194: Copywriting Strategies To Increase Ecommerce Sales With Karon Thackston

Podcast: Download (Duration: 44:16 — 50.9MB)

Stuffing irrelevant keywords into your Amazon backend search terms actively damages your rankings, because shoppers who arrive on a mismatched listing bounce, and Amazon reads repeated bounces as a signal to demote you. Karon Thackston has seen sellers recover rankings simply by removing the broad terms they had added.

Karon has spent 25 years helping ecommerce clients improve search rankings and conversion through copy. She runs MarketingWords.com, has written ten books, and was last on this show in episode 74 talking about optimizing Amazon listings.

This episode covers where Amazon keywords actually belong, the byte limit most sellers still get wrong, how to run a meaningful split test when Amazon gives you no A/B testing tool, and which listing elements are worth testing first.

A note on timing: this is a 2018 conversation. The backend search term limit, the enhanced brand content program, and Amazon’s indexing behavior have all changed since. The underlying principles about relevance and testing discipline still hold.

Key takeaways

• Amazon reverted its backend search term field from 5,000 characters to 250 bytes, and exceeding the limit means Amazon ignores the entire field.
• Keywords in your title, bullets, and description do not need repeating in the backend.
• Broad or loosely related keywords cause bounces, and repeated bounces drag your overall listing performance down.
• Use a substitution test: if a shopper would not accept your product as a replacement for the search term, do not target it.
• Change one element at a time, or you will not know which change produced the result.
• Aim for 1,000 or more sessions per test variant where traffic allows, rather than a fixed time window.
• Check unit session percentage in Business Reports, which is Amazon’s version of a conversion rate.
• Long tail keywords with lower search volume frequently convert better than the highest-volume term.

How many characters can you use in Amazon backend keywords?

Amazon reverted to a 250-byte limit after several years at 5,000 characters, and many sellers never updated their listings. The field will still physically accept more text, which is why the confusion persists.

Bytes and characters are not identical. A letter or number is one byte, while special characters can consume two, though you would rarely use those in search terms anyway.

Spaces no longer count toward the limit. Karon still targets 250 or below rather than trying to recover the handful of characters that exclusion buys.

The failure mode is severe. Exceed the limit and Amazon ignores everything in the field, so a listing with 5,000 characters of keywords may be getting credit for none of them.

You also do not need to fill every search term field. They can all go in one, and the fields are optional by Amazon’s own description.

Where should Amazon keywords actually go?

In your title, bullets, and description rather than the backend. Amazon’s own style guides increasingly say this directly, and Amazon describes backend search terms as a suggestion it is under no obligation to use.

That means duplication is wasted space. Any keyword already present in your copy does not need repeating in the backend.

Karon’s suspicion is that backend search terms eventually go the way of Google’s meta keywords tag, either removed or quietly stripped of weight. Amazon has followed Google’s path on enough other changes to make that plausible.

Her firm ran three tests placing all keywords in the copy with nothing in the backend, and those listings performed well. She is explicit that three tests is not a recommendation, and it is a mild signal that Amazon means what it says.

The working rule is to fit as many keywords into title, bullets, and description as you can without the copy sounding stupid. Rankings are worthless if the person holding the credit card is confused or irritated.

Why do irrelevant keywords hurt your Amazon rankings?

Irrelevant keywords bring shoppers who immediately leave, and Amazon reads that bounce pattern as evidence your listing should rank lower. The traffic actively costs you.

Karon’s example is a stainless steel water bottle targeting “water jug.” Someone searching for a jug clicks, sees a 20-ounce bottle, and clicks away.

Repeat that enough and Amazon demotes you, because it only wants to surface products that sell. Losing rankings costs sales directly, and recovering that traffic through advertising costs money.

The damage is not confined to the broad terms. Amazon evaluates the listing as a whole, so poor performance on several bad keywords drags down the overall data.

What is the substitution test for keyword relevance?

Ask whether a shopper would accept your product as a substitute for what they searched. If not, the keyword is too broad.

The water jug case fails that test. Reasoning that a jug searcher wants water and might therefore accept a bottle is exactly the kind of stretch that produces bounces.

