05 - Keyword Research - SEO Fundamentals

1Keywords and the words a searcher types

The short version

Keyword research is the work of finding out which words people type into Google when they are looking for something Hermetic sells, and writing those words down before any page is written or rewritten. It has three phases: gather widely, value what you gathered, and map each surviving word to a page. This module is the gathering phase. By the end you have a spreadsheet with the topics list, the column headings the next module scores against, and the backup and disaster recovery topic filled from every source available to a ten-person managed services provider in Richmond: Search Console, analytics, Syncro, Nutshell, the people who talk to clients, and Google's own suggestions.

A keyword is the word or phrase a person types or speaks into a search box. It can be one word ("backup") or a whole sentence ("how do I get a deleted file back from the shared drive"). The word "keyword" is singular by habit; most keywords are several words long.

The first thing keyword research shows a company is that its own vocabulary and the searcher's vocabulary differ. Hermetic sells managed services. That phrase appears in the navigation, in the URL /managed-services/, and in the proposals the sales seat sends. An office manager at a Richmond law firm whose previous IT person left does not type "managed services". She has never bought a managed services agreement and does not know the term. She types "IT company Richmond", or "computer support for small business", or "outsourced IT near me". All three describe the same contract Hermetic sells. A page that uses only the industry's term competes for only the industry's search.

Google does understand that "IT company" and "managed IT services" are related, so a page about one can appear for the other. That understanding is partial, and the part Google leaves alone is the title link and snippet the searcher reads on the results page, which are built from our words. A searcher who typed "computer support" and sees a result headed "Managed Services" has to translate before she clicks, and the result below ours that says "Computer Support for Richmond Businesses" needs no translation. Collecting the searcher's words first puts the translation on our side.

Two phrases that differ only slightly count as two keywords. "Backup services Richmond", "Richmond backup services" and "backup service Richmond VA" are three rows in the sheet, even though Google returns much the same page for each. Module 06 values each row separately, and the one that turns out to matter can be any of the three. When you have to choose which form to write down, write the one closest to how a person would say it out loud.

2The three phases of keyword research

Keyword research runs in three phases, in order, and each phase has its own module in this course.

The three phases. The output of each is the input to the next: the raw list feeds the scoring, the scored list feeds the page map that the Rank Math course builds from.

Gather means writing down every word or phrase a qualified searcher might type, from every source you can reach, with no filtering. Value means scoring each gathered row on how relevant it is to what Hermetic sells, how important the service is to the business, where the site ranks for it today, how hard the competition is, and what else Google puts on that results page. Map means assigning each row that survives valuation to one page: an existing service page, a page to be merged from the duplicates module 04 found, or a blog post that does not exist yet. The page map itself is built in Rank Math module 02; this course hands them the scored list.

The obvious shortcut is to filter while gathering, and it costs more than it saves. The filter needs data you do not have at the gathering table: which searches already show the site, where the site ranks, what the competitors' pages look like for that exact phrase. A phrase thrown out early because it sounds wrong for Hermetic ("computer guy Richmond") can be the one a prospect typed on the day she called. Write it down, and let module 06 remove it with a reason attached.

The first gathering pass is the widest, because the site has never had one. Later passes are narrower: after the service pages are rewritten, Search Console shows which new searches the site appears for, and those rows get added and scored the same way. Seasonal words are gathered as their own small pass before the season starts, which section 6 covers.

3The topics list

A topic is the category a group of keywords belongs to. "Backup and disaster recovery" is a topic; "cloud backup for small business Richmond" and "how often should a business back up its data" are two keywords inside it. Topics are the handles you sort the sheet by, and they are listed before any keywords are gathered so that every row lands under one.

Hermetic's starting topics are the ten services in the site's navigation: managed IT services, IT support, cybersecurity, backup and disaster recovery, VoIP, Microsoft 365, phishing protection, vulnerability assessment, co-managed IT, and IT consulting. Each one is a topic because each one has, or will have, one service page, and module 06 hands keywords to pages by topic.

Behind each service sits the client problem that sends someone to Google before they know the service has a name. Hermetic is at the end of that chain: the client has had the problem, searched, read, and called. The search that produced the call happened at the start of the chain, in the client's words. So each topic carries two kinds of keywords: the ones that name the service, and the ones that name the problem. The table gives a candidate problem for each service. These are starting points written at the desk; the sources in sections 5 and 6 replace them with what clients and Google say.

