Ecommerce PPC KPIs: The 5 Numbers That Actually Matter

Five numbers decide whether an ecom PPC account makes money. Profit per order. What a new customer costs you. How much you spent on products that made no sales. The conversion rate on the page your ad points at. And how much of your budget sits on the products that already sell.

That is the whole list. ROAS sits next to it as a signal and a bidding input. Never as the scoreboard. It counts revenue, and revenue is not what you keep.

Open most ecom dashboards and the top row is none of that. Impressions, clicks, click-through rate, impression share. Four numbers that move every single day. Not one of them can pay an invoice.

Which Ecommerce PPC KPIs Actually Matter?

Disclosure before you read any further: ZenoX Media runs Google Ads for ecom brands. We get paid to run the accounts these numbers measure. Weigh it that way.

The top two move fast, so they reward a frequent look. The bottom three sit at product level, where one day of data is mostly noise. Read those on a longer window.

KPI 1: Profit per Order, Not Revenue per Order

What it is. What you keep on an order once everything that order cost you comes off. Build it in two parts.

Part one is what the order is worth before ads. Order value minus product cost, shipping, payment fees and returns.

Part two is the ad cost that sits on that order. This part needs care. No tool ties one order to the exact ad spend that won it. Clicks that never bought still cost money, and some buyers touch several ads first. So you use an average: ad spend in the window divided by orders in the same window. That gives you an average cost per order, not the true cost of one order. Good enough to steer by, as long as you match the two sides. Google Ads spend goes with the orders Google Ads got credit for. Total ad spend across every channel goes with total orders. Mix the two and the number lies to you.

Part one minus part two is your profit per order. It sits above overhead and tax, so it is not your bottom line. It is the number that says whether the order was worth having.

Revenue per order flatters a thin-margin catalog because a discount hits the two halves unevenly. Cut the price and the revenue on that order drops with it. Product cost does not drop. Shipping does not drop. So what you keep falls by more than the price cut did, while average order value barely twitches.

Where you find it. Not in Google Ads by default. You have to hand Google what your products cost you first. There is a slot for that: the cost of goods sold attribute in the product data specification. Fill it in Merchant Center, send cart data, and Google reports gross profit back to you.

The ad half stays yours. Google cannot take ad spend off a single order, so that column lives in your own sheet. No cost column anywhere yet? Export your Shopify orders and build one this week.

What a bad reading looks like. Revenue per order barely moves. Profit per order keeps sliding. Product mix is the first thing I would check. Cheap, thin products taking a bigger share of the orders, with a healthy basket size hiding it. Three other things move that number the same way: supplier cost going up, more returns, or a discount that ran longer than planned. Rule those out before you touch the account.

The one move. Stop bidding the catalog as one pile. Split it by margin band so thin products stop borrowing budget from fat ones. Why profit beats revenue as a target has that argument in full, with the furniture catalog that proves it.

KPI 2: What You Pay for a New Customer

What it is. Ad spend divided by first-time buyers. Not all buyers. First-time ones.

The gap between that and your blended cost per order is bigger than most owners expect. Divide monthly spend by every order and you get the blended figure. Divide the same spend by first-time buyers only and it jumps. If half your orders come from people who bought before, a new customer costs double the blended number. If only a third of orders are new buyers, triple. Same spend, three very different stories.

What makes that number good or bad. On its own, nothing. A new customer cost with no ceiling next to it is just a number going up and down. Set the ceiling yourself, with two questions.

What do you keep on a first order, after product cost, shipping, fees and returns? And how long will you wait to get the rest back?

The strict version: the profit on that first order has to cover what the buyer cost you. Every new customer then pays for itself on day one, and growth funds itself. The loose version: you let the first order lose money and earn it back on repeat orders. That one needs three things. Your repeat rate. What a customer is worth over a year. And the cash to carry the gap until it comes back. Do not know your repeat rate? Run the strict version until you do.

Pick one, write the ceiling on the wall, and judge every month against it.

Where you find it. Shopify tags every order as first-time or returning in the customers reports. Pull that against spend for the same window. Google Ads can do it too, through a new customer acquisition goal.

The two will not match. Shopify reads its own customer records. Google only counts the buyers it can tie back to an ad, and when those signals are missing it cannot make the link at all. Pick one source, write down which one, and never compare the two.

What a bad reading looks like. Blended cost holds steady, maybe even improves. First-time buyer count doesn't move, or it drops. Brand searches and retargeting clicks are cheap and they convert well, so they can drag the blended average down while new demand stalls. Seasonality and a sale to your existing list do the same thing, so check the mix before you call it. The account feels healthy the whole way through either way.

The one move. Get brand traffic out of the bucket before you judge anything. Use brand exclusions and read non-brand campaigns on new customers only.

This matters most in Performance Max, where you do not pick keywords the way you do in Search. You are not blind there though. Google's own list of PMax controls covers search themes, brand exclusions and negative keywords. You also get placement exclusions, plus channel and search terms reporting to read it all back. Product level is still the biggest lever, because that is where you decide what gets shown at all. The PMax spend leak checklist covers that setting and five more.

