Ecommerce Profit Analytics: Profit Per Product, Step by Step

Ecommerce profit analytics is measuring what each product, page and customer leaves you after every variable cost, instead of what it brings in. You take revenue, subtract the landed cost of the product, shipping, payment and platform fees, and the ad spend it took to win the order. What's left is contribution margin. Divide it by the sessions that product page got, and you have profit per session.

That one number tells you which product to push, which to reprice, and which to quietly stop advertising.

Everything else in this guide is how to get that number right, week after week, without a month-end spreadsheet marathon.

Here's the confession. For years the first number I checked every morning was revenue. It's the number Shopify puts at the top of the screen, so it felt like the scoreboard. It isn't. Revenue is the crowd noise. Contribution is the score.

Revenue is what the store brings in. Contribution is what it lets you keep. Only one of them pays for inventory in October.

This page is the hub for the Profit & Analytics side of the blog. It walks the whole system in order: the metrics, the margin math, profit per product and per page, payback, blended ad efficiency, tracking, tariffs, a weekly dashboard, and cash planning for the fourth quarter. Every example that isn't real client data is labeled as hypothetical, and every number recomputes.

Revenue vs profit: the numbers that matter

Here's what founders say when you ask how they track profit. In one Shopify Community thread, a merchant described the problem as revenue looking fine while nobody knows which orders, products or channels were profitable until month-end. Another said margins quietly slip until the drop shows up in the bank balance.

That's the cage. Revenue climbs, the ad account looks fine, and the cash doesn't match.

The fix starts with putting four numbers side by side and knowing which question each one answers.

Metric	Formula	The question it answers	What it hides
Revenue	Orders x average order value	How much came in?	Every cost. A product can grow revenue and lose money on each order.
Contribution margin	Revenue minus landed product cost, shipping, payment and platform fees, returns and discounts, and ad spend	What did these orders leave me before fixed costs?	Fixed costs like salaries, software and rent.
Profit per product (per session)	Product contribution divided by that product page's sessions	Which product earns the most for each visitor I send it?	Customer value after the first order.
Marketing efficiency ratio	Total revenue divided by total marketing spend	Is all my marketing, blended, bringing back enough revenue?	Which campaign or product is doing the work. Needs a break-even line to mean anything.

Read that table from top to bottom and you'll see a pattern. Each row fixes the blind spot of the row above it. Revenue hides costs. Contribution hides which product did the work. Profit per product hides what happens after the first order. The marketing efficiency ratio hides which lever caused it.

So you need all four. But you check them in a particular order.

And underneath all four sits the number I care about most on the selling side: revenue per visitor. It's conversion rate times average order value. A page converting 2.0% of visitors at a $120 average order value earns $2.40 per visitor, which is $24,000 on 10,000 visitors. If you've never calculated yours, start with how to calculate Shopify revenue per visitor step by step, then read what revenue per visitor is and why it beats conversion rate.

Revenue per visitor tells you how well the page sells. Profit per session tells you whether selling it was worth it. You want both on the same screen.

A store can raise conversion rate, raise revenue, and still lose money. The page got better at selling a product that was never profitable to advertise.

That trap is the whole argument of why your Shopify conversion rate is a vanity metric. Conversion rate alone rewards whatever sells easiest, including a discount that gives away the margin. The companion piece, revenue per visitor vs conversion rate, shows why the two numbers send you to opposite fixes.

Contribution margin, step by step

Contribution margin is the money an order leaves you after every cost that grows with that order. Here's the order of subtraction:

1. Start with the price the customer paid, after discounts.
2. Subtract landed product cost (unit cost, freight, duty).
3. Subtract shipping to the customer and packaging.
4. Subtract payment processing and platform fees.
5. Subtract the ad spend it took to win the order.

What's left is contribution per order.

Gross margin stops after step 2. That's why gross margin makes almost every product look healthy. The damage lives in steps 3 through 5.

Here's the best worked example I have, and it's real. Three of our clients moved products from Amazon to their own Shopify stores. The full fee breakdowns are on our results page. Real client numbers, not typical results, and not a promise of what your store will do.

The first product is an animal repellent. Here's the table, exactly as it appears on the results page.