The ten service topics, with the problem behind each one as a client might type it. Candidate wording, to be replaced by gathered wording.
Topic (service in the navigation)	The problem behind it, in the client's words
Managed IT services	"our IT guy quit", "outsource IT small business"
IT support	"computer keeps freezing at work", "IT help desk for small business"
Cybersecurity	"business email hacked what to do", "cyber insurance requirements for small business"
Backup and disaster recovery	"deleted files from shared drive", "server crashed lost everything", "ransomware recovery"
VoIP	"replace office phone system", "business phone that works from home"
Microsoft 365	"move company email to Microsoft 365", "Outlook not syncing"
Phishing protection	"employee clicked a phishing link", "is this email a scam"
Vulnerability assessment	"security assessment for small business", "network scan for insurance"
Co-managed IT	"help for our in-house IT person", "backup for our IT manager"
IT consulting	"IT budget for a small business", "virtual CIO"

Three things cut across every topic and get their own columns rather than their own topic. Place: Richmond, Henrico and Chesterfield are the metro Hermetic serves; Virginia Beach and Lynchburg are pending Jeff's decision, as module 04 recorded. Any keyword in any topic can carry a place ("backup services Henrico"), so place is a column. Brand: searches containing "Hermetic Networks" belong to every topic and none, so a yes-or-no brand column marks them. Industry: the site has four industry pages, and "IT support for law firms Richmond" belongs to the IT support topic with the industry noted. This is enough structure for ten services; if a topic grows past a hundred rows, split it then.

4The keyword plan spreadsheet

The keyword plan is one workbook with three tabs. Raw keywords holds everything gathered, one keyword per row, with no deletions during this module. Keyword plan is where module 06 copies the rows that survive valuation and scores them. Topics holds the list from section 3 so that every row's Topic cell is spelled the same way and the sheet sorts cleanly. Set the workbook up now, with the Keyword plan tab empty apart from its headings.

The Raw keywords tab has six columns in this module. Each is defined below, with the backup row from section 8 as the example.

Columns of the Raw keywords tab, with one example row
Column	What goes in it	Example
Keyword	The phrase as a person would type it, lower case, no punctuation unless it is part of the phrase.	deleted files from shared drive how to recover
Topic	One of the ten from the Topics tab.	Backup and disaster recovery
Place	The place named in the keyword, or blank. Richmond, Henrico, Chesterfield, or one of the pending cities.	(blank)
Intent	Which of module 01's four kinds of search this is: navigational, informational, transactional or local.	Informational
Source	Where the phrase came from: Search Console, GA4, Syncro, Nutshell, a named person, autocomplete, related searches, People also ask, a question tool, Google Trends, or a competitor's title.	Syncro ticket subject
Notes	Anything module 06 needs: the date you saw it in Google, the client who said it, whether it reads like a blog post.	Three tickets with this wording in 2026. Blog post candidate, linking to the backup page.

The intent column uses the four kinds of search from module 01. A navigational search wants a specific site ("Hermetic Networks login"). An informational search wants an answer ("what is the difference between backup and disaster recovery"). A transactional search wants to buy or sign up ("backup and disaster recovery services for small business"). A local search wants a provider nearby ("backup and disaster recovery Richmond VA"). Module 01 mapped these onto the three kinds the Rank Math course uses: local and transactional are "service plus place", navigational is "our name", informational is "a problem". Writing the intent down during gathering costs a second per row and saves module 06 from reading every row twice.

Module 06 adds the valuation columns on the Keyword plan tab: relevance, priority, current rank, difficulty, the results-page features present, an opportunity score, and the page the keyword is mapped to. Leave them out of the Raw keywords tab. A column with no data in it invites someone to fill it with a guess, and a guessed rank is worse than a blank.

A spreadsheet can look like heavy machinery for a company of ten. The reason it is a sheet rather than a document is sorting: module 06 sorts by topic to hand each service page its keywords, by intent to separate the blog post candidates, and by source to see which sources produced rows that survived. The sheet is also the file the page map's "primary search" and "other searches" columns are copied from, so the two documents share wording from the start.

