well.

Queries to block quickly often include:

• Jobs/careers intent
• Support/login if you want to protect acquisition budgets
• Irrelevant product categories
• Price modifiers that conflict with your offer, such as cheap for premium products
• Educational intent like definition, meaning, or what is when you only want demo-ready traffic

Example:

A B2B PPC campaign spends $9,400 in a month. The search term report shows:

• “software engineer jobs analytics” — 14 clicks, $119, 0 conversions
• “analytics software free download” — 19 clicks, $161, 0 conversions
• “analytics dashboard template” — 27 clicks, $216, 0 conversions
• “analytics software login” — 11 clicks, $94, 0 conversions

Those four patterns alone account for $590, or 6.3% of spend. They also signal four different negative classes: employment, free intent, template intent, and existing-user support intent.

A review workflow we would actually use

This is the operational version, not the theoretical one.

1. Pull the search term report for the last 30, 60, and 90 days.
2. Sort by cost, then clicks, then impressions.
3. Tag queries into buckets: irrelevant, low-value, unclear, good.
4. Check whether a bad term is isolated or appears as a wider pattern.
5. Apply the narrowest negative match type that solves the problem.
6. Add recurring patterns to shared negative lists where appropriate.

PPC.co 2024 also notes the manual process inside Google Ads: go to Keywords, then Negative keywords, then click the plus button to upload or add a list. The mechanics are easy. The judgment is the hard part.

The edge case: some ugly queries still assist revenue

This is where rigid rules break. A query like “software comparison” may look weak if direct conversion rate is low. But if your category has a long evaluation cycle, comparison traffic may still influence pipeline and branded search later. Enterprise SaaS teams should be especially careful here. Low immediate conversion does not always equal low commercial value.

So yes, review search terms regularly. But regular review without a decision window can create overreaction. That is why the next piece matters more than most teams realise.

Use a 90-day decision window

One of the most useful recommendations in Search Engine Land’s 2026 strategy piece is the 90-day decision window. The article calls 90 days the balanced default for most paid search accounts when deciding whether to add a negative keyword. It also says 30 days is aggressive and suits short-cycle ecommerce or tightly capped budgets, while 365 days is more conservative for high-consideration B2B or long buying cycles. That is a far better rule than adding negatives every time a query “looks bad.”

When is a query just noise?

A query is often just noise when the sample is too small to support a confident decision. Three clicks and zero conversions tell you almost nothing. Twenty clicks at a high CPC with zero engagement tell you more. Ninety days gives enough room to see whether a pattern repeats.

Example:

Query A: “marketing analytics platform comparison”

• 30 days: 4 clicks, $36, 0 conversions
• 60 days: 9 clicks, $81, 1 assisted conversion
• 90 days: 16 clicks, $144, 2 demo requests

If you had blocked it after 30 days, you would have cut a term that simply needed more time.

When is it safe to block a term permanently?

It is safer when three conditions are true:

• The query has enough data over the chosen window.
• The term shows clear intent mismatch.
• The mismatch is structural, not temporary.

For example, if “free crm template” has produced 41 clicks, $246 spend, 0 conversions, and weak on-site engagement across a full 90-day period for a premium demo-only SaaS product, the evidence is strong. Add the negative.

A decision grid with numbers

Use this quick scoring logic:

The contrarian point: speed can damage learning

Advertisers like decisive action. But adding negatives too early can shrink the account’s ability to discover profitable variants. That is especially dangerous in B2B, where a query may convert rarely but still produce very high revenue when it does.

If your average deal size is $25,000, blocking a query after $180 of spend and zero direct conversions may be shortsighted. If your product sells for $29/month, the same threshold may be perfectly rational. This is why negative keyword decisions should reflect unit economics, not just search term aesthetics.

Once you have a decision window, you need a framework for deciding what counts as genuinely excludable. That is the most useful part of the whole process.

Use a simple exclusion framework

Most teams need a faster rule than “it depends,” but a safer rule than “block everything weird.” Our preferred model is the Intent-Value Filter. It combines three ideas supported by the source base: Deloitte 2021 recommends aligning business goals with the right keywords and target audience, Karooya 2025 emphasises blocking unwanted or low-value searches, and Search Engine Land 2026 stresses the cost of query-ad-page misalignment. Put together, they support a simple operating rule.

The Intent-Value Filter says: block queries that are irrelevant, commercially weak, or structurally impossible to convert. Keep ambiguous terms until the data proves they are bad.

Is the query irrelevant, low-value, or non-converting?

Start by sorting every suspicious query into one of three buckets.

• Irrelevant: the search has nothing to do with your product or buyer.
• Low-value: the topic is related, but the intent is weak for your business model.
• Structurally non-converting: the query conflicts with the offer itself.

Example scoring model:

• Irrelevant = Block immediately
• Low-value + poor economics over 90 days = Block or isolate
• Structurally non-converting = Block at the highest sensible level
• Ambiguous = Keep under review

Worked example with real numbers:

A B2B software campaign reviews these queries over 90 days:

This keeps the framework sharp without becoming reckless.

Does the query match the landing page and offer?

This is the second half of the filter and where many teams miss easy wins. Even if a query is commercially relevant, it may still be wrong for the current page. A term like “pricing” might underperform not because the search is bad, but because it lands on a generic feature page instead of a commercial page.

