Google Ads for ecommerce: you are paying for people who want to open a store, not shop at one!

Ecommerce PPC leaks in two directions at once.

The first leak is people on the wrong side of the counter. A large slice of search around online retail is people who want to build a store, find a supplier or get a job at one. They are researching a business, not buying a product, and your ads reach them anyway.

The second is quieter and costs more. Shopping and Performance Max spend against a product feed rather than a keyword list, so the usual advice about bidding barely applies. If the feed is wrong, or if brand traffic sits in the same campaign as everything else, your return on ad spend is a number that describes nothing.

AdPilot connects to the account and works through it the way a consultant would before quoting you a retainer. Searches that were never an order. Campaigns claiming credit for sales you had already won. Ads that buy traffic and no revenue. Each finding returns as a card, in plain English. You approve it or you skip it. Nothing reaches your account until you do.

And you stay in control the whole way:

• You approve every change. The AI proposes; you decide. Every budget, bid, keyword, or ad edit arrives as a plain-language card you approve or skip with one tap. Nothing is ever applied on its own.
• A snapshot before every change. AdPilot saves exactly how things looked before anything is touched, so you always have a clean “before.”
• One-click revert. Don’t like a change once it’s live? Undo it in a single click, straight back to that snapshot.
• A full audit log. Who, what, when, the old value and the new value, every change recorded. Nothing hidden.
• You keep the keys. You connect with one secure Google sign-in, and it only ever makes changes you approve. The account stays yours start to finish - you never transfer ownership, and you can disconnect anytime.

Where ecommerce ad budgets actually leak

Real US search volumes from July 2026. This is the universal layer, the waste that shows up in almost every store account regardless of what you sell.

Ecommerce is the one industry where a single list cannot do the whole job, and it would be dishonest to pretend otherwise. Your catalog has its own junk: the DIY version of your product, the used market for it, the brands people confuse you with. The method for finding those is below the list.

US monthly search volume, Ahrefs, July 2026. Around 199,000 searches a month before your category adds its own.

Negative keywords for ecommerce stores. Free, no email, no form.

Paste this into your account today whether or not you ever use AdPilot. It costs nothing and takes about ten minutes. Then spend another ten finding your category layer: sort your search terms report by spend, filter to zero conversions, and read the top fifty. The pattern will be obvious and it will be specific to you.

Discount hunters

coupon, coupon code, promo code, discount code, voucher, voucher code, free shipping code, clearance, outlet, liquidation

Wants to sell, not buy

dropshipping, wholesale, bulk, supplier, suppliers, distributor, manufacturer, private label, how to sell, how to start, reseller, warehouse jobs, job openings, careers

Making it themselves

how to make, diy, homemade, pattern, sewing pattern, tutorial, recipe, plans, free template, printable template, how to instructions

Second hand and rental

used, second hand, preowned, refurbished, thrift, rental, rent, for rent, ebay, craigslist, facebook marketplace, poshmark

Repair instead of replace

how to repair, repair guide, repair manual, repair tutorial, fix it myself, user manual pdf, instruction manual pdf, download manual, warranty claim, troubleshooting

Not buying today

images, pictures, wallpaper, meaning, history of, wikipedia, is it worth it reddit, free pdf, free printable, free download

Ecommerce PPC: five things that matter more than your bids

• The feed is the campaign. In Shopping, titles, product types and attributes do the work that keywords do in Search. A store with a tidy feed and clumsy bidding beats a store with clever bidding and a messy feed almost every time.
• Send the value, not just the sale. If every order reports as one conversion, bidding treats a $12 order and a $400 order as identical and dutifully chases the cheap one. Passing revenue with each conversion changes what the algorithm optimizes for.
• Separate brand from everything else. Brand searches convert beautifully because those people already decided. Left in the same campaign as prospecting, they lift the average and hide the fact that your acquisition is losing money.
• Put brand exclusions on Performance Max. Without them, PMax quietly absorbs the traffic you had already earned and reports it as its own success. It is the most common reason a PMax campaign looks like a miracle in month one.
• Let the fourth quarter be its own plan. Competition, click costs and conversion rates all move at once in Q4. A budget and a target ROAS set in August describe a market that no longer exists in November.

What one add to cart is worth to the store

Four numbers your analytics already has. This one is set up differently from the other industry pages on purpose: a store's click price is category-specific, so there is no single honest figure to pre-fill. Take your real average cost per click out of the account and put it in. The cart-to-order rate is the number stores usually DO know, which makes this arithmetic unusually solid here. Nothing is stored.

Everything a store account needs, in order

Everything below is what AdPilot's agent works through on a store account, in the order it works through it, so you can do the whole thing yourself with a spreadsheet and a Merchant Center login. We put it in public because a store owner who follows it gets an account that works whether or not we ever speak, and because you should be able to check a method before you trust software that runs it.

