The safeguards around your ad account, and how the results are checked
Laurence works only through Amazon’s official APIs, inside your own Amazon Ads account, on the campaigns it creates. Every bid sits under a cap, every launch is run by a person, and every bid change is on the record. This page answers what brands ask before they hand over their Amazon PPC: what Laurence can touch, what stops it overspending, who checks it, how your data is kept, and how its results are measured, including what nobody has measured yet.
Handing over your ad account
Is it safe to let an autonomous system run my Amazon ads?
It is when the system works only through Amazon’s own APIs, keeps every bid inside limits you agree, puts a person on every launch and keeps a record of every change. Laurence does all four, and each answer below says how, and how you can check it yourself.
What access does Laurence need to my Amazon account?
Two connections, each approved on Amazon’s own sign-in page: your Amazon Ads account, where Laurence reads results and runs its campaigns, and your Selling Partner account (Seller Central or Vendor Central), which it reads for sales, fees, inventory and listings. Amazon Marketing Cloud access is optional and lets Laurence measure repeat purchases.
Laurence writes to your Selling Partner account in one case only: publishing listing copy, when a person clicks to publish it. It never asks for your Seller Central password, and you can disconnect Laurence from your Amazon accounts at any time.
Is Laurence approved by Amazon?
Laurence connects through Amazon’s official Amazon Ads API and Selling Partner API as a registered solution provider, and Amazon’s partner program lists it as an approved third-party solution. Laurence’s systems work in your account only through those APIs, under the access you granted.
What happens to my existing campaigns?
On the products Laurence runs, it builds its own campaigns, named with a “[Laurence]” prefix, and pauses yours on those products so the results do not mix; you can keep any of yours running instead. Paused campaigns are not deleted, and apart from the ones you agree to pause, Laurence never changes a campaign it did not create.
Can I take my account back?
Yes, at any time. Everything runs in your own Amazon Ads account, so the campaigns and their history stay yours. If you stop, Laurence pauses and archives its campaigns and leaves yours as they were, or you keep its campaigns running yourself; once you disconnect, Laurence deletes your Amazon access tokens within 30 days (privacy policy).
Spend and safeguards
Can Laurence overspend my budget?
Your Laurence team agrees a budget with you and sets each campaign’s daily budget to fit it. The one budget change Laurence makes on its own is for a campaign that is already profitable and running low: when it has used 80% of its daily budget and its ROAS over the last 14 days clears your bar, Laurence raises that day’s budget by half, at most three times in a day, and puts it back at midnight (how Laurence handles budgets).
It raises a budget only once Amazon’s hourly feed has reported at least half of the spend Amazon shows for that campaign that day, so a late feed cannot make a campaign look more profitable than it is.
Is there a limit on how much Laurence bids?
Yes. Every bid sits between a floor of $0.10 and a maximum set for your account, $10 a click unless you agree another, and it can be set separately for a product, a campaign or a single keyword. Placement adjustments apply on top, as they do for every advertiser (how placement adjustments work).
Can Amazon raise Laurence’s bids on its own?
Not on the Sponsored Products campaigns Laurence launches: they use Amazon’s “down only” dynamic bidding, so Amazon can lower a bid when a sale looks less likely but never raise it (dynamic bidding on Amazon).
What does Laurence do when Amazon’s data is late or missing?
It waits rather than guesses. Laurence prices bids from hours that are about a day and a half old, so their orders have had time to arrive, and it does not reprice from an hour whose data has not come in: it tries again for up to a day, then alerts the team.
What stops a mistake from running for days?
Automated checks and people. An hourly check confirms that every account was actually repriced, problems post as alerts to the Laurence team, and your strategist can pause Laurence’s bidding on your account at any time while your reports keep running. Every bid change is in the record, so a mistake can be traced to the hour and put right.
What happens when a product runs low on stock?
Laurence reads every product’s stock each day, and your strategist can switch on its inventory model for a store or a single product, which slows that product’s ads when its stock would run out within two weeks and restores them after the restock (how Laurence handles inventory).
People and oversight
Does a person check what Laurence does?
Yes. Models do the hourly bidding, and a Deployment Strategist runs your account with you: agrees your goal, budget and limits, runs every launch, reviews each keyword set before it goes live and reviews what the models did. Any other change your strategist makes through Laurence is written to Amazon only after they approve it.
What does Laurence change without asking?
Four things, all inside the limits above: it reprices bids every hour, recalculates each Sponsored Products campaign’s placement adjustments every night, raises a profitable campaign’s budget for the rest of a day when it runs low, and pauses its own campaigns that no longer have anything to bid on. New campaigns, new keywords, negative keywords and pausing your campaigns all go through a person.