Karon’s firm only uses backend terms that are highly relevant and carry real search volume. She rejects the advice to fill every available character regardless of what goes in it.

How much of an Amazon title gets indexed?

Multiple Amazon leadership contacts told Karon the first 80 characters are indexed and used for ranking. Anything beyond may be indexed and is not necessarily counted in the ranking algorithm.

Truncation is a separate concern. Titles get cut at different lengths across devices, and mobile can show as little as 35 to 37 characters.

That makes front-loading the practical rule. Put your core message where it survives on the smallest screen.

How do you split test an Amazon listing?

Change exactly one element at a time, because Amazon does not permit two identical listings running side by side. Without a true A/B test, isolation is the only way to attribute results.

The common mistake is changing several things at once. Swap a title keyword, tweak a bullet, and add three backend terms, and a good result tells you nothing about which change caused it.

Sample size matters more than elapsed time. Aim for 1,000 or more sessions per variant where traffic allows, and fall back to a two-week window only on low-traffic listings.

Third-party tools can automate this. Karon’s firm does not offer testing as a service, and clients report good results with Splitly, which she has not personally used.

How do you tell if an Amazon listing test actually worked?

Use Business Reports, specifically the details by child item view, which is more accurate even when your listing has no variations. Three numbers matter.

Sessions tell you how many shoppers reached the page. Units tell you how many bought. Unit session percentage is Amazon’s version of a conversion rate.

Read them together rather than separately. New search terms that raise sessions without raising sales mean you attracted the wrong traffic, and those terms should come out.

Sequential testing carries real risk of confounding. Seasonality and one-off events like a social media mention can move numbers independently of your change.

The defenses are sample size and context. Larger session counts smooth out anomalies, historical 30-day and 90-day comparisons reveal trends, and common sense catches obvious problems like testing a summer product in January.

What should you split test first on Amazon?

Backend search terms, particularly given the character limit. You want to confirm the terms you kept are both driving traffic and converting it.

Your advertising data is a shortcut here. Sponsored ads report performance down to the keyword level, which tells you which terms convert before you commit them to the listing.

Title keyword variations come second. A dog treat seller can test crunchy, chewy, jerky, and allergy variants, all containing the core phrase.

Do not default to the highest search volume. Long tail phrases with lower volume are more specific and frequently convert better.

Is enhanced brand content worth the effort?

Karon’s clients saw some improvement and nothing dramatic enough for her to recommend it universally. She is not telling every seller to pursue brand registry for this alone.

The cost is the issue. Producing it takes significant work, and if you outsource listing creation you pay for that work on every listing.

Sentiment in seller communities matched her data. People who went through brand registry, the additional cost, and the production effort described the result as merely okay.

Frequently asked questions

How many characters can Amazon backend keywords be?

250 bytes, after Amazon reverted from a 5,000 character limit. Exceeding it causes Amazon to ignore the entire search term field, even though the field still accepts more text.

Do you need to repeat keywords in Amazon backend search terms?

No. Anything already in your title, bullets, or description does not need repeating, and Amazon’s own guidance increasingly favors keywords in the copy over the backend.

Can irrelevant keywords hurt your Amazon listing?

Yes. Loosely related terms bring shoppers who bounce, and Amazon reads repeated bounces as a reason to demote the listing, which costs sales and raises advertising costs.

How do you know if a keyword is too broad?

Apply a substitution test. If a shopper searching that term would not accept your product as a replacement for what they wanted, the keyword is too broad.

How much of an Amazon title is indexed?

Amazon staff have indicated the first 80 characters are indexed and used for ranking. Titles also truncate differently by device, sometimes to 35 characters, so front-load your message.

How do you split test an Amazon listing?

Change one element at a time, since Amazon does not allow duplicate listings for a true A/B test. Compare across similar session counts rather than fixed time periods.

How many sessions do you need for a valid test?

A thousand or more per variant where your traffic supports it. Low-traffic listings may have to fall back on a two-week window, which is more vulnerable to seasonality.

Where do you find conversion rate in Amazon Seller Central?

Business Reports, under details by child item. The unit session percentage column is Amazon’s conversion rate, shown alongside sessions and units ordered.