5Internal sources

Internal sources are the records Hermetic already holds and the people who hold the rest in their heads. They come first because they contain the words real clients used, and because the external sources in section 6 work best when seeded with real phrases.

Search Console performance report

Google Search Console is Google's free tool for site owners, and its performance report lists the searches for which Google showed a page from the site, with how many times it was shown (impressions), how many times it was clicked, and the average position. The Queries tab of that report is the only place that shows the exact words searchers typed before reaching www.hermeticnetworks.com. It covers the last sixteen months at most, and it exports to a spreadsheet from the export button at the top right of the report.

Whether a Search Console property exists for the domain, and who owns it, is not known yet; module 04 put finding out on the list. The Rank Math course connects it in Rank Math module 06. Once you have access, export the Queries table for the longest date range offered, paste it into a tab named Search Console export, and copy every query into Raw keywords with source "Search Console" and the impressions figure in Notes. Until then, write "Search Console: not known yet, requested" and the date into the first row of each topic, so the gap stays visible.

The report shows only searches that already show the site. A site with three competing managed IT pages and 57 tourism pages appears for the searches those pages earned, which can be about the Metro Richmond Zoo. It cannot show a search the site has never appeared for, and those are most of the list.

Analytics

Google Analytics 4 (GA4), if it is installed, records visits to each page and where they came from. It does not record the words searched; Search Console is the only source for those. What GA4 does show is which pages receive visits from organic search, and a page that receives them is about a topic people find us for. Read the landing pages report filtered to organic search and add any topic that is missing from the list. Whether GA4 is installed is not known yet.

Syncro ticket subjects

Syncro holds every ticket Hermetic has worked. The subject line is written either by the client in the email that opened the ticket or by the technician from what the client said on the phone, and either way it is closer to the client's words than anything on the website: "can't get into my email", "shared drive is gone". Search tickets for each topic's plain words (for backup: "backup", "restore", "deleted", "recover", "ransomware", "crashed") and copy the subject lines as written, misspellings included, with source "Syncro ticket subject". A subject that appears on several tickets gets one row and a count in Notes. These rows are almost all informational intent and most are blog post candidates.

Nutshell lead notes

Nutshell holds the leads and the notes from the first conversation with each one. Those notes record what the prospect said they wanted before anyone at Hermetic had named it for them: "they want someone to handle all of it", "their current guy is slow", "they got a cyber insurance questionnaire and cannot answer it". Read the notes on every lead from the last year, won or lost, and copy the phrases that describe the need. Where the source field says the lead came from Google, the words they searched are worth asking about.

The people to ask

Five groups of people hold words that are in no system. Ask each the same four questions, and write the answers down verbatim.

Jeff

Why the company exists, which services it most wants to sell next year, and what he wants the company to be known for. His answers set the priority column in module 06, so record them now.

The technicians

The words clients use when something breaks, which are the words they type into Google an hour before they call. "The internet is down" when the firewall failed; "Outlook is broken" when the mailbox is full.

The sales seat

The words prospects use before they buy, the questions every prospect asks before signing, and the objections, especially about price. The sales seat also knows which marketing phrases prospects ignore.

Clients

How they found Hermetic, and if it was Google, what they typed. Ask the last five clients who signed. Most will have come from a referral, and the one who came from a search gives you a real keyword.

Prospects who did not buy

What they searched for, who else they talked to, and what the winner's page said that ours did not. Nutshell has their names. A short email with one question gets more answers than a survey.

The four questions

What words describe what we sell. What words do clients use for the same thing. What do prospects ask before they buy. What do people search for before they know they need us.

The answers will include Hermetic's own vocabulary, because the people answering work here. Push for the client's phrase each time: "what did she say on the phone" rather than "what do we call it". Both go into the sheet, as two rows under one topic, and module 06 decides between them with data.

6External sources

External sources are what Google shows anyone who searches, plus one tool for seasonality and the competitor titles module 04 collected. They are free, they need no access to anything, and they extend the internal list with phrasings nobody at Hermetic would have thought of.

Where the rows come from. Every source writes to the same six columns, and the Source column records which one.