Can I see what Laurence is doing?
Yes. The platform lists every target Laurence runs with its bid and whether each change raised, lowered or held it, and for Sponsored Products targets it shows the CVR Laurence expects at each placement and the bid it set there. Ask Laurence answers questions about any of it on the platform or in Slack, and you can turn on a weekly email report.
Can I see why Laurence changed a bid?
Your strategist explains any change you ask about. On the platform, a bid’s main inputs sit beside it: Laurence bids what a click is worth to you, the CVR it expects times what an order is worth, so the expected CVR at each placement shows why a bid sits where it does. The platform does not write out a reason for each hourly change.
Can my own AI tools check Laurence’s work?
Yes, read only. Laurence MCP lets the AI assistant your team uses, such as Claude or ChatGPT, read your Laurence data, including what Laurence changed this week, and it cannot change anything.
Your data
Is my data safe with Laurence?
Laurence encrypts data in transit (TLS) and at rest (AES-256) and protects your Amazon credentials with envelope encryption. Production access is limited to authorized staff, whose own logins to production data are personal and read only, and the keys our systems hold are rotated every 90 days. Details are in our privacy policy.
How does Laurence use my data?
Only to run and improve Laurence for your own account. Laurence never sells your data and never uses it to benefit your competitors; where our models learn across accounts, they use only anonymized, aggregated data that cannot identify you, your products or your business (privacy policy).
Incentives
Does Laurence’s fee push it to spend more of my money?
No: the fee is a share of ad sales, not of spend, so spending more without selling more earns Laurence nothing. Bids are priced toward the goal you set, such as net profit, revenue or a target cost per order, and never above what a click is worth to you (why a share of ad sales).
Doesn’t a fee on ad sales reward easy branded sales?
It could, so Laurence runs branded, competitor and non-branded searches in separate campaigns with their own targets, and reports them apart, so you always see how much of your ad sales came from shoppers already searching for you (branded vs non-branded on Amazon).
How results are checked
How do I know Laurence is working?
Judge it on your own business numbers, product line by product line, against the days before Laurence. Your pilot view on the platform, which your strategist opens for you, compares each product line’s days before launch with the same number of days after it on total sales and profit, not only on ad results.
- Total sales, not only ad sales: the comparison includes the store’s total revenue, so sales that move from organic to ads do not count as growth.
- Margins held fixed: profit is computed with the same margins on both sides, so a price or cost change cannot pass as a gain.
- Sale days left out: Amazon’s big sale days are skipped on both sides, so a Prime Day cannot flatter one window.
- Agreed up front: the pilot’s KPIs and what counts as success are set with you before anything changes.
Why shouldn’t I judge results on the last few days?
Because Amazon credits orders late. A day read two days after it ends shows about 8 to 10% fewer orders than it will end with, and orders reach 99% of their final count only 5 to 6 days after the click. Judge a week once it has settled, and compare results read the same way: the same attribution window and the same products.
Do ad results show what Laurence actually added?
Not entirely, and no one’s do. Amazon credits an order to the last ad clicked, so some ad sales would have happened anyway, most of all on branded searches and where your listing already ranks near the top; nobody has measured how large that share is, and how much ads lift organic rank has not been measured either. That is why Laurence judges results on total sales and runs branded searches apart.
The clean way to measure it is a holdout test: pause the ads on one product, or one group of searches, for a set time and compare. Your strategist can design one with you.
How does Laurence avoid betting on a lucky streak?
It holds each keyword’s CVR as a range, not one number, and bids from the cautious end of that range while the data is thin. A keyword with little history starts from similar searches for the same product, so a few early sales or misses move its bid only a little, and recent results count more than old ones (how Laurence sets bids).
What results have brands seen with Laurence?
Pit Viper, which judges its ads by net profit, raised its net profit a day by 32% while the Sponsored Products campaigns Laurence ran spent 63% less a day, with store sales level; SPYLT, which judges them by total store revenue, doubled its daily Amazon sales over Prime Day, alongside its own Prime Day deals. Read the Pit Viper and SPYLT stories for how.
Who is behind Laurence?
A New York team of quants, engineers and researchers from Jump Trading, Five Rings, Google and Meta, backed by Y Combinator, Susa Ventures, Box Group, Jump Trading and SV Angel, with angel investors from OpenAI and Amazon. More about us.
Start with a pilot measured on your own numbers, with the KPIs agreed before anything changes.
More guides
- Updated November 19, 2026
Amazon ad bids, set from what a click is worth
How to set Amazon Sponsored Products bids from what a click is worth: break-even CPC, target CPC, new keywords and placements, with figures from 25 brands.