Two things make a store different from every other business on this site, and both are in here. Your margin decides whether ads are possible at all. Your product feed does the job keywords do everywhere else. Budget two days and give the feed one of them. After that it's an hour every second week for as long as the store trades, plus a proper sit-down before peak, because the numbers you set in the quiet months stop being true in November.

Sell a service instead of a product? Every other version of this build lives here

1. 1

   Start with your margin, because it decides whether ads are possible at all

   Before a keyword, before a campaign, work out what one order actually leaves you. Not revenue - what's left after the cost of the goods, the payment fee, the box, the pick and pack, and whatever you absorb on shipping. Call it contribution margin. A store running 60 percent can pay roughly three times as much for a customer as a store running 20 percent, and no amount of clever bidding closes that gap. It's the number that decides whether the rest of this page applies to you, and most owners have never written it down.

   Marlowe Coffee Roasters is a store that doesn't exist, and its numbers run the whole way down this page. Average order $42, contribution margin 45 percent, so $18.90 comes out of an order and everything else was spent before the ad ran. Then the click price, which is the one number nobody should hand you: what a click costs depends entirely on what you sell, so any figure printed on a page like this is worth nothing to your account. Pull yours out of the account itself, last 90 days, non-brand only. Marlowe's is $0.62. Yours will be different, and borrowing theirs will only mislead you.

   Then one division does more work than every tool in the account. Cost per click divided by conversion rate is what an order costs to buy. Marlowe converts 2.4 percent of its ad clicks, so $0.62 divided by that 2.4 percent is $25.83 to produce one order that leaves $18.90 behind. Call it 42 clicks a sale. That store loses $6.93 every time the ads work perfectly, and almost nobody's done the sum.

   There are two honest ways out and neither one is a bid adjustment. The customer comes back, or the order gets bigger. Marlowe's buyers average 2.6 orders in the first year, which turns one new customer into $49.14 of margin instead of $18.90. At that number $25.83 to acquire one is a good trade, provided the bank account can wait 90 days for the second order. If nobody reorders and the average order won't move, the honest answer is that paid search doesn't work for this catalog at this click price, and you deserve to hear that from a page trying to sell you software. One caveat decides whether that trade is real. Google counts every order the ads touch, including the second and third from a customer you already owned, and your brand campaign is where most of those land. Before you lean on $49.14, pull the new-versus-returning split out of your store platform. If half the ad-driven orders are people you were keeping anyway, $25.83 is buying you fewer customers than the report says it is.

   One more thing before you leave this step. Margin isn't one figure, it's a column with a row per product, and the product that sells most is very often the one that pays least. Marlowe's 12 oz house bag is the bestseller; the 2 lb bag and the subscription are what cover the rent. Sort your catalog by contribution per order rather than by units sold. That column decides which product groups get the bid in step three, and which ones you stop advertising altogether.

   • Contribution margin per order: revenue minus goods, payment fees, packaging, pick and pack, and the shipping you absorb.
   • Cost per click out of your own account, last 90 days, non-brand only. Never off somebody's blog.
   • Cost per order: cost per click divided by your conversion rate. Compare it to the margin, not to the order value.
   • Repeat rate and orders per customer in year one - that's what makes a thin first order survivable.
   • New versus returning on the ad-driven orders, out of your store platform. If most of them were already yours, the year-one number is flattering you.
   • Contribution by product, sorted. The bestseller is often the worst payer.
   • If cost per order beats year-one margin and nothing can move, stop here. That's a real answer, not a failure.

   Marlowe Coffee Roasters, all the way to the uncomfortable part

   • Average order $42, contribution margin 45% - $18.90 left per order.
   • Non-brand cost per click, taken out of their own account: $0.62. Not a benchmark. Yours is yours.
   • Conversion rate on ad traffic 2.4%, so about 42 clicks to buy one order.
   • $0.62 divided by 2.4% = $25.83 to produce an order worth $18.90. Every good day loses $6.93.
   • Break-even return on ad spend at 45% margin is 2.2x. They're running 1.6x.
   • Year one: 2.6 orders per customer, $49.14 of margin. So $25.83 buys a customer, not an order - as long as the order really came from a new one.
   • Target set at $26 an order. Every budget, bid and product decision below hangs off that one figure.
2. 2

   Sort the shoppers from the shopkeepers

   There are about 198,700 searches a month around online retail before your own category adds a thing, and 150,600 of them come from people who want to open a store rather than buy from one. Dropshipping alone is 145,000. Another 44,500 are hunting a discount code, usually for somebody else's shop. 3,600 want a job. None of them will check out, all of them click cheaply, and cheap is exactly why they arrive first and flatten the day's budget by lunchtime.

   Now the test that saves the most money, and it takes ten seconds a term. Read the keyword cold, with nobody standing next to you explaining what you sell, and ask whether it could match a search for something you don't stock. Bare “coffee” is cafes, jobs, a machine, a mug and a definition. Bare “espresso” is the drink. One-word category terms are never the seed of a small campaign, and no match type rescues them - phrase still carries every other meaning those words have. Add the word that makes the term yours: “buy espresso beans online”, “2 lb dark roast beans”, “espresso beans free shipping”.

