Amazon PPC match types are usually taught as a neat ladder. Broad match gives reach. Phrase match gives a little more control. Exact match gives precision. Learn in broad, promote winners into exact, add negatives, repeat. That advice is not wrong. It is just too small for a brand owner managing advertising spend with real margin pressure.
The expensive part of match types is not the definition. It is the budget behaviour. Broad, phrase and exact do not only decide which customer searches can trigger your ad. They decide how quickly uncertainty is allowed to spend money. In a self-service account, that difference matters more than the textbook explanation.
The named mistake I see is treating match types as keyword settings instead of spend-risk classes. A team puts the same keyword into broad, phrase and exact campaigns, sets similar bids, gives each campaign a daily budget, and assumes the structure is mature. In reality, the account now has three different learning speeds, three different waste profiles and three different ways to hide margin leakage.
My stance: Amazon PPC match types need a budget firewall. Not just isolation. Not just negatives. A practical rule set that limits how much broad and phrase are allowed to learn, when exact is allowed to scale, and when advertising software must stop a query because the SKU cannot afford another lesson.
This guide is for brand owners managing Amazon Ads themselves, usually from around €1.5K monthly ad spend. The same operating logic also applies when you translate search demand into bol Sponsored Products, Walmart Connect or other retail media platforms. FiveX helps here by connecting ad performance to SKU margin, stock cover, product profitability and automation rules, so match types are judged by commercial permission instead of campaign neatness.
What current match-type advice gets right
The best competitor content explains the mechanics well. SellerSprite gives a useful plain-English breakdown: broad match creates maximum reach and maximum noise, phrase match shapes intent, and exact match gives the highest control for proven winners. Perpetua’s older Amazon PPC guides do a good job separating customer search terms from the keywords advertisers bid on. BidX correctly points out that campaign-level budgets make match-type separation useful, because Amazon can otherwise consume budget in one place before another match type gets a fair test.
Agency and software content from Eva, Amazoniac and ScaleA2Z also adds an important profitability layer. They recommend separating discovery campaigns from profit campaigns, isolating match types, using negative keywords and calculating break-even ACOS before scaling. That is sensible. If you currently run one manual campaign with broad, phrase and exact in the same ad group, fixing that structure will almost certainly improve clarity.
Reddit discussions show the operator problem behind the theory. Sellers ask whether they should use all three match types or only phrase and exact. Others describe broad campaigns swallowing budget, keywords competing with each other, and conflicting guru advice about “peel, stick and block” workflows. The pain is not that sellers do not know broad is broader than exact. The pain is that they do not know how much uncertainty their budget can safely buy.
That is the gap. Most advice answers, “Which match type should I use?” The better question is, “How much paid uncertainty can this SKU afford this week?”
The budget firewall: a better way to think about match types
A budget firewall gives each match type a commercial job and a spending boundary. It prevents discovery from stealing money from proven demand, but it also prevents exact match from receiving unlimited budget simply because it looks tidy.
Here is the simple version:
- Broad match is research spend. It should buy search-term information, not carry the account’s revenue target.
- Phrase match is validation spend. It should test shaped intent and close variations, not behave like a cheaper exact campaign.
- Exact match is execution spend. It should scale proven queries only when contribution margin, stock and incrementality still support the bid.
Notice the trade-off. If you make the firewall too tight, you starve discovery and the account stops learning. If you make it too loose, broad match becomes a very confident intern with the company credit card. Cute for about four hours. Expensive by Friday.
In FiveX, this is exactly where advertising data becomes more useful when it is connected to commerce data. A broad query that spends €18 without a sale may be fine for a high-margin hero SKU with 44 days of stock. The same €18 may be unacceptable for a low-margin variation with 9 days of stock and a 14% return rate. Match-type rules should not live in the ad platform alone. They need SKU economics beside them.
Example 1: KitchenEase and the broad-match learning cap
Imagine KitchenEase sells a stainless-steel garlic press on Amazon.de for €24.95. After referral fees, fulfilment, landed cost and expected returns, the product has €6.10 contribution margin before ads. Its break-even ACOS is roughly 24%. The team has a monthly Amazon Ads budget of €2,400 and wants to find new non-branded search terms.