KPI 3: How Much Spend Went to Products That Made No Sales

What it is. Money that bought clicks on products with zero tracked sales value inside the window you are looking at. Not products that could never sell. Products that brought back nothing in those 30 days.

We measured this instead of guessing. We pulled 106,400 products across 137 stores.

Fashion products with spend and no tracked sales

71.7%

Fashion ad spend those products ate

14.7%

Home decor products with spend and no tracked sales

80.5%

Home decor ad spend those products ate

29.5%

Two niches, one measurement, and the waste share is twice as big in one as the other. Those are pooled averages across many stores, so treat them as the pattern, not as a forecast for your account. Yours could be half that. It could be worse. Nothing inside Google Ads cuts these products for you either way. Somebody has to look.

Where you find it. The Products tab. Set a 30-day window. Sort by cost. Filter to zero conversion value. Add up the cost column.

Use conversion value, not conversion count. A newsletter signup or another non-sale action can make a dead product look alive if you count conversions. Sales value is the same cut we used in the study, so your number and ours mean the same thing. And read it at product level, never campaign level. A campaign average can look survivable while most of the products underneath it bring back nothing at all.

What a bad reading looks like. The share climbing month over month. That means more of your budget is drifting somewhere that gives nothing back.

Flat is not a pass though. If a quarter of your spend does nothing this month, flat only means it will do nothing again next month. Judge the size first, then the direction.

The one move. A monthly cut, same day every month. Split the list in two first. Products with a data problem get fixed, not cut: missing identifiers, a broken title, out of stock. The rest, real traffic and no sales, get excluded.

Give a product more than one window before you pull it. One bad month can be a fluke. Leave new products still ramping, out-of-season items, and variants that support a winner. Pick a window that fits how fast your catalog turns over.

KPI 4: Conversion Rate on the Page the Ad Points At

What it is. Of everyone who lands on that exact product page from an ad, how many buy. Page level, not site level.

Where you find it. The landing pages report in Google Ads, or your analytics landing-page view filtered to paid traffic. Then line each page up against the spend it took.

What a bad reading looks like. A site average that reads fine while one high-spend page sits far below the rest of your catalog. Averages are very good at hiding a broken page.

We publish no benchmark conversion rate, and I would be careful with anyone who does. A cheap impulse buy and a big-ticket item do not convert at the same rate. A brand people already know does not convert like one they have never seen. Your own catalog median is the only fair comparison you have.

The one move. Read the funnel before you touch the page. The drop-off point tells you roughly who owns the problem.

Lots of add-to-carts and few checkouts usually points past the page. Checkout friction, shipping cost, delivery time. That is normally site-wide, so a landing page edit will not move it. The Baymard Institute keeps a running list of independent cart abandonment studies worth reading first.

Few add-to-carts points more at the page itself: price, photos, size information, stock. Fix the biggest spender first, one page at a time, or you will not know what worked.

KPI 5: How Much of Your Budget Sits on Proven Winners

What it is. The share of spend landing on the products that actually produce revenue. Ecom catalogs are far more top-heavy than the old 80/20 line suggests. In the same 106,400-product study, the top 1% of fashion products drove 45.2% of all Google Ads revenue. The top 10% drove 85.6%.

One percent of the catalog does almost half the revenue. Treating every product the same is fighting your own data.

Where you find it. Same Products tab. Sort by revenue. Mark the top 10%. Then check what share of total cost those products got. Most of the revenue on a small slice of the spend? That is where the work is being done. The rest of the budget is the question.

What a bad reading looks like. Revenue concentrated and spend spread flat. The winners are quietly paying for the tail.

The one move. Give the proven products their own campaign and their own budget so they stop being averaged with everything that never sells. It costs nothing in extra media. One condition: splitting also splits your conversion data, and Smart Bidding gets worse with less to learn from. Only split a group off once it brings steady sales of its own week after week. Below that, keep one pot and steer with product-level exclusions instead.

Which Number Do You Check First When Two Go Bad?

1. Profit per order down? Stop there. Nothing else matters yet. Go find what changed: mix, supplier cost, returns, or a discount.
2. Cost per order up, ROAS flat? Your average order value moved. Check product mix before you touch a bid.
3. ROAS up, new customer count flat? Brand traffic is the usual suspect. Repeat buyers and retargeting look the same on this reading. Split brand out and read it again before you decide.
4. Conversion rate down, CPC steady? Look at the page and the checkout first. Then rule out stock, a price change, a device shift, or a change in which searches you are matching.
5. Everything soft at once? Check the dead-product pile first. It is the cheapest thing to fix.

One habit worth keeping: change one thing at a time. Move bids and landing pages in the same week and you lose the read. Next month you are guessing again.

So Where Does ROAS Fit In?

ROAS updates fast and Smart Bidding takes it as an input. It just cannot see your costs.

Three things make it lie. Product mix is the sneaky one: cheap items take a bigger share of orders, revenue holds, profit drops, and ROAS never blinks. Brand traffic inflates it with sales you already had coming. And attribution keeps rewriting the same week's number for days afterwards. Conversions from the last 24 to 48 hours are still landing, so today's ROAS is not final. Judging a change on yesterday's number is how good campaigns get killed.

