Search-term review / guides

N-gram analysis for Google Ads search terms, in plain words

By Abhishek Pal, performance marketer (apships.com). Last updated 2026-11-02.

N-gram analysis groups your search terms by repeated words and phrases, then adds up their cost, clicks and conversions. It helps you spot a recurring problem that is spread across many small rows. I use it to choose what to inspect next. A costly word alone is not enough reason to exclude it.

What an n-gram means

An n-gram is a word or a short sequence of words from a search. The "n" is the number of words. For example, the made-up search "free brass table lamp" contains the single words "free", "brass", "table" and "lamp". Its adjacent two-word phrases are "free brass", "brass table" and "table lamp".

The free browser view shows every word and two-word phrase in your export. Beside each is its summed cost, clicks and conversions, sorted by cost. You can see which pieces of language account for the most spend without reading every search first.

Why a single-term list can hide a repeated problem

Imagine sorting individual searches by cost. A few expensive rows get your attention. Many smaller rows sit further down. If those smaller rows share the same unwanted intent, reviewing only the first few searches misses the combined cost.

For example, a fictional lamp store has 14 different searches containing "free". Together they cost $93, receive 46 clicks and record zero conversions. The following made-up groups account for those 14 searches. In this example, none of the searches belongs to more than one group.

Phrase in the example searchesNumber of searchesCostClicksConversions
free lamps7$49240
free lamp3$20100
free shipping4$24120

The word "free" brings that $93 into view. It gives you a review priority. It does not establish that all $93 was waste, or that another $93 will disappear after an exclusion.

Read cost and conversions together

Start with expensive words that have no recorded conversions. Then read their full searches. Ask whether the searches describe something you sell, something you give away or something you do not offer.

High cost with conversions needs a different question. For example, a made-up "desk lamp" row with $180 in cost and six conversions has a cost per conversion of $30. Whether that works depends on what those conversions are worth. The n-gram table does not contain your margin decision.

Recent zeros also need time context. Google explains that conversion delay can leave recent periods with full spend but conversions still to come. Check that before calling a pattern wasteful.

The same word can point to different needs

Return to the fictional lamp example. "Free lamps" suggests someone wants the product without paying. "Free shipping" can describe someone willing to buy a lamp while avoiding a delivery charge. If the store offers free delivery, those searches deserve different decisions.

I would write down the unwanted intent before choosing the negative. For this example: "We sell lamps; we do not give them away." That supports inspecting phrases about free products. It does not justify blocking every search containing "free".

Keep the wording that carries the unwanted intent. Removing a word from a phrase can make the exclusion much wider than the evidence you reviewed.

Turn a repeated phrase into a reviewed negative

  1. Find the source rows. Return to the original CSV and collect the full searches containing the candidate phrase. Keep their campaign names and date range.
  2. Check the offer. Read those searches against the products or services in that campaign. Look for any useful interpretation of the same phrase.
  3. Check the results. Review conversions and the time available for them to arrive. Record any reason to hold the decision.
  4. Choose the wording and scope. For example, consider "free lamps" for the fictional store's paid-lamp campaign only after reviewing the matching searches. Keep "free shipping" out of that decision.
  5. Record and revisit. Save the chosen negative, match type, campaign, reason and date in your review notes. Return to those notes when reviewing the next export.

Google's negative keyword documentation says a phrase negative blocks searches containing its words in the same order, including searches with extra words. An exact negative applies to that ordered phrase without extra words. Review the possible reach of the exclusion before choosing. The match types guide walks through that choice.

This is a manual decision after reading the n-gram table. The table does not turn a repeated word into an instruction to block it.

Do not total overlapping rows

For example, one made-up search for "free brass table lamp" costs $4. That same $4 contributes to "free", "brass", "table", "lamp" and the two-word phrases. Adding those n-gram costs would count the same search repeatedly.

Use the table to find patterns, then count each source row once when estimating spend under review. Keep those estimates separate from achieved savings. For help preparing the file, use the search terms export guide; for the wider review, use the CSV analysis guide.

Open the free n-gram view to inspect words and two-word phrases in your browser, with nothing uploaded. If you also want suggested negatives and reasons, the search terms review offers a free summary and a $39 full report. Every suggestion needs review, and savings remain estimates.

Questions

What is an n-gram in a search terms review?
It is a word or a sequence of words taken from a search. Grouping searches by those words lets you see their combined cost, clicks and conversions. This browser tool shows single words and two-word phrases.
Is the Google Ads n-gram tool free?
Yes. The n-gram view is free and runs in your browser, with nothing uploaded. It shows words and two-word phrases from your export, sorted by summed cost.
Should I exclude every word with cost and no conversions?
No. Read the full searches, check whether the offer fits and allow for conversion delay. A word can appear in both useful and unwanted searches. The table tells you where to investigate.
Can I add together all the n-gram cost rows?
No. A search contributes to several words and phrases. Adding those overlapping totals would count the same spend more than once. Return to distinct search-term rows when estimating the spend under review.

Sources

Have an export ready? The search-term review finds the wasted spend in it for free. The full report is $39. Run it on your file.

Guides