.png)
.svg)
Introduction
We have audited accounts where the Search Term Report had not been touched in six weeks. Not because the account was performing well. Because nobody had a documented process for what to do with it. Meanwhile the STR was sitting there showing a $4,200 spend on search terms that had never converted once.
The exact match portfolio had not grown in two months. The negation list had not been updated in three. And the account was spending more every week on terms it already knew did not work.
Across the 50+ Amazon accounts we manage at Olifant Digital, the weekly harvest cycle is the single most consistent compounding factor in long-term efficiency. Skip it and the account flatlines. Run it correctly and TACoS bends downward week after week.
Now, let’s get right into how to structure keyword harvesting and negation on Amazon.
What Keyword Harvesting Means in a Managed Account
Amazon keyword harvesting is the process of pulling converting search terms from your Search Term Report and promoting them into exact match campaigns where they can be bid and tracked on their own.
That is the definition. Here is why it matters week to week.
Your broad match and auto campaigns are designed to cast wide. Most of the searches they surface will never convert... but some will, and often ones you would never have thought to target yourself. The harvest cycle finds those terms, promotes them, and negates them from the source campaign so your budget stops splitting across two campaigns chasing the same traffic.
Do that every week and the account gets more efficient with every cycle. Skip a week and the waste compounds just as fast in the other direction.
The Foundation: Why the 1-1-1-1 Structure Makes Harvesting Possible
The harvest cycle only works cleanly if the structure underneath it is built correctly and that structure is the 1-1-1-1 method.
The reason behind the one campaign per ASIN, per match type, per ad type, per targeting group framework is that when you're dealing with multiple keywords in one ad group, the conversion data always blends together. Because of this you can't really tell which keywords actually drove sales.
So when you sit down to run the harvest cycle, the data you need to make a promotion decision is already too muddied to act on.
In a 1-1-1-1 structure every keyword tells its own story. So when a term in a broad match campaign hits four to five conversions at or below target TACoS within a 30-day window, you know exactly which keyword drove them.
Then the decision is simple because now you can either promote it or give it its own exact match campaign. And from that point forward every conversion in that campaign is clean data you can actually act on.
The structure is the prerequisite. The cycle is what compounds from there.
The Weekly Cycle: Step by Step
We run the harvest cycle on a rolling daily basis so by the time we pull the Search Term Report there is always a full seven days of conversion data ready to act on, weekends included. The cycle runs in the same order every time and the order matters.
We pull the report on the same day because from that we can compare the same window and the data is never stale and nothing slips through the gaps between sessions.
Step 1 — Pull and Prepare the Search Term Report
We head into Campaign Manager then go to Reports then create a Sponsored Products Search Term Report. The date range gets set to the last 7 days to match the weekly cycle window.
We add a column for TACoS contribution per search term, total ad spend on that term divided by total account revenue, then sort by spend descending. We then filter down to broad match and auto campaigns only, because those are the ones casting wide enough to surface new converting terms. Exact match campaigns are already controlled so there is nothing to harvest from them. We start where the money is going first.
Step 2 — Identify Harvest Candidates
Four to five conversions within the 30-day window at or below our TACoS target per ASIN is the threshold we work from.
TACoS, meaning total advertising cost of sale which measures ad spend against total revenue including organic is the lens we use because it captures the full picture of what a term is contributing to the account.
One conversion does not qualify. A single data point could be coincidence and promoting on that is how exact match portfolios fill up with terms that never compound into anything.
The flip side of this filter is worth running at the same time. Any term that has accumulated spend equal to or above your average order value without generating a single conversion is a negation candidate, not a harvest candidate. We flag these in the same STR pull by sorting for spend above AOV with zero orders. They go straight to the negation list.
Specificity is the second filter we apply.
A term like "kitchen organizer" converting twice is still not a candidate because in exact match it would pull the same broad traffic without the precision that makes exact match worth using. We look for terms specific enough that isolating them actually improves the quality of traffic coming through.