   Buy words in a store account are easy to spot once you look: buy, order, online, a size, a weight, a color, a model number, a brand plus a product, free shipping. Compare words - best, top, versus, review - are real buyers on a longer path, and margin decides them for you. A 55 percent store can afford the extra clicks a comparison costs. A 20 percent store can't, and should let those searchers arrive through the free results instead.

   The list further up this page is the universal layer, and honestly it's half the job. Every catalog has its own junk and nobody can write that part for you. Three questions find it. What's the do-it-yourself version of your product - for a roaster, “how to roast coffee at home”. What's the used and repair market around it - “used espresso machine”, “grinder burr replacement”. Which businesses do people confuse you with - a coffee store collects “coffee shop near me” and “barista jobs” all day long. Write your answers down before you spend a cent, then let the search terms report correct you every second week.

   Split brand from everything else in your own head from the first day, because Google won't do it for you. There's no brand column on the report. Take your store name, take the misspellings people manage, and mark those rows yourself. Blended into one average, brand traffic makes acquisition look profitable when it isn't, and that's the most common way a store account looks healthy and isn't.

   • Read every term cold. Could it match something you don't stock? Then it isn't a keyword yet, it's a draft.
   • Buy words: buy, order, online, a size, a weight, a model number, a brand plus product, free shipping.
   • Wrong side of the counter, as patterns to read for rather than lines to paste: dropshipping, wholesale account, bulk supplier, private label, how to start a business, warehouse jobs. Turn each one into a phrase before it goes anywhere near the account.
   • Discount words are their own decision. A coupon hunter converts once and never comes back at full price.
   • Write your category layer down before you spend: the search that ends in a video tutorial, the search that ends on eBay, the search that ends at somebody else's counter. Then let the search terms report add to it every second week.
   • Mark your brand rows yourself, then report brand and non-brand as two numbers. Never one.

   The universal layer, and where a coffee store's own junk hides

   • dropshipping - 145,000/mo - wants to open a store, not shop at one. Block it.
   • promo code - 26,000/mo - hunting a discount, often to spend somewhere else.
   • coupon code - 12,000/mo - same crowd, and rarely a second order.
   • discount code - 6,500/mo - same again.
   • how to make money on amazon - 5,600/mo - wants to sell, not buy.
   • shopify jobs - 3,600/mo - looking for work.
   • Category layer, coffee edition: how to roast coffee at home, used espresso machine, coffee shop near me, barista jobs, coffee machine descaler. On no published list anywhere, and every one of them clicks.
   • Buyers in the same catalog: buy espresso beans online, 2 lb dark roast beans, espresso beans free shipping, decaf whole bean coffee. No volumes printed here on purpose - the ones that matter are in your category, and looking them up is your ten minutes.
3. 3

   The feed is your keyword list

   Shopping and Performance Max don't take keywords. There's no list to write and no bid to set per term. Google reads your product feed and decides which searches your products deserve, which means your titles, product types, attributes, prices and images are the keyword list. Most stores spend their attention on bids and leave the feed exactly as the theme exported it. That's backwards, and it's the biggest single difference between a store account and every other kind.

   A title isn't a name. “House Espresso” is what you call it in the warehouse; nobody types it. Put the words a buyer would use at the front and keep the brand at the end, unless people genuinely search for you by name. Roughly the first 70 characters are what a shopper sees, so front-load them. Do this for the twenty products that carry your margin - the ones off the column you sorted in step one - not for all nine hundred.

   Then the fields nobody enjoys. GTINs, because without them Google has a weaker idea of what the product even is and the comparison surfaces skip you. Price accuracy, because a feed price that disagrees with the page price gets the product disapproved, and the disapproval is silent: your campaign quietly serves a slice of the catalog and reads like a bidding problem. Availability, so you stop buying clicks on the seasonal blend that sold out in November. And a main image on a clean background with no sale badge burned into it, because a badge baked into the picture is a policy rejection, not a design choice.

   product_type describes the catalog; custom labels describe the money. Use product_type the way a shopper would sort your shelves, then put the margin column from step one into custom_label_0: margin_high, margin_thin, do_not_advertise. Those five custom label slots exist so you can bid on something the shopper never sees, and they're how you bid espresso up, push low-margin accessories down and stop advertising anything under about $12 where shipping swallows the whole contribution. Excluding products feels like switching off revenue. It's switching off the products that were paying you nothing to sell them.

   Before any of that matters, open Merchant Center and read the diagnostics. Disapproved items, products pending review, a feed that last fetched nine days ago. A store whose feed is half approved has a catalog problem being reported to it as an advertising problem, and no card, no consultant and no bid change fixes that first.

   One habit that pays every time: when a product launches or a seasonal line lands, the feed has to be right before the campaign is. A launch with a thin title and no GTIN spends the two weeks it'll never get back being shown to nobody in particular.