Line	Amazon FBA	Shopify
Product price	$17.95	$24.95
Landed costs	-$4.50	-$4.50
Shipping fees	-$4.24	-$6.30
Platform fees	-$2.69 (15%)	-$0.75 (3%)
Gross profit per sale	$6.52	$13.40
Advertising costs	-$2.69	-$26.36
Average order value	$17.95	$101.38
Net profit per sale	$3.83	$48.27

Walk the Amazon column with a calculator. $17.95 minus $4.50 minus $4.24 minus $2.69 is $6.52. Minus $2.69 in ads leaves $3.83. That's contribution margin in its purest form: a $17.95 sale that leaves $3.83 after paying for ads and inventory. About 21 cents of every dollar.

Now read the bottom of the Shopify column. The advertising line is almost ten times bigger, $26.36 against $2.69. A founder staring at ad spend alone would panic. But the average order value is $101.38, because on their own store the brand could sell more than one unit per order, and the net profit per sale is $48.27, after paying for ads and inventory.

In the client's own words, instead of making $17 per order, they started banking a $48 profit per order.

The same pattern holds on the other two products.

Product	Amazon net profit per sale	Shopify net profit per sale	Shopify average order value
Animal repellent	$3.83	$48.27	$101.38
Rodent product	$7.73	$78.87	$177.75
Third animal product	$17.58	$134.28	$210.86

All three are after paying for ads and inventory. And here's the lesson the table teaches better than any definition could.

The channel with the smaller ad bill was the one making less money. Cost lines don't tell you where the profit is. The bottom line does.

A founder who managed by ad spend would have kept everything on Amazon. A founder who managed by contribution per order moved the winners. If you sell on Amazon, the full fee story lives on why Amazon sellers are moving their winners to Shopify, and our Shopify vs Amazon revenue per visitor study covers the per-visitor side.

One caution before you copy this into a spreadsheet. Contribution margin only works if the costs are the real ones for that order. Averages hide the product that ships in a heavy box, the one that gets returned twice as often, and the one that only sells with a discount code. Which brings us to the product level.

Profit per product and per page

Store-level contribution tells you whether the business works. Product-level contribution tells you where to put the next dollar. And profit per session is the fairest way to compare products, because it accounts for how much traffic each one needed.

Run the math on a store like this, a hypothetical kitchen brand with three products and 10,000 product page sessions a month.

The cutting board. 5,000 sessions, 3.2% conversion rate, $62 average order value. That's 160 orders and $9,920 in revenue. Revenue per visitor: $1.98. Per order, it costs $16 landed, $14 to ship (it's heavy), $1.86 in 3% fees, and $24 in ads. Contribution per order: $6.14. Total: $982.40. Profit per session: about $0.20.

The chef's knife. 4,000 sessions, 2.0% conversion rate, $120 average order value. That's 80 orders and $9,600 in revenue. Revenue per visitor: $2.40. Per order: $38 landed, $9 shipping, $3.60 in fees, $40 in ads. Contribution per order: $29.40. Total: $2,352. Profit per session: about $0.59.

The sharpener. 1,000 sessions, 2.5% conversion rate, $45 average order value. That's 25 orders and $1,125 in revenue. Revenue per visitor: $1.13. Per order: $9 landed, $5 shipping, $1.35 in fees, $8 in ads. Contribution per order: $21.65. Total: $541.25. Profit per session: about $0.54.

Now look at the ranking by revenue.

Cutting board $9,920

Chef's knife $9,600

Sharpener $1,125

Sessions x conversion rate x average order value.

The cutting board wins. It converts best and brings in the most money. If you run the store by the Shopify revenue report, it's your hero, and it gets the biggest ad budget.

Now rank the same three products by contribution for every 1,000 sessions.

Chef's knife $588

Sharpener $541

Cutting board $196

Product contribution after landed cost, shipping, fees and ads, divided by sessions, x 1,000.

The order flips. The best-selling product is the worst place to send a visitor. The small sharpener, which nobody on the team talks about, earns almost as much per session as the knife.

Here's the thing. Nothing about the cutting board page is broken. It converts at 3.2%. The problem is heavy shipping and a $24 ad cost on a $62 order. You could spend a month testing headlines on that page and never touch the leak.

The best-selling product and the most profitable product are often different products. Your ad budget should know which is which.

So what do you do with this?

