GUIDEUPDATED NOVEMBER 19, 2026

New Amazon ad keywords, bid from what similar searches already show

Start a new Amazon PPC keyword at the target CPC its expected CVR supports, taking the CVR from similar keywords for the same product and starting a little under your established keywords. Then let its own clicks move the bid a little at a time, because a run without an order says less than it seems: across 14 brands we run, 64% of keywords that ended at or above their store’s CVR had at some point gone twice the store’s usual clicks per order without one.

Laurence8 min read

What starting bid should I use for a new Amazon keyword?

Start at the target CPC the keyword’s expected CVR supports: (contribution margin per order − profit you keep per order) × expected CVR. Take the expected CVR from similar keywords for the same product, or the product’s own average if it has none, and start a little under the keywords you already run.

New keywords convert a little worse than established ones. Across 19 US brands we run, keywords in their first 90 days converted at 0.9 times the rate of keywords the store already had clicks on, at the median store (0.7 to 1.0 times for the middle half of stores), and lower in 14 of the 19 stores.

  1. 01
    Expected CVRSimilar keywords for this product convert at 10%, so the new keyword starts at about 0.9 × 10% = 9%.
  2. 02
    Target CPCWith $12.00 of contribution per order and $3.00 kept as profit, ($12.00 − $3.00) × 9% = $0.81 (the bidding guide’s worked example uses the same product).
  3. 03
    Starting bid$0.81, then checked against Amazon’s suggested bid (below). Its break-even CPC is 9% × $12.00 = $1.08, the most a click is worth on the first order.

How do I bid on Amazon keywords that have little data?

Bid from what similar keywords show and let the keyword’s own clicks move the bid a little at a time, rather than reacting to its first few results. A run of clicks without an order is normal even for a good keyword: in our accounts, most keywords that ended at or above their store’s CVR had at some point gone twice the store’s usual clicks per order without one.

KEYWORDS THAT ENDED AT OR ABOVE THEIR STORE’S CVR: LONGEST RUN WITHOUT AN ORDER, MEDIAN STORE
LONGEST RUN WITH NO ORDER, AT LEASTAT A 10% CVRSHARE OF THESE KEYWORDS
1 × the store’s clicks per order10 clicks94.8%
2 × the store’s clicks per order20 clicks63.7%
3 × the store’s clicks per order30 clicks26.1%

14 US brands, Sponsored Products keywords (exact, phrase and broad), August 14 to November 11, 2026: 621 non-branded keywords with 100 or more clicks whose CVR ended at or above their store’s non-branded CVR on keywords. A store’s clicks per order is 1 ÷ that CVR. A run counts the clicks on consecutive days with no order; clicks on a day with an order never count, so real runs are if anything longer. Each figure is the median across stores.

  • Long runs without an order are normal. At the median store, 64% of keywords that ended at or above the store’s CVR had gone at least twice the store’s usual clicks per order without one: 58% to 70% for the middle half of stores, and half or more in 12 of 14 stores.
  • The longest run was 2.3 times the store’s clicks per order for the median such keyword: about 23 clicks at a 10% CVR, over about 290 clicks in the quarter.
  • So a dry start says little. Fifteen or twenty clicks without an order is a run most good keywords have had. It should lower the bid a little, not end the test.

Should I use Amazon’s suggested bid for a new keyword?

Use it as a check, not a starting point. Amazon builds the suggested bid from the bids that won impressions for ads like yours over the past 7 days (Amazon Ads), so it tells you what wins the auction, not what a click is worth to you.

If your target CPC sits far below the suggested range, the keyword may be too expensive for this product at today’s prices: start at your target CPC and accept fewer clicks, rather than paying more per click than the product can earn back. More on how to set bids from what a click is worth.

How do I use similar keywords to bid on a new one?

Blend the similar keywords’ CVR with the new keyword’s own record, counting the similar keywords as a fixed number of clicks, so the keyword’s own clicks take over as they add up. After a few clicks the similar keywords decide the bid; after a few hundred, the keyword’s own record does.