- Updated November 17, 2026
Amazon ad keywords, chosen from what shoppers already search
Find Amazon PPC keywords in your search term report, then choose by profit per click. Across 20 brands, small searches brought 25% of non-branded orders.
- Updated November 17, 2026
Amazon ad budgets, paced to the clicks worth buying
Pace an Amazon PPC budget month to date and fix campaigns that run out. Across 15 brands, morning clicks converted 1.4x as often as evening clicks.
- Updated November 20, 2026
Short answers about Laurence
Yes, Laurence runs Amazon Sponsored Products, Sponsored Brands, Sponsored Display and DSP. Short answers on how it bids, what it needs and how to start.
- Updated November 20, 2026
What makes Laurence different
What sets Laurence apart on Amazon ads: quant bidding every hour, a dedicated Deployment Strategist, custom AI for your brand and a fee on ad sales.
- Updated November 19, 2026
Amazon ad costs, counted per order rather than per click
What Amazon PPC costs: across 20 brands, the median click cost $1.63 and each ad order $9.14. CPC and cost per order by store, month and placement.
- Updated November 19, 2026
Amazon ACOS, judged against your own margin
A good ACOS sits below your margin. Across 20 Amazon brands we run, the middle half ran non-branded ACOS at 37% to 84% and branded at 13% to 28%.
- Updated November 19, 2026
Break-even ACOS on Amazon, from the margin each product keeps
Break-even ACOS is your margin before ads divided by price. Calculator, formula and worked example; across 13 Amazon brands we run, the median was 45%.
- Updated November 19, 2026
Break-even CPC on Amazon, and the bid that sits below it
Break-even CPC is CVR × margin per order: the most an Amazon click is worth. Bid below it: across 20 brands, the best searches converted 10% less next.
- Updated November 19, 2026
Amazon ad placements, each bid on what its clicks are worth
How to bid top of search vs product pages in Amazon ads. Across 25 brands, top-of-search clicks converted 2.1x product-page clicks, worth about +108%.
- Updated November 19, 2026
Dynamic bidding on Amazon ads, weighed against what a click is worth
Use down only unless your data says otherwise. Across 25 brands, top-of-search clicks converted 1.37x the average click, while up and down can pay 2x.
- Updated November 19, 2026
New Amazon ad keywords, bid from what similar searches already show
Bid a new Amazon PPC keyword from similar keywords’ CVR, then judge it on clicks. Across 14 brands, 64% of good keywords had long runs without an order.
- Updated November 20, 2026
Your whole Amazon keyword universe, and the searches worth paying for
Size your Amazon markets, find every search, pay where clicks earn. Across 25 brands, phrase, broad and auto targeting found half of non-branded orders.
- Updated November 10, 2026
Amazon ad types, starting where shoppers search
Sponsored Products vs Sponsored Brands vs Sponsored Display, and when to use each. Across 20 brands, Sponsored Products took 93% of Amazon PPC spend.
- Updated November 16, 2026
Amazon Sponsored Brands, priced by where they show
How Amazon Sponsored Brands ads work, what they cost and where they pay. Across 21 brands, top of search took 68% of their spend and converted best.
- Updated November 16, 2026
Amazon DSP and the shoppers who left without buying
What Amazon DSP is now that it lives in Amazon Ads Agent, what it costs and when it pays: in our stores, most product page visits end without a purchase.
- Updated November 17, 2026
Amazon ad management, matched to the size of the job
Amazon PPC agency, software your team runs, or a full-service autonomous system: what each does and costs. Our median store met 1,907 searches a week.
- Updated November 17, 2026
Amazon PPC software for a five-million-dollar brand
Which Amazon PPC software fits a $5M brand, and what each costs. The median $5M brand we run spends $51,900 a month on Amazon ads.
- Updated November 17, 2026
Amazon ads that ease off before the stock runs out
Don’t pause Amazon ads when stock runs low: slow the least efficient targets first. Across 10 brands, 17% of advertised products ran out in 11 weeks.
- Updated November 17, 2026
Contribution margin on Amazon, worked out product by product
Work out contribution margin per Amazon product from price, COGS and fees. Across 12 brands we run, fees took 37% of price and COGS 16%, leaving 45%.
- Updated November 17, 2026
Amazon customer LTV, measured product by product
Count Amazon LTV in margin, by the product a customer bought first. Across 13 US brands, 25.6% of new customers bought again within 12 months.
- Updated November 10, 2026
Amazon ads terms and the formulas behind them
Amazon PPC terms with their formulas, from ACOS and TACOS to match types and placements, plus median ACOS, CPC and CVR across 20 brands we run.