• Push the knife. Each extra session there is worth about three times a cutting board session.
• Test the sharpener with more traffic. It earns $21.65 per order on only 1,000 sessions. Watch whether its conversion rate and ad cost hold as spend grows.
• Fix the cutting board's economics, not its copy. Raise the average order value with a board-and-oil bundle, renegotiate shipping, or move its ad budget elsewhere. Our breakdown of Shopify bundle strategy mistakes that kill margin covers how to build that bundle without giving the margin back.

Profit per page works the same way. Landing pages, collection pages and the homepage each get sessions and each produce orders, so each has a revenue per visitor and a profit per session. If a collection page earns $0.15 per session and the hero product page earns $0.59, sending paid traffic to the collection page is a decision to earn less.

And this is where the selling side and the profit side meet. The biggest lever on profit per session is usually the page itself, because conversion rate and average order value both sit at the top of the math. Our flagship bedding client went from a conversion rate of 1.0% and a $125 average order value to 3.5% and $231. Revenue per visitor went from $1.25 to $8.10. On 10,000 visitors, that's $12,500 before and $81,000 after. The case study numbers are here, and the method behind them is in how to increase revenue per visitor on Shopify.

Want to see how a lift like that moves your own numbers? Here's the math from our homepage: 1% conversion at a $50 average order value is $0.50 per visitor, or $5,000 on 10,000 visitors. At 3% and $80, it's $2.40 per visitor, or $24,000. Plug in your own.

Then take each product's revenue per visitor and multiply it by that product's contribution margin percentage. In the kitchen example, the knife keeps 24.5% of revenue after all variable costs, the sharpener 48.1%, the cutting board 9.9%. Revenue per visitor times that percentage is profit per session. It's the same number, reached from the selling side.

Before you move on, read the $60K store that was actually losing money. It's the story version of everything in this section.

Acquisition cost, payback and lifetime value

Profit per session looks at the first order. But most stores don't live on first orders. They live on the second and third.

That's where customer acquisition cost and lifetime value come in.

Customer acquisition cost is what you spend on marketing to win one new customer. Take the ad spend aimed at new buyers for a month and divide it by the new customers you got that month. Don't divide by all orders, because returning customers dilute the number and make acquisition look cheaper than it is.

Lifetime value is how much contribution a customer leaves you over time. Use contribution, never revenue. A customer who spends $400 with you over two years at a 10% contribution margin is worth $40, not $400.

Payback is how long it takes a group of customers to earn back what you spent to acquire them.

Run the math on the cutting board again, hypothetical. Blended across all orders, ads cost $24 per order. But new customers cost more: say $38 each. Each board order leaves $30.14 before ads ($62 minus $16, $14 and $1.86).

Buy 100 new customers in January. That's $3,800 in acquisition cost. Their first orders leave 100 x $30.14, which is $3,014. So January's group is $786 underwater on day one.

Now say 30 of those 100 come back within six months for another board at the same $30.14 before ads. That's another $904.20. The group has now left $3,918.20 against $3,800 spent. It paid back, in month six, with $118.20 to spare.

Six-month lifetime value on a contribution basis: $39.18 per customer. Customer acquisition cost: $38. That's a ratio of about 1.03 to 1.

A product can lose money on the first order and still be a good business. It just needs the second order to show up, and you need to know when it does.

That's why I want a cohort view instead of a single average. A cohort is every customer acquired in the same month, tracked forward. If the January group paid back in month six and the March group is still underwater in month eight, something changed: the offer, the audience, or the product's repeat rate. An average across all customers would blur that into one comfortable number.

Three rules for this section:

1. Measure lifetime value in contribution, not revenue. Revenue-based lifetime value makes thin-margin products look like heroes.
2. Pick a payback window you can afford. If cash is tight before the fourth quarter, a 12-month payback is a loan you're giving your customers.
3. Separate new-customer acquisition cost from blended cost per order. Blended cost flatters. New-customer cost tells the truth.

Average order value matters here twice. It raises first-order contribution, which shortens payback, and it raises every repeat order. For category benchmarks, see our DTC average order value statistics for 2026. And if your plan to raise it involves a sitewide discount, read whether discounts increase Shopify conversion rate first. A discount raises conversion rate by handing over the exact margin this whole section depends on.

Blended efficiency: marketing efficiency ratio

Platform dashboards each grade their own homework. Meta reports the sales it thinks it caused. Google reports the sales it thinks it caused. Add them together and you often get more sales than the store actually made.