So work out your own floor. Break-even ROAS is 1 divided by your gross margin. Take your real margin after product cost and shipping, do that one division, write the answer on the wall. We do not publish ours and you should not borrow anyone else's. A margin you did not measure gives you a target that is not yours. When a target return bid strategy is the right call covers the bidding side.

One more thing if you run Microsoft or Meta next to Google. Two platforms will claim the same order, so per-channel ROAS gets shakier with every channel you add. Judge new customer cost on total spend against total first-time buyers. Use the per-channel numbers to decide where the next chunk of budget goes, not whether last month worked.

Which PPC Metrics Can You Ignore in Ecommerce?

If most of your budget sits in Performance Max, this list gets harsher. PMax does not give you keyword-level bidding, and the reporting is rolled up across several surfaces at once. So click-through rate and cost per click there are a blend, and a blend is hard to act on. Product level is where you steer with the most precision.

Ignore these as goals. Each still has one honest job.

• Impressions. Useless as a health metric. Good for one thing: a sudden drop to near zero flags a feed problem or a policy hit before revenue moves.
• Clicks. Useless alone. Pair with cost and you get cost per click, which tells you whether the auction got pricier or your volume dropped.
• Click-through rate. A bad goal on Shopping. Its honest job is to send you looking. A low rate on a product with plenty of impressions has a long list of possible causes. Price, image, title, missing reviews. A promo a rival is running. Where you sit on the page. Matching to searches that were never for your product. It flags. It does not name the cause.
• Impression share. Mostly noise. The one reading worth having is impression share lost to budget. That tells you the budget is the cap. It says nothing about whether raising it pays. Check what the campaign returns first.
• Quality score. Search only. It does nothing for Shopping or PMax. A very low score on a keyword you spend real money on means the ad and the page do not match the search.

Check these after one of the five has already gone bad. Never before.

Who Should Not Hire an Agency for This?

Plenty of stores. Run your own numbers before you talk to anybody.

Start with your own dead-product spend from the Products tab. Do not use ours. The 14.7% and 29.5% in the study are pooled averages across many stores, and a pooled average is not a forecast for one account. Yours is the only figure worth doing maths on.

Then be honest about both sides of it. Nobody claws back all of that spend. And cutting dead products is one job out of many. Weigh a fee against the whole scope: the feed, account structure, bidding, conversion tracking, the hours it takes off your plate. One recovered number on its own is a bad test in either direction.

Doing it yourself is genuinely small: three reports and a spreadsheet with your product costs in it. A freelancer for a few hours a month is the next rung, and for a lot of stores that is the honest answer. Our free Skool community teaches the same playbook and costs nothing. We are the agency, so weigh that. How we work and what we charge is published, and on a small account the answer is still no.

Frequently Asked Questions

Which ecommerce PPC KPIs should I track daily?

Two of them move fast enough to be worth a frequent look. Profit per order, and what a first-time buyer costs you. Both swing with your product mix. The other three sit at product level: dead-product spend, page conversion rate, and how much of your budget sits on proven winners. One day of product-level data is too thin to mean anything. Read those over a week or a month, and make your cuts on the longer window.

Is ROAS a good KPI for ecommerce PPC?

It earns its place. Smart Bidding bids straight off it, and it updates fast. Just do not trust the last day or two. Conversions from the last 24 to 48 hours are still landing, so a fresh ROAS reads low and keeps moving as attribution catches up. It also only counts revenue, and revenue is not what you keep. ROAS can sit perfectly still while your profit drops, because it cannot see what your products cost you. Work out your own break-even instead: 1 divided by your gross margin. That number is yours, and no article can hand it to you.

How much ecommerce ad spend goes to products that never sell?

We measured it. Our study covered 106,400 products across 137 stores. In fashion, 71.7% of products with ad spend made no tracked sales in their 30-day window. Those products ate 14.7% of fashion ad spend. Home decor was worse: 80.5% of products, and 29.5% of spend. Those are pooled averages across many stores. So check your own. Google Ads Products tab, 30-day window, filter to zero conversion value, add up the cost column.

What is a good conversion rate for an ecommerce PPC landing page?

There is no single number, and any article handing you one range for all of ecommerce is guessing. A cheap impulse buy and a big-ticket item do not convert at the same rate. Neither do a brand people know and a brand they have never heard of. Compare each page against the median of your own catalog. The page to fix first is the one sitting well under your own median while taking a big slice of the spend.

Which PPC metrics can I ignore in ecommerce?

Impressions, clicks, click-through rate, impression share and quality score, as goals. None of them tell you whether you made money. Each has one honest job as a diagnostic. Impressions near zero flag a feed or policy problem. Impression share lost to budget tells you the budget is the cap, and nothing about whether spending more would pay. A low click-through rate on a product with plenty of impressions is a prompt to look at price, image, title, reviews and which searches you match. Read them after something else has gone wrong, never as the scoreboard.