The obvious setup is a broad campaign for “garlic press”, “stainless garlic crusher” and “kitchen garlic tool”. The suggested bids sit around €0.72. In the first week, broad match spends €138, generates €310 in attributed sales and shows a 44.5% ACOS. A normal weekly optimisation might lower the bid and add a few negatives.
The firewall view is sharper. At €24.95 selling price and €6.10 pre-ad contribution margin, KitchenEase can only afford about €6.10 of ad cost per order before profit turns negative. If conversion rate on broad traffic is 7%, the maximum CPC is about €0.43. A €0.72 bid means broad is buying learning at a price the SKU cannot repeat profitably.
The rule should be: broad match may spend a maximum of €45 per week on this SKU until it has either two profitable converting search terms or a conversion rate above 10% across at least 80 clicks. Any search term that spends €12 without a basket event is quarantined for review. Any term that converts twice below 24% ACOS is promoted to phrase or exact, but only if stock cover remains above 21 days.
That is a FiveX-style product hook in practice: ad software should calculate the learning cap from margin, not from yesterday’s campaign budget. It should also know whether the product has enough stock to deserve more discovery.
Example 2: NordicNest and phrase match that quietly becomes broad
NordicNest sells a €79.00 electric coffee grinder on Amazon.fr. The product has €19.50 contribution margin before ads, a 4.4-star rating and 37 days of stock. The team runs phrase match on “burr coffee grinder” because broad was too noisy and exact did not create enough volume.
For two weeks the phrase campaign looks acceptable: €420 spend, €1,620 attributed revenue, 25.9% ACOS. The target ACOS is 28%, so nobody panics. But the search-term report tells a different story. Three queries drive most of the revenue: “burr coffee grinder electric”, “coffee grinder for espresso”, and “quiet coffee grinder”. Two are profitable. The third converts, but it sells the cheaper black variant more often, where contribution margin is only €11.20 after a coupon.
This is the phrase-match trap. Phrase can feel controlled because the wording looks relevant. But if it routes demand into the wrong variation, the blended ACOS hides a SKU-level margin problem. The campaign is not wrong; the reporting layer is incomplete.
The firewall rule: phrase match can validate intent, but every material query needs a SKU-level profit check before it graduates. “Burr coffee grinder electric” can move to exact at a €0.95 bid because it converts at 13% and usually sells the silver variant with €19.50 margin. “Quiet coffee grinder” stays capped at €0.48 because it pulls the couponed black variant and breaks even only at a lower CPC. If the black variant stock drops below 14 days, the phrase target pauses even if campaign ACOS still looks good.
This is where FiveX’s product profitability and variation-level reporting matter. Amazon Ads can tell you the campaign converted. FiveX can help show whether the converted unit was actually the unit you wanted to fund.
Example 3: BrightBaby and exact match that deserves less budget, not more
BrightBaby sells a rechargeable night light on Amazon.com for $32.99. The exact keyword “baby night light rechargeable” looks like a winner: $890 spend, $4,180 attributed sales, 21.3% ACOS against a 30% target. The operator sees this and raises the exact campaign budget from $40 to $90 per day.
Then total account profit barely moves. Why? The keyword already ranks organically in the top three. A large share of exact-match clicks are capturing shoppers who would likely have bought anyway. At the same time, broad discovery is capped so tightly that the account is not finding new use cases such as “night light for breastfeeding” or “toddler travel night light”.
The named mistake here is assuming exact match always deserves the safest money. Exact match deserves controlled money. Sometimes the right move is to hold exact at a profitable but limited share of voice, then move the next $500 into measured discovery or Sponsored Brands Video.
A practical firewall rule: if an exact keyword has strong organic rank, stable conversion and rising CPC, do not scale budget automatically. Run a small incrementality check. For seven days, reduce the exact bid by 20% during non-peak hours and watch total keyword-level sales, organic rank and TACOS. If ad sales fall but total sales hold, the keyword was over-collecting credit. If total sales drop materially, restore the bid. FiveX supports this operating model by putting TACOS, organic context, ad spend and SKU margin in the same decision view instead of letting exact-match ROAS win the argument alone.