The marketing efficiency ratio sidesteps the argument. Shopify defines it as total revenue divided by total marketing spend, and says marketing spend should include paid ads, influencer fees, creative production, marketing tools, and agency or contractor costs. It's blended on purpose. It doesn't care which platform claims the sale.

But a ratio with no target is just a number. You need a break-even line.

Here's the math on the hypothetical kitchen store. Total revenue: $20,645. Total ad spend across the three products: $3,840 plus $3,200 plus $200, which is $7,240. Marketing efficiency ratio: $20,645 divided by $7,240, about 2.85.

Is 2.85 good? You can't know until you compute what the store keeps before marketing. Contribution after ads was $3,875.65. Add the $7,240 of ads back, and the store keeps $11,115.65 before marketing, which is 53.8% of revenue.

Break-even marketing efficiency ratio = 1 divided by 0.538, about 1.86.

At 1.86, every marketing dollar is paid back and nothing is left over. At 2.85, this store keeps $3,875.65 a month in contribution, 18.8% of revenue, before fixed costs. That's the number that pays for software, people and inventory growth.

A marketing efficiency ratio of 3 can be a great month for one store and a slow bleed for another. The break-even line is what turns it from a number into a decision.

The same break-even logic works per campaign. Our break-even return on ad spend calculator does the math for a single product: put in the price, cost of goods, shipping and fees, and it shows the line your ads have to clear.

Now, the honest limit. The marketing efficiency ratio tells you whether marketing is working in total. It doesn't tell you which product or page is causing it. That's why it sits next to profit per product on the dashboard, never instead of it. And if your instinct when the ratio drops is to spend more, read why more ads won't fix conversion before you raise a budget.

There's one more trap here. Earnings per click, the number affiliate and traffic buyers live by, looks a lot like revenue per visitor but answers a different question. We break the difference down in revenue per visitor vs earnings per click.

Attribution, first-party and server-side tracking

Every number above is only as good as the data under it. And ecommerce data leaks in three places.

Leak one: test orders and junk sessions. Shopify says its own test orders and simulated transactions don't display in payouts or reports. Good. But anything that counts purchases outside Shopify's reports (an ad pixel, a spreadsheet export, a homemade dashboard) needs its own filter. One $500 test order in a month with 265 real orders inflates revenue and, worse, flatters the conversion rate of whatever page it came through. Bots and internal team sessions do the opposite: they add visitors who were never going to buy, which drags revenue per visitor down.

Leak two: browser-only tracking. Pixels that live only in the browser miss events when scripts get blocked or pages close early. That's why the ad platforms now offer server-side routes. Meta describes its Conversions API as a connection between an advertiser's marketing data and the systems Meta uses to target ads and measure outcomes, and it accepts events from a server, a website platform, an app or a CRM. Google's server-side tagging moves tag processing onto a server you control, where only you have the data until you choose to send it elsewhere.

Leak three: platform-reported sales. Each ad platform claims its own conversions, with its own attribution window. Treat those claims as directional. Your store's order data is the source of truth, which is exactly why the marketing efficiency ratio uses store revenue.

The practical setup looks like this:

• First-party tracking that records sessions and orders on your own domain, tied together, with test orders and internal traffic excluded before anything gets reported.
• Server-side events to the ad platforms, so the algorithms learn from real purchases.
• One source of truth for revenue: store orders, net of refunds, never the sum of platform claims.
• Consent respected. Server-side tracking doesn't replace consent. Send only what your privacy policy and your visitors' choices allow.

Here's where AI changes the job. The math in this guide was never hard. The work was collecting it: exporting orders, matching ad spend to products, updating cost sheets when a freight quote changed. That's the chore founders in the Shopify Community thread describe as stitching together the store, the ad platform and a spreadsheet just to answer whether they made money that week.

That stitching is what we automated for our own clients. The system we run does first-party tracking that excludes test orders, shows a before and after revenue per visitor strip for every page we rebuild, calculates profit per product, gives a customer acquisition cost to lifetime value view with monthly cohorts, and runs an offer lab that compares offers by profit per session instead of by conversion rate. The last one matters most. An offer that converts better and earns less per session loses in our lab, even when it would win a normal A/B test.