   • Titles front-loaded with the words buyers type. Brand at the end unless they search your name.
   • GTIN on everything that has one.
   • Feed price and page price identical today, sale prices included.
   • Availability accurate, so last winter's blend stops collecting clicks.
   • Main image on a plain background. No SALE badge burned in - that's a rejection, not a design choice.
   • product_type written the way a shopper would sort your shelves.
   • custom_label_0 carrying the margin column from step one: margin_high, margin_thin, do_not_advertise.
   • Merchant Center diagnostics read and disapprovals cleared before a dollar moves.

   Three product titles, before and after

   • Before: Marlowe House Espresso
   • After: Espresso Beans 2 lb - Dark Roast Whole Bean Coffee - Marlowe
   • Why: the words a buyer types are in front, the size sits in the title where it belongs, and the brand goes where a non-brand shopper can ignore it.
   • Before: Winter Blend (2026)
   • After: Christmas Coffee Gift Blend 12 oz - Whole Bean - Marlowe
   • Why: seasonal stock is only findable in the season's own words, and “(2026)” is a thing nobody has ever typed into Google.
   • Bundle: 3-Bag Espresso Sampler 36 oz - Whole Bean Coffee Gift Set - Marlowe
   • Subscription refill: Espresso Beans Subscription 12 oz - Ships Every 2 Weeks - Marlowe
   • Twenty of these beats a month of bid tinkering, and it takes an afternoon.

4. None of this is difficult. It is a long run of small decisions, and each one quietly costs money when it goes the wrong way. The agent makes them for you and shows you every single one before it touches the account.

   Start free - no credit cardFree for 7 days. No card, no call, cancel whenever.
5. 4

   Put the money where a store's money works: Shopping first

   For a store, Shopping is usually the workhorse, and Standard Shopping isn't dead however many times you've been told to migrate. It has no sunset date, and Google no longer hands Performance Max automatic priority over it for the same products - Ad Rank decides which one serves, and running both in one account is normal. That's the sort of rule Google moves without telling anybody, so check the current help page before you build a quarter on it. Either way, start here. A shopper who types a product and sees your product, your price and your photo has done most of the qualifying already.

   Performance Max earns its place later, and it comes with conditions. Conversion tracking that reports real revenue per order rather than a count. A feed that's fully approved. A budget of roughly three times what a sale costs you, on its own rather than shaved off the others. And then the one nobody mentions: brand exclusions. Without them PMax swallows people searching your store by name, traffic you'd already earned, and files it as its own triumph. That's why a new PMax campaign looks like a miracle in month one. Switch the exclusions on and the return on ad spend drops, hard. The lower number is the real one.

   Search still has two jobs in a store account. The phrases you have to own, where a text ad can say something a product tile can't - free shipping over $40, refills shipped every two weeks, a refund on any bag. And your own name, always in its own campaign, always separate.

   One rule governs how those campaigns share traffic, and it's the only real lever you've got over PMax. A query that exactly matches an exact-match keyword in your Search campaign goes to Search; PMax's own themes rank more like phrase and broad. So the moment a search term inside PMax or Shopping starts producing orders, add it to Search as exact. That's how you take a winner out of the box and give it its own ad and its own landing page.

   Now the arithmetic, out loud, before any campaign exists. A new campaign has to clear two floors at once: a daily budget of at least three times what one sale costs you, and enough clicks to be readable, call it ten a day at the very least. Marlowe's target is $26 an order, so three times that is $78 and Shopping opens at $80 a day. At $0.62 a click that's roughly 129 clicks and three orders a day at the 2.4 percent this account converts at - thirty conversions banked inside ten days, so the bidding is out of learning before your first proper read. What throws a month away is spreading that $80 over four campaigns to see which one works. Four starving campaigns learn nothing; one fed campaign learns.

   Then leave it alone. Google says a smart bidding strategy usually settles inside a week and can take two. On a catalog this size, go further: don't act on anything until two weeks have passed or 30 conversions have landed, whichever comes later. Week one is noise. Editing a target, switching bid strategy, pausing and restarting - each of those restarts Google's clock. Budget changes don't, so move budgets rather than targets when something has to move. Adding negatives off the search terms report is always safe, and in week one it's the only thing you should be doing.

   • Shopping first. A product tile with a price on it is the cheapest qualified click a store gets, and Standard Shopping has no sunset date whatever your last agency told you.
   • Each campaign funded at three times what one order costs you - per campaign, never spread across them.
   • Ten clicks a day minimum at your own cost per click, or a month of data says nothing.
   • Brand in its own campaign from day one.
   • Performance Max last: clean feed, real revenue reporting, brand exclusions switched on before the first impression, and a budget of its own.
   • Two weeks untouched. If something has to move, move a budget, not a target.