Google autocomplete

Autocomplete is the list of predicted searches that drops down under the search box as you type. Google builds it from searches people have run, filtered by its policies, and shaped by your location and language. Type a seed slowly ("backup and disaster recovery") and record every prediction. Then add a space and a letter ("backup and disaster recovery a", then "b", and so on) to pull predictions starting with each letter, and then add "Richmond" and "for small business". Each prediction is one row with source "autocomplete" and the date in Notes, because the list changes. Use a private browser window so your own history does not shape the list; Google still uses your location, and predictions seen from Richmond are the right ones, since the clients are here too.

Question tools

Two third-party tools do the People also ask and autocomplete collection in bulk and draw the result as a map of related questions: AlsoAsked, which builds from People also ask, and AnswerThePublic, which builds from autocomplete. They are options, and neither shows anything you cannot see by running the searches yourself. For ten topics, running the searches by hand teaches you what the results pages look like, which module 06 needs you to know.

Google Trends and seasonality

Google Trends, at trends.google.com, shows how interest in a search term has risen and fallen over time, on a scale where 100 is the term's own peak in the period chosen. It compares up to five terms against each other on the same scale, and it can be narrowed to Virginia or to the Richmond metro area. It shows no counts. The number it gives is relative interest, and it goes in Notes as a comparison ("Trends: 'data backup' above 'disaster recovery' in Virginia over five years"), never as a volume. Volume figures belong to module 06 and to whichever data source module 06 chooses.

Seasonality is a pattern that repeats at the same time each year. For a Richmond MSP there are at least three. Tax season, from January through mid-April, when accounting clients cannot tolerate downtime and search for help with the systems they depend on. The school year, for any client whose budget and projects run August to June. The Atlantic hurricane season, June through November, when a business owner reads about a storm and searches for backup for the first time. Trends shows whether a term has such a pattern: set the range to five years and look for the same month rising each year. Gather the seasonal rows a few weeks before the season, note the season in Notes, and module 06 treats them as a separate small set.

Competitors' page titles

Module 04 listed the competitors that appear for the service-plus-place searches and recorded the title of each of their service pages. A competitor's title is the keyword that competitor chose for that page. "Managed IT Services Richmond VA | Company A" is a row with source "competitor title" and the competitor's name in Notes. Record the phrasings that differ from ours ("IT services company", "IT solutions", "technology partner") as well as the ones that match. Where several competitors chose the same words, that is evidence the words are searched, and module 06 weighs it.

7Natural-language questions

A natural-language question is a search typed or spoken as a complete question: "how often should a small business back up its data". People search this way on phones, and especially by voice, far more than they do at a desk, and module 01 covered how much of local search happens on a phone. Each service topic should have at least one question in the sheet, and the backup topic will have a dozen once People also ask has been read.

Questions come in a sequence that follows the client's situation. Before the problem has a name, the question is about a symptom: "why did my files disappear from the shared drive". Once the problem has a name, the question is about the fix: "how do I recover files from a crashed server". Once the fix is understood, the question is about prevention: "how often should a small business back up its data", "what is the difference between backup and disaster recovery". The last question in the sequence names a vendor: "backup and disaster recovery services Richmond". The earlier a page of ours answers a question in that sequence, the earlier Hermetic's name is in front of the client, and before any competitor's.

A question keyword is answered by a blog post, and the post links to the service page. The mechanism is below.

A question keyword's path. The blog post is written to answer the question completely; the service page is written to be chosen. One link joins them, and its anchor text is the service's own words.

Three rules make the post do its job. Its title is the question, in the searcher's words, because the title link on the results page is what she compares against what she typed. Its body answers the question fully, including the answer that involves hiring nobody, because a post that withholds the answer to force a call is the kind of content module 08 shows Google demotes. And it contains a link to the service page whose anchor text, the clickable words of the link, is the service's own phrase: "backup and disaster recovery for Richmond businesses". That link carries the reader who wants help to the page that offers it, and it tells Google what the service page is about in a searcher's words.

During this module you write no posts. Each question is a row with intent "informational" and "blog post candidate" in Notes, and module 06 decides which questions earn a post first.