How to set match-type budgets from €1.5K monthly spend
At €1.5K monthly ad spend, you cannot give every campaign a generous learning budget. You need a clean default that protects the account but still creates new information.
A useful starting split is:
- 55-65% execution budget: exact match, proven ASIN targets and brand defence with known contribution margin.
- 20-30% validation budget: phrase match and tightly themed product targeting that can graduate into exact.
- 10-15% research budget: broad match, auto targeting and exploratory ASIN tests.
This is not a universal law. A launch account may need more research. A mature account with thin margins may need more execution. The point is to make the trade-off explicit. Discovery is valuable, but it is not allowed to quietly become the account’s biggest line item just because broad match finds impressions faster.
Set a weekly learning allowance per SKU. For a product with €8 pre-ad contribution margin, 10% conversion rate and €1.5K monthly spend across a small catalogue, broad match might only receive €20-€35 per week until it proves a query. For a hero SKU with €22 margin, strong stock and a review advantage, €75-€120 per week may be reasonable. The number should come from margin and confidence, not from the campaign name.
The rules your advertising software should enforce
A match-type budget firewall becomes powerful when it is automated. Not because the software is magically smarter than the operator, but because the operator should not manually babysit every search term at midnight.
Use these rules as a practical starting point:
- Research cap: broad and auto campaigns cannot spend more than a fixed percentage of the SKU’s weekly ad allowance unless they produce profitable search terms.
- Evidence threshold: a search term needs a minimum number of clicks, orders or contribution-margin-positive sales before promotion.
- Promotion with context: moving a term to exact also copies the SKU margin target, max CPC, stock rule and campaign role.
- Negative quarantine: do not instantly block every weak term. Quarantine it when data is thin, block it when spend and intent clearly fail.
- Variation check: if a query sells a lower-margin variation, the bid ceiling follows the sold SKU, not the advertised parent.
- Stock brake: phrase and broad pause automatically when stock cover drops below the threshold you set, even if ACOS is attractive.
- TACOS override: exact match does not receive unlimited budget when total sales are flat and ad-attributed sales are simply taking credit.
These are the kinds of guardrails FiveX is built for: connect marketplace ads, profitability dashboards, inventory insights and AI recommendations, then turn them into operating rules a brand owner can actually use.
A simple weekly operating cadence
Once the firewall is in place, the weekly routine becomes calmer.
On Monday, review broad and auto research spend. Ask: what did we learn, what did it cost, and which SKUs are no longer allowed to keep learning? On Tuesday, review phrase validation. Which queries have enough evidence to become exact? Which queries are relevant but too expensive for the SKU they actually sell? On Wednesday, review exact execution. Which targets deserve more budget, which are over-attributing existing demand, and which need a lower bid because CPC inflation has eaten the margin?
On Thursday, check inventory and pricing changes. A keyword that was safe last week may become unsafe after a coupon, fee change, stockout risk or Buy Box issue. On Friday, update the rules before the weekend, because Amazon has a special talent for spending enthusiastically while humans are enjoying dinner.
The operator voice matters here. Do not let the account become a museum of clever structure. If a campaign cannot answer what job it has, what margin it protects and when it must stop, it is not a structure. It is decoration.
The takeaway
Broad, phrase and exact are not just match types. They are spend-risk classes. Broad buys uncertainty. Phrase validates intent. Exact executes proven demand. All three can be useful. All three can waste money when they are allowed to spend without SKU-level permission.
The better self-service setup is a match-type budget firewall: research caps for broad, validation gates for phrase, execution rules for exact, and profit guardrails across all of them. That is how brand owners can keep learning without letting Amazon’s auction turn every lesson into a margin leak.
FiveX helps by connecting the pieces that Amazon Ads keeps separate: ad spend, search terms, SKU profitability, stock cover, pricing changes, TACOS and automation rules. So instead of asking “broad, phrase or exact?”, your team can ask the question that actually protects profit: “how much uncertainty is this SKU allowed to buy today?”