   Budget arithmetic for a store, done before the campaign exists

   • Target cost per order $26, straight out of step one. Three times that is $78, so Shopping opens at $80/day.
   • At $0.62 a click that's about 129 clicks a day, three orders a day, and 30 conversions banked inside ten days - well past the ~10 clicks a day below which a month of data says nothing.
   • Google may spend up to double the daily budget on one day, but never more than 30.4x it in a month. $80/day is a $2,432 ceiling, not a promise of exactly $80.
   • Non-brand Search costs more a click in this account but converts better: about $24 an order, floor $72, so it runs at $75/day.
   • Brand orders cost about $3 here, so the floor is $9 and $10/day clears it by a dollar. Brand is the one campaign where that's fine - you're defending traffic you already earned, not buying new.
   • Performance Max would need a fourth budget of roughly $80/day of its own, plus brand exclusions and a clean feed. That's why it's added last.
   • The month-killer: $80 split across four campaigns to test them. Thirty days later you've got four campaigns and no verdict.
6. 5

   Write the ad to the page it lands on

   A responsive search ad takes 3 to 15 headlines at 30 characters and 2 to 4 descriptions at 90. Count them yourself, spaces included, or Google will count for you at the worst possible moment. Aim for 10 to 12 headlines that are genuinely different, because a near-duplicate is worse than an empty slot: it burns a combination Google could have tested and teaches it nothing. Google bolds whatever matches what the shopper typed, wherever that headline lands, which is the real reason to write “espresso beans” in their words rather than the words on your packaging.

   The page decides the copy, so pick the page first. A category page is right when the ad group is a category - espresso beans, decaf coffee - because the shopper still has a choice to make and a category page is where choices get made. A product page is right for one thing with one search behind it: a launch, a bundle, a specific size. The homepage is right almost never, and “there's nowhere better to send them” isn't an exception - it's a page you haven't built yet, admitted sideways. Build it, then buy the clicks. Don't point an ad group at a single SKU that can sell out either, or you'll pay for clicks to an out-of-stock product before anyone thinks to tell you.

   Here's the part almost nobody does. Before you save the ad, find the three headlines that name the product least, read those three alone as a complete ad, and ask whether a stranger could say what's for sale. “Small Batch, Big Flavor”, “Freshness You Can Taste” and “Free Shipping Over $40” is a finished ad for gin, granola or hot sauce. Nothing in it is coffee, and you'd be paying full retail every time Google served that combination.

   One job per headline, and the job never repeats even though the product does. The search term, a size, the price, the roast, the subscription, the bundle, proof, the refund, the shipping threshold, one plain ask. What ruins a set is four headlines making the same freshness claim in four different outfits. Vary the length on purpose as well: if everything you write is 29 characters, Google often fits only two of them and you lose a third of the ad.

   Put a real number in one headline. A price in the ad text does a negative keyword's job for free - people who were never going to pay $17 for a bag of beans read it and don't click, and you don't pay for them. The shipping threshold works the same way from the other side: “free shipping over $40” nudges the order up and filters out the person who wanted a $6 sample.

   • Pick the landing page before you write a word. Category ad group, category page. One product with one search behind it, product page.
   • Ten to twelve headlines, no near-duplicates, every one counted to 30 characters with the spaces in.
   • Two descriptions minimum at 90 characters. Google won't save the ad with one.
   • One real number in the set: a price, a size, or the shipping threshold.
   • Lengths spread from about 16 to 30, or Google fits two headlines instead of three.
   • Read your three weakest together. If a stranger can't name the product, rewrite one.

   Headlines, counted by hand, with two that failed and their fixes

   • Buy Espresso Beans Online (25) - the search term, spelled the way they type it.
   • Dark Roast Beans (16) - short, and it still says what's in the bag.
   • Espresso Beans From $17 (23) - the price, once.
   • Espresso Beans, 12oz Or 2lb (27) - the size question, answered in the ad.
   • Coffee Refills Shipped Monthly (30) - the subscription, and the longest one in the set.
   • Espresso Beans, 2,300 Reviews (29) - proof, with a number in it.
   • WEAK: Small Batch, Big Flavor (23) - could be gin, granola or hot sauce.
   • FIXED: Small Batch Coffee Beans (24)
   • WEAK: Freshness You Can Taste (23) - freshness of what? Nothing in it is coffee.
   • FIXED: Coffee Roasted To Order (23)

7. Everything past this point is the work that never finishes: reading search terms, adding negatives, checking what the budget did overnight. That is exactly the part the agent takes off your hands.

   Start free - no credit cardFree for 7 days. No card, no call, cancel whenever.
8. 6

   The negative list you paste before the first impression

   Negatives go in before the campaign serves its first impression, not when the report starts looking bad. A store campaign without them buys home-roasting guides, machine repair videos and directions to a cafe from hour one. The biggest entry is the one nobody can write for you: everything you don't stock. Walk your own site, list what you sell, then list the neighboring things people will assume you sell. That second list is worth more than the first.

   Then the trap that makes people say negatives don't work. A negative keyword doesn't stretch the way a normal one does. No plurals, no synonyms. “Sample” doesn't block “samples”. “Cheap” leaves “inexpensive” running all month. Add both forms every time, then read the list back hunting for the words you only wrote once.