Strategically important head keywords get treated differently. A term driving ranking on a high-volume search may justify promotion even when direct efficiency sits slightly above TACoS target, because the organic ranking it builds reduces TACoS at account level over time. We treat that as a deliberate investment decision, not a threshold exception.
Step 3 — Promote to Exact Match
We handle promotion through bulk operations rather than the campaign builder. In Campaign Manager we go to Bulk Operations then download the template then build out the new exact match campaigns in 1-1-1-1 structure: one campaign per ASIN, per match type, per ad type, per targeting group. Then we upload the file and verify the following day that everything went live correctly.
We use bulk operations because it is more accurate. Clicking through individual campaign creation screens is where settings get missed.
A structured file leaves a record and removes that margin for error.
Step 4 — Add Negatives at the Right Level
Once a term gets promoted to exact match it gets negated from the source broad match campaign at ad group level immediately. Without that negative both campaigns bid on the same search simultaneously, CPCs inflate and conversion data splits between two places making neither readable.
Negative targeting runs on a slightly longer review cadence than the harvest itself though. We review every two weeks rather than every week because search terms need enough time to accumulate statistically meaningful data before we make a call to exclude them. Negating too early based on a handful of impressions risks cutting off terms that would have converted with more volume behind them.
When we do negate, the term gets added at ad group level, not campaign level. Campaign-level negatives sweep across every ad group in that campaign which in most cases is broader than what we want. And we use exact match type negative, not phrase, so the exclusion is surgical rather than catching variations we may still want traffic from.
Step 5 — Prevent Internal Cannibalization
As the exact match portfolio grows week over week the same term can end up in multiple campaigns through different promotion paths.
Two exact match campaigns targeting the same keyword creates the exact same problem Step 4 just solved, just deeper in the account where it is harder to spot.
We run the cannibalization scan weekly through Olifant AI, our proprietary Amazon management platform. Senior Amazon specialists with 7+ years of experience review the platform's output every week and make the promotion, pause, and negation calls before the structure drifts.
We pull all active exact match keywords across the account then sort alphabetically then scan for duplicates. When we find one we pause the lower-performing campaign then add that term as a negative in the one we are keeping. The portfolio compounds better when the structure stays clean. Left unchecked it compounds the wrong way.
The Negation Decision Framework
Negation is where most accounts make a mistake and it's not because they negate the wrong terms. The mistake they make is that they negate at the wrong level and this means either the exclusion sweeps too wide and kills traffic you still wanted, or it is too narrow and the problem keeps running underneath it.
There are two levels to work with. Campaign and ad group. Knowing which one to use and when is what separates a clean account from one that slowly bleeds efficiency in a place nobody is looking.
When to Negate at Campaign Level vs Ad Group Level
Campaign-level negatives apply to every ad group inside that campaign. Ad group-level negatives apply only to the specific ad group you add them to.
We default to ad group level for almost everything in the harvest cycle. When we negate a promoted term from its source broad match campaign, we want that exclusion contained to the ad group where the traffic was coming from, not swept across every ad group in the campaign. The difference matters more as campaigns grow and ad groups multiply.
Campaign-level negation has its place though. Irrelevant terms that have nothing to do with the product and never will, think searches that landed on a gardening product from someone clearly searching for something unrelated, those get negated at campaign level because there is no scenario where any ad group in that campaign should be showing for them.
The rule we follow is that surgical exclusions go at the ad group level, broad categorical exclusions go at the campaign level.
One technical limit worth knowing before building out your negation list: negative phrase match keywords are capped at 4 words maximum. Negative exact match keywords allow up to 10 words. Anything above those limits will not save and the exclusion will not apply, which is easy to miss when adding multi-word terms through the campaign builder.
Paused vs Archived Keywords Still Count — The Limit Trap
Amazon's negative keyword limit sits at 1,000 per ad group and paused and archived keywords count toward that limit the same as active ones.