Ask two questions:

• Does the query align with the offer format? Demo, trial, download, or consultation.
• Does the query align with the landing page promise?

If not, do not rush to add a negative. First ask whether the page should change. This is one of the clearest intersections between keyword filtering and conversion work. Teams often try to solve a page problem with a query block, or a query problem with new page copy. You need to diagnose correctly.

A second framework for team use

The Query-Offer Fit Check is a simple companion to the Intent-Value Filter:

1. Can this searcher realistically buy what we sell?
2. Would our current ad and page make sense to them?
3. Do our unit economics justify testing this traffic longer?

Example:

A query like “best landing page software for startups” may underperform in a campaign promoting enterprise landing page optimization services. But the issue may be segment mismatch, not total irrelevance. Instead of blocking it everywhere, you might route it to a startup-specific campaign or a tailored page. That is a smarter move than broad exclusion.

This matters if your account also spans adjacent topics such as B2B PPC strategy, ad copy refinement, or broader acquisition messaging. Negative keywords should sharpen routing, not flatten demand.

The edge case: profitable exceptions deserve protection

Some terms break the rule in a good way. Competitor searches often have weak CTR and mixed lead quality, yet they can still produce high-value deals in the right account. Educational queries may convert poorly on a last-click basis but still feed retargeting and branded search efficiently.

That is why our contrarian take is simple: the best negative keyword strategy is not about blocking more terms faster; it is about being selective enough to protect good traffic while aggressively removing only the queries that will never pay you back. If your framework cannot distinguish between those two groups, it is too blunt.

Once you adopt that mindset, the larger strategic benefit becomes easier to see.

Why negative keywords stay underrated

The obvious benefit of negative keywords is lower wasted spend. But that is not the full story. Skai 2024 says negatives can reduce wasted spend, improve CTR, and enhance ad performance. Karooya 2025 adds the downstream effects: stronger ad relevance, better Quality Score, lower CPC, and improved ROAS. Those are not isolated outcomes. They are all signs of cleaner campaign signal.

What improves after you cut waste?

Four things usually improve first:

• CTR rises because fewer irrelevant users see your ads.
• CPC often softens as relevance strengthens.
• Conversion rate improves because more clicks come from aligned intent.
• Reporting quality improves because noise no longer masks genuine trends.

Example:

A lead gen account spending $30,000/month removes low-intent search classes worth 12% of spend. Over the next six weeks:

• CTR rises from 3.9% to 4.6%
• Average CPC falls from $9.20 to $8.50
• Landing page conversion rate rises from 5.1% to 6.0%
• CPA drops from $180 to $141.67

That is not magic. It is simply what happens when the platform stops buying as much irrelevant attention.

Why does this matter more as automation scales?

Because modern campaign automation can spend money faster than teams can manually inspect every query. Google’s 2025 Performance Max update is a useful signal here: it adds campaign-level negative keywords, a search terms source column, and a usefulness indicator for search themes. Google would not add those controls if query governance were no longer important.

The practical implication is clear. As bidding, matching, and campaign expansion become more automated, negative keywords become more important as a strategic boundary. Automation is good at finding volume. It is not automatically good at respecting your business model.

The enterprise angle most teams ignore

Skai’s 2024 guide notes that teams managing high query volume often use an enterprise paid search platform to centralize search term insights and negative keyword governance across campaigns. That is less about software preference and more about operating discipline. Once an account spans many campaigns, markets, and product lines, negative keyword drift becomes a governance problem.

The edge case is very small accounts. If you spend a few hundred euros per month on a tightly defined local campaign, enterprise-style governance is unnecessary. A lightweight review habit and a short negative list may be enough. But as soon as the account expands across geographies, products, or funnel stages, the cost of messy exclusions rises quickly.

A final worked example: the economics of cleaner signal

Consider two accounts with the same $15,000 monthly spend.

Account A has weak negative control:

• CPC: $7.50
• Clicks: 2,000
• Conversion rate: 3.5%
• Conversions: 70
• CPA: $214.29

Account B removes enough low-intent traffic to improve CTR and lower CPC by 8%, while conversion rate rises to 4.3%:

• CPC: $6.90
• Clicks: 2,174
• Conversion rate: 4.3%
• Conversions: 93.5
• CPA: $160.43

That is roughly 33.5% more conversions from the same spend. Not because bidding got cleverer. Because filtering got stricter where it needed to and more selective where it mattered.

Negative keywords remain underrated because they look unglamorous. No one boasts about them in a board deck. Yet they often determine whether the account is learning from real demand or just paying to inspect noise. The final question is how to make that discipline repeatable without building more manual work than your team can maintain.

Put query control on rails

If the problems in this article sound familiar, the fix is not another spreadsheet full of one-off exclusions. You need a system that watches search intent, surfaces recurring waste, and connects traffic quality back to landing page performance and revenue metrics. That is exactly where dynares.ai fits. Our platform helps teams identify low-value query patterns faster, connect those patterns to page-level conversion outcomes, and generate clearer optimization priorities across Google Ads, PPC landing pages, and broader acquisition workflows. Instead of manually bouncing between search term reports, page tests, and reporting dashboards, you can spot where poor-fit traffic is draining budget and where tighter routing will actually improve results. If you want to stop treating negative keywords as cleanup and start using them as a performance control, this is the moment to put that process on a firmer system.