   Block phrases, not words, and in a store account breaking that rule costs real orders. Never bare “free”: “free shipping” is one of the strongest buying signals you get, so block “free samples” and “free download” and leave the word alone. Never bare “repair” or “fix” if a repair kit is a product on your site - block “repair video” and “how to repair”, not the word your customers use to find the part. Never bare “cheap”, because “cheap espresso beans” is somebody buying espresso beans. One row on the list higher up this page needs your judgment before you paste it: it blocks “used”, “second hand”, “preowned” and “refurbished” together, which is right for a store that sells new only and wrong the day refurbished becomes a product line. If you sell refurbished, pull those four out and block “used” only where it's genuinely somebody else's stock - “used espresso machine ebay”, not the bare word. And never your own brand name anywhere, except inside a campaign you deliberately decided should be brand-free.

   Standard Shopping takes negative keywords at campaign level exactly like Search does. It just has no positive keyword list, so negatives are the only word control you get there, which makes this list more important in Shopping, not less. Performance Max takes campaign-level negatives as well, with brand exclusions on top for your own name. Paste the list into all of them. Then check what's already blocked account-wide before you add anything, or you'll spend an afternoon rebuilding a list that's been running for two years.

   Pick the narrowest form that stops the bleeding. Negative broad blocks any search containing all your words in any order, which is blunt. Negative phrase blocks those words in that order, and it's what you want almost every time. Negative exact blocks one query and nothing else, which is right for a competitor's product name you'd otherwise catch by accident.

   • The list goes in before the first impression, not after the first bad report.
   • Everything you don't stock, written out from a walk through your own site.
   • Singular and plural of every term. Negatives don't stretch.
   • Two words minimum on every line. Read each one back and ask what paying search it also kills.
   • Account-level lists checked for what's already blocked before you add a thing, so Shopping and PMax get the same protection.
   • Brand exclusions on Performance Max. That's where your own name gets swallowed.

   A day-one list for a coffee store. Two words minimum on every line, with one deliberate exception

   • how to make coffee, how to make espresso, coffee recipe, coffee recipes, coffee to water ratio, cold brew ratio
   • coupon code, promo code, discount code, voucher code, free stuff, coffee giveaway, giveaway entry
   • dropshipping, wholesale account, bulk supplier, private label, sell coffee online, coffee business plan, how to start a coffee business, how to start a coffee shop, how to start a coffee roastery
   • coffee shop near me, cafe near me, barista jobs, barista job, coffee jobs, hiring baristas, now hiring
   • Things you don't stock, and this is always the biggest group: buy espresso machine, espresso machine deals, coffee maker reviews, drip machine, pod machine, k cups
   • Competitor products by their exact name, and only if you truly don't stock them: dunkin original blend, nespresso pods
   • The only single word on the whole list is “dropshipping”, because nobody who types it is buying a bag of beans. Every other line is two words or more - one word always has a second meaning, and the second meaning is usually a customer.
   • Careful: never bare “free” - “espresso beans free shipping” is a buyer. Never bare “repair” or “fix” - a grinder repair kit is a product. Never bare “used” if refurbished grinders are on the site. Never bare “cheap” - “cheap espresso beans” still wants beans. Never bare “sample”; block “free samples” instead.
   • Never block a brew method either. “French press”, “pour over”, “moka pot” and “cold brew” are how coffee shoppers describe the beans they want - “best coffee for french press” is a buyer with the bag already chosen, which is exactly why “cold brew ratio” is on the list and “cold brew” is not.
   • And never truncate. “How to start a coffee” would block “how to start a coffee subscription”, which is somebody trying to start one with you. Never your own name as a negative either, outside a campaign you meant to keep brand-free - that's how a store stops bidding on itself and hands the click to a reseller.
9. 7

   Send the revenue, then read the account every second week

   Most store accounts count orders and stop, which tells bidding that a $12 bag and a $400 gift hamper are the same event. It'll then go and find you more $12 bags, efficiently, forever. Send the actual order total with every conversion. In the conversion action that means picking “Use different values for each conversion” - the fallback value sitting under it is only used when your tag sends nothing at all. The setting that quietly flattens your revenue is “Use the same value for each conversion”: choose that and the $12 bag and the $400 hamper both report as whatever number you typed in.

   Two things to check before you call anybody in. Auto-tagging: with it off there's no click ID landing on your site, so Google Ads can't tie an order back to the click that produced it, and campaigns that are working look dead and get paused for looking dead. And second, the conversions Google already built out of your analytics and quietly left switched off. That one's a toggle, not a project, and it's been sitting there in more accounts than you'd think. Neither one needs a developer.

   First forty-eight hours, two questions and nothing else. Is money leaving, and is everything approved - and for a store that means the products as well as the ad. A campaign that can't serve looks exactly like a campaign that's warming up, and the silence gets blamed on your keywords. Day seven you open the search terms report, add negatives, and close it again. No targets, no opinions, no harvesting. The first week is noise, and a strategy still learning will punish you for touching it.