What this means is that an account that has been running for a year and has been adding negatives inconsistently can hit that ceiling without realising it. Due to this new negatives simply do not save, and the account keeps spending on terms it has already decided to exclude.
The way we audit for this is through a bulk operations download. We pull the full keyword list then filter for negative keywords then filter again for paused and archived status. Anything that has been sitting inactive for more than 90 days and has no strategic reason to be kept gets removed. It frees up limited headroom and keeps the negative list clean enough to actually manage.
If this has never been done on an account, the first audit usually surfaces more inactive negatives than anyone expects.
Branded Term Negation and Defense Campaign Structure
Branded search terms need to be handled deliberately, not left to default campaign settings.
In our broad match and auto campaigns we negate branded terms. The reason is that branded search traffic, meaning someone who already knows the brand and is searching for it by name, should be captured by a dedicated brand defense campaign running exact match. If branded terms are left open in broad and auto campaigns they compete with the defense campaign for the same traffic, which inflates CPCs on the searches most likely to convert.
The defense campaign runs separately with its own budget. It is not shared with generic prospecting campaigns because the two serve completely different purposes and mixing the budgets makes both harder to evaluate. Branded traffic converts at a higher rate than cold traffic so it deserves its own structure, its own bid logic, and its own performance targets.
What Happens If You Skip a Week
Missing one cycle is not a disaster. Miss two in a row though and then you're dealing with an account that starts drifting in a direction that takes longer to correct. It basically creates what we call harvesting gaps and you lose the ability to scale high-converting terms while failing to negate wasted spend at the same time.
And while that is happening your broad match and auto campaigns keep running regardless. They keep surfacing new search terms, some of those terms convert, and without the weekly pull none of that gets captured. Those terms stay buried in campaigns bidding at broad match with no isolation and no TACoS data to evaluate them against.
On top of that your wasted spend keeps accumulating. Terms that should have been negated two weeks ago are still pulling budget, the exact match portfolio stops growing, and the account flatlines at whatever efficiency level it was at when the cycle stopped.
Then there is a data problem that most people only discover too late. Amazon only retains Search Term Report data for 60 days, which means a two-week gap in your harvest cycle is not just two missed weeks of optimization. If those converting terms are not captured before the 60-day cutoff they are gone permanently.
The TACoS is usually where you first notice something is off. Instead of a declining TACoS alongside growing revenue, you start seeing the ratio hold flat or creep upward. More spend going in, organic velocity not building, and your account becoming increasingly reliant on paid traffic just to hold revenue steady.
While one week missed is recoverable, missing two weeks or more becomes a structural problem.
The Compounding Effect Over 3 to 6 Months
before the numbers do.
The weekly cycle does not feel like much in the first few weeks. And that is fine, because what you are doing in those early cycles is not optimization yet. You are building the foundation that optimization compounds from.
Week one is where the structure takes shape. Your first STR pull surfaces the obvious converting terms, the ones performing in broad match that clearly belong in their own exact match campaigns. You promote them, negate them from source, and for the first time the account has clean keyword-level data to work with.
Week four is when the first real signal emerges. The exact match campaigns promoted in week one now have enough conversion history to tell you whether the promotion decision was right. Some terms compound quickly. Others plateau and need their bids adjusted downward. The negation list is growing too, and broad match campaigns are becoming noticeably more efficient as irrelevant spend gets cut each cycle
By month two the exact match portfolio has a meaningful conversion history behind it. Amazon's algorithm has started rewarding those campaigns with stronger organic placement for the terms they are built around because consistent conversion velocity is exactly what the algorithm responds to. Along with that your TACoS starts moving, not dramatically but the direction changes.
Month three is where it becomes visible in the numbers. The organic revenue share inside your total revenue has grown. Paid dependency has reduced. You are spending the same or close to it but the account is generating more revenue overall because organic is doing work that was not happening before. TACoS reflects that shift directly.
And then by month six, if the cycle has run consistently without gaps, the exact match portfolio is a completely different structure from what you started with. Terms that were converting in broad match eighteen months ago are now driving organic ranking for head keywords you could not have afforded to dominate on paid alone.