You don't need our system to do any of this. A clean spreadsheet and a Monday habit will get you most of the way. The point is to measure profit per session, whatever tool does the adding.

Tariffs, landed cost and repricing

Landed cost is everything it takes to get one unit onto your shelf: what the supplier charges, freight, insurance, customs brokerage, and duty. It sits at step 2 of the contribution margin math, which means every dollar of new duty comes straight out of profit per order.

Tariff rules have changed several times since 2025, and they're still moving, so treat this section as a snapshot dated September 28, 2026, and check the current rules before you act.

As of this writing, here's what U.S. Customs and Border Protection says. In a June 24, 2026 release, CBP said the executive order suspending duty-free de minimis treatment went into effect August 29, 2025, and that new regulations indefinitely suspend duty-free de minimis treatment for imports valued at $800 or less. In plain words: low-value shipments that used to come in duty-free now pay duty. If your model relied on shipping single orders straight from an overseas supplier, your landed cost changed.

Duty rates themselves depend on the product's classification and country of origin. The official source is the Harmonized Tariff Schedule published by the U.S. International Trade Commission. Look up your products' codes, confirm the rate with your customs broker, and update your landed cost sheet. I'm not a customs broker, and a guide like this can't classify your product for you.

What I can show you is the repricing math, because that part never changes. For the full absorb, pass-through or split decision, see our tariff pricing strategy breakdown.

Back to the hypothetical cutting board. Say new duty adds $3 to its landed cost, from $16 to $19. Ad spend stays at $3,840 a month. Each order now leaves $27.14 before ads instead of $30.14. On 160 orders, that's $4,342.40 minus $3,840 in ads: $502.40 a month in contribution, down from $982.40. Half the product's profit gone, and the revenue report didn't move a cent.

Now raise the price from $62 to $66. Fees go to $1.98. Each order leaves $66 minus $19, $14 and $1.98, which is $31.02 before ads. If the page still converts 160 orders, contribution is $4,963.20 minus $3,840: $1,123.20. Better than before the duty.

But will it still convert 160 orders? Here's the question that matters: what's the lowest conversion rate the new price can survive?

To match the old $982.40, the store needs $982.40 plus $3,840, which is $4,822.40, from orders worth $31.02 each. That's 155.5 orders, so 156. On 5,000 sessions, that's a conversion rate of 3.12%.

So the board can fall from 3.2% to about 3.1% and still break even on the price change. Not much room. At 3.12% and $66, revenue per visitor is $2.06.

A price increase is a bet that your page can carry the new number. Know the lowest conversion rate you can survive before you place it.

That's why repricing is a page problem as much as a pricing problem. A page that explains why the product costs what it costs holds its conversion rate after a price increase. A spec sheet doesn't. Three moves, in this order:

1. Re-run landed cost for every hero product with current duty and freight quotes.
2. Compute the survival conversion rate for each proposed new price, the way we just did.
3. Strengthen the page before the price goes up, not after. Bundles, clearer value copy, and proof all help the page hold its conversion rate at a higher price.

If you sell on Amazon too, run the same math on referral fees. Our Amazon page quotes the line founders feel most: platforms like Amazon take 15% of every sale, which is $150,000 in referral fees for every $1 million in sales.

A weekly profit dashboard

Month-end is too late. By the time the profit and loss statement shows a bad campaign, you've paid for four weeks of it. A founder in that same Shopify Community thread said exactly that: the number arrives too slowly to act on.

So run a weekly dashboard. Monday, 30 minutes, the same 12 lines every week. Keep it boring. Boring is how you spot the week something changed.

Top half: is the store selling?

• Sessions, conversion rate, average order value, and revenue per visitor, for the store and each hero product.
• The same four numbers from last week and the same week last year, next to this week's.

Bottom half: is the selling worth it?

• Contribution margin per product, in dollars and percent.
• Profit per session per product, ranked.
• Marketing efficiency ratio, with the break-even line drawn under it.
• New-customer acquisition cost and the latest cohort's payback status.

The trick is the ranking. If the profit per session ranking changes, something happened: a cost moved, an ad got expensive, a page broke. Find out which before you touch anything.

That last line matters more than the other 11. A dashboard that doesn't produce a decision is a screensaver.

For a quick outside check on the whole store, the ecommerce health scorecard scores traffic, conversion, order value and repeat rate together and lists the fixes by revenue.