A NEW KEYWORD’S CVR, BLENDED WITH SIMILAR KEYWORDS AT 9%
THE KEYWORD’S OWN RECORDBLENDED CVRTARGET CPC
No clicks yet9.0%$0.81
15 clicks, no order6.9%$0.62
15 clicks, 2 orders10.0%$0.90
60 clicks, 3 orders6.8%$0.61
200 clicks, 30 orders13.8%$1.24

Illustrative. The similar keywords’ 9% counts as 50 clicks of the keyword’s own (4.5 orders in 50 clicks), added to its own clicks and orders: 15 clicks with no order gives 4.5 ÷ 65 = 6.9%. Target CPC is ($12.00 − $3.00) × the blended CVR.

Similar keywords are ones the same product’s shoppers use with the same intent: the same search in another match type, close variants and other wordings of it, and the product’s other searches in the same group. The more clicks they have between them, and the closer they sit to the new keyword, the more weight they deserve.

How long should I wait before changing a new keyword’s bid?

Wait for clicks, not days: judge a new keyword on clicks counted against your CVR, whether they take two days or two months. A keyword that truly converts at 8% still goes 15 clicks without an order 29% of the time (0.92 to the 15th power).

Days matter only for orders still on their way. Amazon credits a seller’s Sponsored Products ad with orders placed within 7 days of the click (Amazon Ads), so a click from yesterday may yet bring an order: read a new keyword’s CVR over clicks at least a week old, or expect it to rise as late orders arrive.

Should I lower the bid on a new Amazon keyword with no orders?

Yes, a little with each click, not all at once after a few. Each click without an order lowers the CVR the keyword is likely to have, so the bid should drift down with the evidence, as in the blended table above, rather than drop at an arbitrary count.

Check relevance first: a keyword that does not fit the product needs no data, so exclude it now. For a relevant keyword, pause only once it has had enough clicks to show it converts below break-even and still loses money at the bid its CVR supports. The keyword guide works out how many clicks before pausing an Amazon keyword.

Should I bid high on a new Amazon keyword to get data faster?

Not above its break-even CPC. A higher bid buys clicks sooner, but the clicks tell you the same thing at any price, and every click charged above what it is worth loses money on the first order.

If the keyword wins almost no clicks at its target CPC, raise the bid in steps toward break-even ($0.81 up to $1.08 in the example above) and cap the test by clicks or spend before you start: 40 clicks at $0.90 cost $36.

How Laurence bids on new keywords

Laurence prices every new keyword this way, for every keyword at once, and reprices it every hour as its clicks come in. More on how Laurence sets bids.

  • From similar searches first: each keyword’s CVR is held as a range, not one number. A new keyword starts from the CVR of similar searches for the same product, and its own clicks move the range as they come in.
  • From the cautious end: Laurence bids from the cautious end of that range, so a keyword with few clicks is bid below what its average suggests, and the gap closes as its record grows.
  • A little with every click: each click without an order lowers the CVR Laurence expects from that keyword, and its bid follows every hour. There is no fixed click count to pause.
  • Orders given time to arrive: an hour’s results enter the bid once Amazon has had about a day and a half to credit its orders, and orders credited later are folded in as they arrive, so a click whose order is still on its way is never read as a click that failed.
  • From your margins: the bid is your average order value times your margin after Amazon’s fees and the cost of goods you enter (or the ROAS target you set), times that cautious CVR.

Laurence starts every new keyword from what similar searches show and reprices it every hour as its own clicks come in.

How we measured this. Sponsored Products search term reports (Amazon Ads API) for US stores on the Laurence platform, August 14 to November 11, 2026, at Amazon’s default attribution window. A keyword is one keyword text and match type (exact, phrase or broad) per store, across the campaigns that bid on it; automatic and product targets are left out. Only clicks and orders from non-branded searches count: searches for a store’s own brand or its own products are branded, and every other search, competitors’ brands included, is non-branded. The run table covers 14 stores with 20 or more keywords of 100 or more clicks that ended at or above the store’s non-branded CVR on keywords. New keywords had their first click in the window after none from June 15 to August 13, 2026; the ratio covers 19 stores with 1,000 or more clicks and 50 or more orders on new keywords. Each figure is computed per store first, then summarized as the median across stores; no store is named.

More guides

Tomorrow's brandsgrow differently today.

Backed by

  • Y Combinator
  • Susa Ventures
  • Box Group
  • Jump Trading
  • SV Angel
  • & angel investors from

    OpenAIAmazon