MatchaBar is the clearest example we have of this.
Within 90 days of running the full weekly cycle consistently the organic revenue share had grown to the point where paid dependency reduced on its own. Not because we cut spend. Because the cycle built the organic foundation that eventually carried the account. That contributed directly to $114,305 added in monthly Amazon revenue.
The cycle did that. Balanced Tiger followed the same trajectory on a different timeline with the same structural outcome.
The compounding only works if the cycle runs. Every week without a gap, in the same order, with the same criteria applied every time.
Frequently Asked Questions
How often should I run a keyword harvest on Amazon?
Weekly for any account spending above $3,000 per month in total ad spend. New accounts in the first 30 days need daily STR checks because conversion data is thin and irrelevant terms need faster negation to protect launch budget. Below $3,000 per month, bi-weekly works.
How many conversions does a search term need before I promote it to exact match?
Four to five conversions at or below TACoS target within a 30-day window. Fewer than that is not enough data to act on. The exception is high AOV products above $80, where two to three conversions significantly below TACoS target can justify early promotion.
What is the difference between harvesting from auto campaigns vs broad match campaigns?
Auto campaigns cast the widest net and surface searches Amazon associates with your product semantically, including terms you would never have targeted yourself. Broad match campaigns surface variations of keywords already in your strategy. Both get mined weekly but for different reasons. Auto is where unexpected winners come from. Broad is where phrase and exact variations get refined.
Can I have the same keyword in both a broad match and exact match campaign simultaneously?
Yes, but only before the negation step. Once a term is promoted to exact match it gets negated from the source broad match campaign at ad group level. Without that negative both campaigns bid on the same search simultaneously, CPCs inflate, and conversion data splits between two places making neither readable.
Should I harvest keywords from Sponsored Brands and Sponsored Display in the same cycle?
Sponsored Brands has its own STR so we run a separate harvest cycle for SB campaigns with different promotion thresholds, because SB converts at a lower direct rate but contributes to brand search lift. Sponsored Display uses audience and ASIN targeting, not keywords, so there is no STR to harvest from.
How do I handle search terms that convert well but have CPCs that push above my TACoS target?
Three options. Accept the higher TACoS on that term as a strategic investment in ranking velocity if it is a head keyword worth owning organically. Promote to exact match at a bid below market CPC and accept reduced impression share while still capturing some of that traffic profitably. Or leave it in broad and monitor whether organic ranking improvements reduce the CPC required over time. The right call depends on how strategically important that keyword is for organic positioning.
Get Your Free Amazon PPC Plan
If your account does not have a documented weekly harvest cycle, or you are running one but the compounding effect has not materialised, we will show you exactly where the cycle is breaking down. We pull your Search Term Report, review your exact match portfolio and negation structure, and give you a clear picture of what needs fixing.
Olifant Digital is a full-service Amazon agency and Amazon PPC agency for established brands, managing $100M+ in annual client revenue across 50+ Amazon accounts. Every engagement is backed by a 60-day money-back guarantee — if we do not improve your Amazon results, you do not pay.
Get Your Free Amazon PPC Plan →
Alex is the founder and CEO of Olifant Digital, where his team manages over $100M in annual Amazon client revenue across 50+ brands, and he runs a 7-figure Amazon brand of his own. That operator background shapes how the agency works: every tactic is tested with his own money before it reaches a client account. He oversees PPC methodology, creative, and conversion rate across all client accounts to make sure Olifant Digital scales brands profitably.
Mike reviews every Amazon article on this blog for strategic and technical accuracy before it publishes. As Director of Amazon Growth at Olifant Digital, he sets marketing strategy across client accounts and personally audits PPC at every stage of growth. He brings 8 years of daily Amazon operations across 7 and 8-figure brands including Beauty by Earth, Ekster, and Bullstrap, the kind of hands-on depth most agency directors delegate away.
.svg)
.svg)