   Day fourteen to thirty is the real read, and a store's version has an extra column. Cost per order and return on ad spend, brand and non-brand kept apart, then the same numbers per product group - because the account average is usually hiding the fact that three product types are carrying the whole thing. Search terms again, this time for waste and for gold: anything that produced orders goes back in as an exact match keyword with its own ad. Harvesting converters out of Shopping and PMax into Search is the most useful hour you'll spend all month.

   Month two, each campaign gets one verdict: feed it, fix it, or switch it off. Then the decision only a store has to make. Whatever you settle on in August describes a market that's gone by November. Click costs, competition and conversion rates all move at once in the fourth quarter, so a target set in the quiet season will either strangle delivery or overspend in the weeks that pay for the year. Plan the peak as its own thing, with its own budget and its own target.

   One honest limit before you go. The search terms Google shows you never add up to the money you spent - low-volume queries sit below a privacy threshold, and Shopping shows you less than Search does. Act on the rows you've got, and never tell yourself you've accounted for every dollar.

   • Auto-tagging on. One switch, no downside.
   • Conversions imported from analytics and left disabled - switch them on.
   • Real order revenue on every purchase, set to different values per conversion rather than one flat number for all of them.
   • Brand and non-brand reported as two rows. Google won't split them for you - there's no brand column.
   • Return on ad spend per product group, not only per campaign.
   • A date in the diary every second week. The search terms arrive whether you read them or not.

   Two weeks of search terms in a coffee account, and what each row earns

   • “buy espresso beans online” - 22 clicks, 3 orders -> add as exact match, and give it its own ad.
   • “2 lb espresso beans” - 14 clicks, 2 orders -> exact match. Size terms convert because the decision is already made.
   • “monthly coffee subscription” - 9 clicks, 1 order -> exact match, pointed at the subscription page rather than the shop.
   • “how to roast coffee at home” - 31 clicks, 0 orders -> negative phrases “how to roast” and “roast at home”. Your day-one list blocked “how to make”; it never saw this wording.
   • “used espresso machine” - 18 clicks, 0 orders -> negative phrase “used espresso machine”, all three words. Not bare “used”, because the refurbished grinders are real products with real buyers, and not “espresso machine” on its own either - “best beans for espresso machine” is somebody buying beans.
   • “coffee machine descaler” - 12 clicks, 0 orders -> negative phrase “machine descaler”, because you don't stock it.
   • “free coffee samples” - 9 clicks, 0 orders -> block “free samples” and “free sample”. Not bare “free”: “espresso beans free shipping” converted twice in the same two weeks.
   • “coffee sample pack” - 11 clicks, 1 order -> leave it running. You sell a 3-bag sampler, so this is a buyer, and it's the whole reason bare “sample” never went on the list.
   • “starbucks pike place” - 7 clicks, 0 orders -> negative phrase, their exact product name. Their bag, their traffic.
   • Every one of those got past a list that already blocked “how to make”, “dropshipping”, “coupon code” and “barista jobs”. The list is never finished, and that's what the hour every second week is for.

One finished ad for the espresso beans group

This is one ad group - espresso beans, in the non-brand Search campaign - written for a coffee store that doesn't exist. Twelve headlines, counted by hand. The card draws three of them, because three is the most a shopper ever meets at once, and the other nine sit underneath as what they are: spare parts.

• You write twelve, a shopper sees three, and you don't choose which three.. That's the whole reason this ad is built the way it is. You're not shipping twelve headlines, you're shipping every trio they can form, and you pay full price for each one that serves. So the question is never whether your best headline is good. It's whether your worst three, read together by a stranger, still sell a bag of coffee.
• The weakest trio still says coffee out loud.. Pick the weak three deliberately, because any three you happen to sample will pass. Here the ones that name the product least are “Dark Roast Beans”, “Refund On Any Coffee Bag” and “3-Bag Coffee Bundle”. Read them cold as one ad and a stranger knows exactly what's for sale. Any set whose weakest trio could be read out at a farmers market stall selling something else entirely is unfinished, however good the other nine are.
• The search term goes in one or two headlines, not twelve.. Google bolds the words that match what somebody typed, wherever that headline lands, which is why “Buy Espresso Beans Online” is written in the shopper's words rather than the packaging's. But Google's own guidance asks for at least three headlines without the keyword in them, so the rest name the product differently - beans, coffee, dark roast, whole bean - instead of repeating the term twelve times.
• The $17 is doing a negative keyword's job.. A price in the ad text is the cheapest filter you own. Somebody who was never going to pay $17 for 12 oz of coffee reads it and moves on, and you don't pay for that. If your prices move too often to print, name the range or lead with the shipping threshold instead - but print something, because the alternative is paying to discover who your price scares off.
• “Free” earns its keep here, and nowhere on your negative list.. “Beans Ship Free Over $40” raises the average order and screens out the person hunting a $6 sample. It's also the exact reason bare “free” must never go on a negative list: block the word and you block the buying signal along with the freeloaders. Block “free samples”, keep “free shipping”.
• Two descriptions minimum, and a shopper only ever sees one.. Google won't save the ad with fewer than two and won't take more than four, then it chooses which one runs. The one above carries the roast date and the shipping threshold. Write a second that carries the money - the $17 and the refund - and a third only if it makes a claim the first two don't. Rewording the first in fresh clothes just hands Google a weaker twin to spend your budget on.
• This ad points at the espresso category page.. The ad group is a category, so the landing page is that category - the shopper still has a choice to make between roasts and sizes. A launch or a bundle gets its own ad group and its own product page. Sending this to the homepage would throw away every specific promise in the twelve headlines, and pointing it at one SKU means paying for clicks the week that SKU sells out.
• Lengths run 16 to 30 characters on purpose.. Google shows up to three headlines, but only when they fit the space it has. A set where everything is 29 or 30 characters often gets squeezed down to two, and you lose a third of your ad. “Dark Roast Beans” is 16 and it still names the product, which is the only rule a short headline has to obey.
• Your Shopping ads get none of this, and that's the point.. No headlines, no descriptions, no sitelinks. The feed writes those ads: the title, the price, the image, the review count. So if you've got an hour and a store, twenty rewritten product titles beat a twelfth headline every time. The text ad is for the phrases a product tile can't answer - the subscription, the shipping threshold, the refund.

