Keyword match types determine how far Google Ads can expand the searches that make your ads eligible. Broad, phrase and exact provide different degrees of reach and control. None replaces a clear offer, meaningful measurement or review of the actual queries. Your choice depends on what you need to learn and which mistakes the business can afford.
For a software implementation company, “software course” and “company that implements software” imply different needs. A shared word does not make both clicks commercially useful. Before expanding reach, decide which business outcome matters and which requests the company cannot serve.
What broad, phrase and exact match change
Google’s match-type documentation describes broad as reaching related searches, phrase as preserving the keyword’s meaning, and exact as focusing on equivalent meaning or intent. Exact therefore does not require a character-for-character copy of the keyword.
| Type | Practical purpose | What to watch |
|---|---|---|
| Broad | Explore additional ways people express a relevant need. | Query relevance, conversion quality and the cost of exploration. |
| Phrase | Work within a defined need while allowing additional wording. | Ambiguous services and apparently related searches that do not fit. |
| Exact | Limit exploration around a particular intention. | Semantic variants, available demand and missed opportunities. |
The input format differs too: broad uses no delimiters, phrase uses quotation marks and exact uses square brackets. This configures a keyword; it does not automatically describe the quality of the person who clicks.
Choose according to the account’s starting point
When conversion measurement is unreliable
First check what the account is learning from. Duplicate form submissions or a button click recorded as a sale can make any match type look successful. Restricting terms may create a more understandable sample, but it does not repair a bad business signal. Our introduction to how Google Ads works explains the wider campaign structure.
When useful commercial queries are already visible
Use them as a baseline for a test. Instead of asking whether broad match is better, ask whether expansion finds additional opportunities at a sustainable cost. Distinguish new customers, branded demand and requests rejected by sales. Cheap conversions can hide expensive commercial work.
When the offer has narrow eligibility
A particular technology, limited service area or enterprise contract needs explicit boundaries. Phrase or exact may be part of the initial design, accompanied by ads and pages that explain those conditions. Adding a city to a keyword is not a substitute for configuring the campaign’s geographic settings.
Read search terms rather than only the keyword list
A keyword is a targeting instruction; a query is what someone searched. The search terms report helps connect the two. Its classification of a query may differ from the match type assigned to the keyword that triggered it.
- Choose a reporting period that allows for the usual conversion delay.
- Review the query and destination together, rather than click totals alone.
- Classify searches as potential customer, information, employment, support or another offer.
- Check valid inquiries against sales records or the CRM where available.
- Record what you will retain, exclude or investigate further.
The report has privacy and coverage limits; it is not a complete log of every search. An isolated query with no sale also should not automatically become a negative when the buying cycle has not yet elapsed.
A hypothetical decision example
Consider an ERP implementation company serving business projects only. A test reveals “ERP for distributors”, “ERP consultant jobs” and “free ERP template”. The first requires a review of fit and outcome; the others may indicate needs the company does not serve. These illustrate classification, rather than a universal negative-keyword list.
The company tests expansion in one service line. It keeps the destination consistent, records which queries generate accepted contacts and agrees on a spending limit before the test. More forms accompanied by fewer valid opportunities would not count as success simply because the advertising dashboard shows additional conversions.
Test one hypothesis and maintain a change log
Record the original match type, proposed change, hypothesis, budget and review date. Avoid changing ads, geography, bidding and landing pages at the same time. Otherwise any difference becomes difficult to explain. Where available and appropriate, use an experiment instead of replacing the setup across the account.
Evaluate cost per accepted opportunity, contact validity and sales capacity alongside CPC and conversion rate. If the sample is too small, the finding may be “we do not know yet”. That is more useful than scaling a decision based on two contacts. Our guide to cost per click separates the price of traffic from its value.
Mistakes that distort the result
- Assuming exact match is strictly literal.
- Repeating every keyword in every match type without a clear purpose.
- Combining sales and micro-actions in one conversion total.
- Excluding terms without checking whether they also remove useful demand.
- Judging a change before delayed conversions arrive.
The next adjustment should address evidence: an intention to restrict, an opportunity to explore or measurement to repair. Match type is one control within that process, rather than an acquisition strategy on its own.
Prepare a test brief before changing match types
A test needs a decision that can be made when it ends. Write down the service you want to expand, the customers you can serve and the outcome that counts as useful. Include boundaries too: locations excluded by operational capacity, requests outside the offer and conditions buyers need to understand. This turns a general discussion about reach into an operational criterion.
Define an observable hypothesis. For example: expanding this query family may uncover implementation needs we currently miss while preserving an acceptable share of valid enquiries. Then agree what acceptable means for this business and how much it can invest in learning. A number borrowed from another account is not automatically an appropriate decision rule.