8A gathering pass on backup and disaster recovery

This section runs the whole method on one topic so that the other nine can follow the same path. Backup and disaster recovery is the topic because module 04 found it has three URLs on the site today, /backup-disaster-recovery/, /backup-disaster-recovery-richmond/ and /backup-disaster-recovery-old/, which means the merged page will be rewritten anyway, and the rewrite needs these words.

The sources, in the order consulted

The site itself. The three URL slugs give "backup disaster recovery" and "backup disaster recovery richmond". Page titles beyond the homepage are not known from the crawl, so open the three pages in a browser and copy each title from the tab.
Search Console. Not known yet. One row carries the dated note.
Syncro. Tickets matching "backup", "restore", "deleted", "recover", "ransomware", "crashed". Subject lines as written.
Nutshell. Lead notes from the last year mentioning backup, data loss, ransomware or an insurance questionnaire.
People. The technician who has handled the most restores, and the sales seat, each asked the four questions about backup.
Autocomplete, related searches, People also ask. Seeds: "backup and disaster recovery", "data backup for small business", "cloud backup", each alone, then with "Richmond", then with letters added.
Google Trends. "data backup", "disaster recovery" and "cloud backup" compared in Virginia over five years.
Competitor titles. Each competitor's backup page title from module 04's list.

The resulting rows

The table shows the shape of the result. The rows sourced from the site are real. The rows with Google sources are the kind of phrase those features return for these seeds; you type the seeds and record what Google shows you on the day, which may differ. The Syncro and people rows are the kind of line a technician writes or a client says; the real ones come out of Syncro and out of the conversation. No row carries a volume, because no source consulted here provides one.

Raw keywords, backup and disaster recovery topic, after one gathering pass. Illustrative rows for the Google, Syncro and people sources; replace with what you record.
Keyword	Place	Intent	Source	Notes
backup and disaster recovery		Transactional	Site navigation	Three URLs today; merged page per module 04
backup and disaster recovery richmond	Richmond	Local	Site URL slug	/backup-disaster-recovery-richmond/
(Search Console queries)			Search Console	Not known yet. Access requested, date
server crashed can't open files		Informational	Syncro ticket subject	Illustrative. Count tickets with this wording
deleted a folder on the shared drive		Informational	Syncro ticket subject	Illustrative. Blog post candidate
ransomware recovery		Informational	Syncro ticket subject	Illustrative. Also appears in Nutshell notes
cyber insurance asking about backups		Informational	Nutshell lead note	Illustrative. Ties to vulnerability assessment topic too
we lost everything when the server died		Informational	Technician, named	Verbatim from the conversation
how fast can you get us back up		Informational	Sales seat	Asked by prospects before signing. FAQ for the service page
data backup for small business		Transactional	Autocomplete	Date seen. Replace with what Google shows
cloud backup for small business		Transactional	Autocomplete	Date seen
backup services richmond va	Richmond	Local	Autocomplete	Date seen
what is the difference between backup and disaster recovery		Informational	People also ask	Blog post candidate
how often should a small business back up its data		Informational	People also ask	Blog post candidate
what is a disaster recovery plan for a small business		Informational	Related searches	Blog post candidate; also an IT consulting row
data backup		Informational	Google Trends	Relative interest versus "disaster recovery" in Virginia; note any late-summer rise. No volume
(competitor's backup page title)	Richmond	Local	Competitor title	Competitor name. One row per competitor from module 04

Read top to bottom, the table shows what one topic produces: two rows that name the service, a gap where Search Console will go, informational rows from clients and technicians that are mostly blog post candidates, transactional and local rows from autocomplete, question rows from People also ask, a seasonality note from Trends, and one row per competitor. Across ten topics that is a few hundred rows, which is why valuation is a separate module.

What is not known yet

Which searches already show the site, how many impressions any page gets, and whether GA4 is installed. All three come from access that module 04 asked for. Nothing in this module waits on them; the rows they produce are added to the same sheet when the access arrives, and module 06 can be rerun.