Q4 cash flow planning

Profit and cash are different animals, and the fourth quarter is where founders learn it the hard way.

Here's why. You pay for holiday inventory in September and October. You pay for the ads in November, when auctions get busier. The money arrives as orders ship, and some of it goes back out in January as returns. A store can be profitable on paper all quarter and still run out of cash in the middle of it.

Run the math on the hypothetical knife, since it's the product we'd push. Say you want 600 knives in stock for November and December. At $38 landed, that's $22,800 paid before a single one sells. If each sale leaves $29.40 after ads, those 600 knives return $17,640 in contribution, but only as they sell, spread across eight weeks, with the ad spend going out first.

Four questions to answer before October ends:

1. How much cash goes out before the sales come in? Inventory, deposits, and the first weeks of ad spend.
2. What happens if ad costs rise and conversion rate holds? Re-run contribution per order with a higher ad cost per order. If the product goes negative, cap its budget now.
3. Which products get the budget? Rank by profit per session. The fourth quarter is when the cutting board problem gets expensive, because heavy shipping and busy auctions hit together.
4. What comes back in January? Estimate returns from last year's rate for each product and hold that cash.

The fourth quarter doesn't reward the store with the most revenue. It rewards the store that still has cash in January.

And the cheapest cash in the quarter is conversion rate you already paid for. Every visitor who lands on a page that sells better is a visitor you don't have to buy twice. That's why we rebuild the hero product pages before the spike, never during it.

Start here: the full reading list

This guide is the hub for the Profit & Analytics side of the blog. Here's where to go next, in the order I'd read it.

The core metric

• What revenue per visitor is and why it beats conversion rate
• How to calculate Shopify revenue per visitor step by step
• What a good revenue per visitor looks like for a Shopify store

Metrics that mislead

• Why your Shopify conversion rate is a vanity metric
• Revenue per visitor vs conversion rate, and which to track
• Revenue per visitor vs earnings per click, explained

Where profit leaks

• The $60K store that was actually losing money
• Whether discounts increase Shopify conversion rate
• Shopify bundle strategy mistakes that kill your margin
• Why more ads won't fix conversion
• The tariff pricing strategy that protects margin

Benchmarks and channels

• DTC average order value statistics for 2026
• Shopify vs Amazon revenue per visitor, the 2026 study

FAQ

What is ecommerce profit analytics? Measuring what each product, page, channel and customer leaves you after every variable cost, instead of what it brings in. Revenue minus landed cost, shipping, fees and ad spend is contribution margin. Divided by sessions, it's profit per session.

How do I calculate profit per product on Shopify? Take the product's revenue, subtract landed cost of goods, shipping, fees, discounts and the ad spend that drove it. Divide that contribution by the product page's sessions to compare products fairly.

What's the difference between contribution margin and gross margin? Gross margin only subtracts product cost. Contribution margin also subtracts shipping, fees, returns and ad spend, which is where most of the profit disappears.

What is a good marketing efficiency ratio for ecommerce? It depends on your margins. Your break-even is 1 divided by your contribution margin before marketing. A store keeping 54% before marketing breaks even around 1.86.

How do tariffs change my product profit? Duty raises landed cost, which comes straight out of contribution per order. Re-run each product with current rates from the Harmonized Tariff Schedule, then find the lowest conversion rate a price increase can survive.

Which metrics should be on a weekly ecommerce profit dashboard? Revenue per visitor and its two inputs for each hero product, contribution margin, profit per session, new-customer acquisition cost, cohort payback, and marketing efficiency ratio against its break-even line. Add cash and inventory cover before the fourth quarter.

What to do next

Pick your top three products. For each one, write down last month's sessions, orders, revenue, landed cost, shipping, fees and ad spend. Calculate contribution per order and profit per session. Then rank them.

If the ranking doesn't match where your ad budget goes, you've found your first move for this week.

---

Book Your Profit Audit

On a profit audit, we calculate revenue per visitor and profit per session for your hero products and show you exactly which page is leaking the most money per click. Then we show you how to rebuild a high-converting product sales page in less than 15 minutes.

Book Your Profit Audit →

Or go here to check it out → revenueflows.ai

P.S. Revenue is the crowd noise. Contribution is the score. Stop celebrating the product that sells the most until you know it's the one that keeps the most.