The shape a store on a real budget actually runs

Three campaigns, listed in the order you build them: Shopping does the volume, your own name goes on a small budget of its own where it can't flatter anything, and a narrow Search campaign comes last for the phrases a product tile can't answer. Performance Max is deliberately absent, and the caption says exactly when to add it. Every budget below clears the same two floors: three times what an order costs, and enough clicks a day to be readable. All Search keywords start as phrase match, and exact match is earned later by the searches that produce orders.

Why it's shaped like this. Shopping goes first and takes the biggest budget because a store's cheapest qualified click is a product tile with a price on it, and Standard Shopping has no sunset date whatever your last agency told you. Your own name sits alone on a small budget, because brand clicks are cheap and convert far better, so blended in they lift the average and hide that acquisition is losing money underneath. The Search campaign is narrow on purpose - it exists for the two things a tile can't say, the subscription and the shipping threshold - and it's the last one you fund, not the first. Espresso and subscriptions are two ad groups rather than four because each one needs enough orders a month to teach the bidding anything; split a small account into six themes and all six go quiet. Performance Max is missing on purpose, and here's exactly when to add it: after the feed is fully approved, after real order revenue is reaching Google, with brand exclusions switched on before the first impression, and with a fourth budget of roughly $80/day of its own rather than a slice cut off the others. Add it up before you commit. $165 a day is about $5,016 a month, roughly 190 orders at a $26 target. If that's bigger than your whole ad budget, don't shave all three - run Shopping at $80 and brand at $10, about $2,736 a month between them, and leave Search until Shopping has paid for itself. Two fed campaigns beat three that are all just under the line.

PPC management for ecommerce, shown as cards you approve one at a time.

What you actually pay an ecommerce agency for is the hands-on part. Getting the feed and the campaign structure right, splitting brand from prospecting, digging through search terms for money leaks, writing ads that match the product someone typed. AdPilot’s AI does that work and shows what it based each call on.

Examples of what the cards look like - not results from your account

A bad week in Q4 does not wait for the monthly report.

You keep control, and you stop being dependent.

Most store owners never learn how their own Google Ads account works, which is awkward, because ad spend is usually their largest controllable cost. AdPilot is built the other way round. Every card tells you:

• What it changes - for example, pause a search term that spent $418 and returned $61.
• Why it helps - the reasoning, not just the instruction.
• What it is based on - the numbers in your own account.

Approve one and a snapshot is saved first. Then a before and after dashboard shows what happened to return on ad spend, cost per order and revenue, in plain words. After a month or two you know which levers actually move the margin.

AdPilot is new - and we’d rather earn your trust than fake it.

We could plaster this page with five-star quotes and a “trusted by 10,000 businesses” banner. We won’t - because AdPilot is new, and inventing proof would be lying to you. Instead, here’s what actually protects you:

• You risk nothing to try it. 7 days free, no credit card. If it isn’t useful, you walk away having spent nothing.
• Your account is never at risk. Nothing is ever applied without your one-tap approval, a snapshot is saved before every edit, every change has one-click revert, and a full audit log records it. The worst case is a suggestion you don’t like - so you just don’t approve it.
• Guardrails you can’t trip by accident. AdPilot never deletes campaigns and never touches shared budgets. Some doors simply don’t exist.
• Built by an operator, not a VC-funded black box. Software made to give small businesses the Google Ads work an agency charges a retainer for - without agency prices or the loss of control.

Judge us on the work, the math, and the safeguards - on your own numbers, not our adjectives.