The brief should name the person assessing outcomes. If marketing labels an enquiry commercial and sales rejects it without recording a reason, the test cannot answer its question. Establish a shared definition and allow a pending category when information is incomplete. A sale that has not happened yet does not prove that the original intention was wrong.
Group needs that can share a meaningful response
Do not organise solely around similar words. ERP migration and ERP support can share vocabulary while requiring different staff, ads and destinations. A useful group contains searches that can receive a coherent commercial explanation. That coherence makes it possible to assess whether a match type expands one need or introduces an entirely different offer.
Inspect ambiguous terms too. A technical expression may have an educational meaning and an enterprise purchasing meaning. Write a few examples that fit and a few that do not. They do not predict every search; they test whether the team understands the boundary. Where the term is highly ambiguous, the offer and landing page must help qualify the request.
Broad match warrants an additional check of bidding and the underlying conversion signal. Google recommends pairing it with Smart Bidding in its match options guide. This technical recommendation does not replace examining which conversions feed the account. If the objective is a superficial click, expanded automation may pursue more of that interaction rather than the intended commercial outcome.
Use a decision matrix for each query family
| Finding | Interpretation to investigate | Reasonable response |
|---|---|---|
| Related new queries and serviceable enquiries | Potentially useful additional coverage exists | Retain the test and evaluate cost once outcomes mature |
| Unrelated queries concentrated around one meaning | The keyword or ad may be ambiguous | Limit that meaning and clarify the offer |
| More forms with the same opportunity count | Expansion may be buying volume without quality | Compare rejection reasons and cost per opportunity before scaling |
| Few queries and outcomes still pending | The evidence cannot support a firm conclusion | Complete follow-up or redefine the scope of the test |
This matrix does not establish universal profitability rules. It prevents the team from applying the same response to different problems. Every decision should connect to evidence, an owner and a verification date. Keep exclusions at the appropriate scope: a valid restriction for one service can block another service’s buyers.
Assess quality without overinterpreting a small sample
Imagine two hypothetical groups with ten forms each. Sales has reached nine people in the first group and only four in the second. Comparing their opportunity counts as though both groups were complete would penalise unfinished follow-up. Before judging the match type, align assessment status or display the difference explicitly in the report.
Even after complete follow-up, a few outcomes cannot establish that the same proportion will continue. Separate observations from assumptions. You can state that a query family generated compatible enquiries during the reviewed period without promising the same result next month. Seasonality, a promotion or a change in service capacity may alter the context.
Record duplicates, existing customers and support requests when the goal is new customer acquisition. Do not delete those records to improve a rate; classify them. They may reveal that the ad or page does not clearly separate purchasing, customer support and learning. The classification can therefore lead to a copy or navigation correction rather than a match-type change.
Run a review that produces specific actions
Structure the recurring review around three questions: which intentions were purchased, what reached the business and what needs changing. Show query families in the first section, rather than an endless list. Add commercial status and rejection reasons in the second. In the third, propose limited changes with a defined method of verification.
Maintain an exception log. If an apparently informational query produced a suitable opportunity, investigate what that person needed before excluding similar wording. If an apparently commercial search produced support requests, examine its actual meaning. The goal is to learn how this audience buys this offer, rather than forcing every query into a category chosen in advance.
At the end of a test, record whether the decision is to retain, expand, correct or stop, and explain why. Include limitations that could change the decision and the next check required. The account then preserves learning when its operator changes and avoids repeating experiments already rejected on evidence.
Finally, distinguish a missing answer from an unsuccessful test. If the team cannot link queries to valid enquiry outcomes, the next task may be improving that connection. Changing the match type repeatedly will not repair the information gap. A test that identifies the missing evidence still produces a practical next step, provided its limits are stated clearly and spending remains within the agreed learning budget.
Revisit the test brief when the offer changes. A service newly available to smaller companies can make a previously unsuitable query relevant. Record that commercial change before reusing an old exclusion or repeating an earlier match-type conclusion.
Your first Google Ads audit is free. Let’s review what limits your account and what to improve first. Request my free audit.
Frequently asked questions
Does exact match only trigger an identical phrase?
No. It can include searches with equivalent meaning or intent. Review actual queries and whether they fit the offer.
Is broad match the best starting point?
That depends on the offer, measurement, budget and ability to evaluate valid contacts. Define a hypothesis and test limit before expanding the whole account.
Should every keyword use all three match types?
Not as a general rule. Each grouping should have a clear purpose. Repeating settings without a reason makes account structure and testing harder to understand.
Can SEOMOS review which match types suit my account?
Yes. Your first Google Ads audit is free. We review the account context and agree on scope before recommending changes or implementation.