9Do this now

Create a workbook named SEO keyword plan with three tabs: Raw keywords, Keyword plan, and Topics. On Raw keywords, add the six headings from section 4: Keyword, Topic, Place, Intent, Source, Notes. On Keyword plan, add the same six and leave the rest for module 06. Save it where the page map from the plugin course will also live.
On the Topics tab, list the ten services from the navigation in column A, the places in column B (Richmond, Henrico, Chesterfield, then Virginia Beach and Lynchburg marked pending), and the four industries in column C, copying the spelling from www.hermeticnetworks.com as you type.
In Syncro, search tickets for "backup", "restore", "deleted", "recover", "ransomware" and "crashed". Copy at least ten subject lines, as written, into Raw keywords with topic "Backup and disaster recovery", source "Syncro ticket subject", and the number of tickets with that wording in Notes.
Open a private browser window. Type "backup and disaster recovery" slowly and record every autocomplete prediction. Repeat with "Richmond" added and with the letters a through f added. Run the search, record the People also ask questions and the related searches at the bottom. Put today's date in every Notes cell.
Open trends.google.com, compare "data backup", "disaster recovery" and "cloud backup", set the region to Virginia and the range to five years. Write one sentence in Notes on the "data backup" row describing which is higher and whether any month rises each year.
Ask one technician and the sales seat the four questions from section 5 about backup. Write the answers verbatim as rows with the person's name as Source. Then email the last three lost prospects in Nutshell one question: what did you search for when you were looking.
Follow up on module 04's Search Console request. If access exists, export the Queries table for the longest range offered into a tab named Search Console export and copy the backup-related queries into Raw keywords. If it does not, write "Search Console: not known yet, requested" with the date into the first backup row. The topic is done for this module when every source in section 8 has either rows or a dated gap.

11Glossary

Keyword: The word or phrase a person types or speaks into a search box. Usually several words long.
Keyword research: Finding the words people type when looking for what Hermetic sells, and writing them down before pages are written. Three phases: gather, value, map.
Topic: The category a group of keywords belongs to. Hermetic's topics are the ten services in the navigation.
Keyword plan: The workbook with the Raw keywords, Keyword plan and Topics tabs. The scored list in it feeds the page map.
Raw keywords: The tab holding everything gathered, one phrase per row, with nothing deleted during gathering.
Intent: What the searcher wants, in module 01's four kinds: navigational (a specific site), informational (an answer), transactional (to buy), local (a provider nearby).
Source: The column recording where a keyword came from: a system, a person, or a Google feature.
Search Console performance report: The report in Google Search Console listing the searches that showed a page from the site, with impressions, clicks and average position. Its Queries tab is the only record of the exact words searchers typed.
Impression: One appearance of a page from the site on a results page, whether or not it was clicked.
Autocomplete: The predicted searches Google shows under the search box as you type, built from searches people have run.
Related searches: The linked phrases at the bottom of a results page that people who ran this search also ran.
People also ask: The box of expandable questions on a results page. Each is a question people have typed.
Google Trends: Google's tool showing relative interest in a term over time, on a scale where 100 is the term's own peak. Shows no counts.
Seasonality: A pattern in searches that repeats at the same time each year, such as tax season or hurricane season.
Natural-language question: A search typed or spoken as a complete question. Common on phones and by voice. Answered by a blog post.
Anchor text: The clickable words of a link. The link from a blog post to a service page uses the service's own phrase.

12Sources

Google Search Central, SEO Starter Guide: writing for the words readers use, and how titles and links describe a page.
Google Search Central, In-depth guide to how Google Search works: how a query's words are matched to pages.
Google Search Central, Creating helpful, reliable, people-first content: why a post answers the question in full.
Google Search Central, Get started with Search Console, and Search Console Help, Performance report (Search results): queries, impressions, clicks, position, the sixteen-month limit and export.
Google, How Google autocomplete works in Search: predictions are built from real searches and shaped by location and language.
Google Trends Help, FAQ about Google Trends data: values are relative interest scaled to 100, with no counts.
National Hurricane Center, Tropical cyclone climatology: the Atlantic hurricane season runs June 1 to November 30.
Stephan Spencer, Eric Enge and Jessie Stricchiola, The Art of SEO, 4th edition (O'Reilly, early release 2021), chapter 4.
Site facts from a crawl of www.hermeticnetworks.com on 2026-09-29: the navigation, page-sitemap.xml, and the three backup and disaster recovery URLs.

Documentation read 2026-10-01. Google changes often; where this page and the documentation disagree, the documentation is current.

Keyword Research