Where to Start With Search Marketing

Four places this can fail, and they are not interchangeable

Most people arrive at search marketing with the solution already chosen. They need content, or links, or a faster site, or someone to run ads. Occasionally that instinct is right. More often it is a guess made before anyone looked, and the money goes to the part of the machine that was already working.

There are only four places a search channel breaks, and they sit in a chain. Search engines have to be able to reach and store your pages. The stored pages have to appear for things people actually search. The appearances have to earn a click. The clicks have to turn into something you care about. Each link in that chain has a different remedy, a different cost, and a different clock. Publishing more content when the problem is that half the catalog is not indexed produces nothing but a bigger unindexed catalog.

So the first month is diagnosis. Not a purchase, not a strategy deck. You are trying to establish which link is failing, and you can get most of the way there yourself with instruments you already have.

The instruments, and the one question each is allowed to answer

Confusion about which tool answers which question causes more bad decisions than any technical fault. Keep the boundaries strict.

Google Search Console — the only source that reports impressions and position at all. Google defines its four metrics precisely: clicks are the number of times a user clicked your site from Google Search, impressions are how many times your site appeared in results, CTR is clicks divided by impressions, and average position is the average position of the topmost result from your site, not the average of everything you rank with. It also carries the Page Indexing, Crawl Stats and Manual Actions reports, which answer the reach-and-store question.
Your analytics — sessions, engagement and conversions. It never sees an impression or a position. Anything phrased as a ranking question cannot be answered here.
A crawler — what URLs exist, what status code each returns, what each one declares about itself through canonicals, robots directives and internal links. This is the inventory nothing else gives you.
Field performance data — what real visitors actually experienced, reported at the 75th percentile and split by mobile and desktop. Not a score from a single simulated load on a machine in a data center.
Keyword and demand data — how many people search for the thing at all, and in what words. This is the only one that tells you whether the ceiling you are pushing against is your site or the market.

The first four weeks, in order

Run these in sequence. Each step is cheap, and each one can end the investigation early.

Get access, at the right level. Search Console verified as a domain property so that subdomains and both protocol variants are covered rather than one slice of the site. Analytics with admin rights. The CMS. If server logs are obtainable, find out now, because on many hosted platforms they simply are not.
Export everything before it ages out. Search Console retains sixteen months of performance data and caps the interface and its spreadsheet export at 1,000 rows. Whatever you do not export this month is gone in the shape you will want it later. Pull queries, pages, countries and devices, and keep the file.
Crawl the site. You are looking for the count of URLs that return 200, and for anything that returns 404, 410, 5xx or a chain of redirects. Compare that count with what the Page Indexing report says is indexed. A large gap between what exists and what is stored is the highest-value finding available in the first week.
Read robots.txt and the indexing directives yourself. A single stray Disallow or a noindex shipped from a staging environment outranks every other explanation for a site that gets no traffic.
Segment the Search Console data before you total it. Split branded queries from everything else, then split by template — product pages, category pages, articles, the homepage. Site-wide averages hide the pattern that tells you what to do.
Check demand. Look up search volume for the ten terms you assume matter. Some businesses discover their category is barely searched, which is a strategy finding, not an SEO finding, and it should change the plan rather than the tactics.
Record the baseline. Clicks, impressions and average position by segment; conversions and revenue by landing page and template; field performance at the 75th percentile; the indexed-page count. Date it. Every argument you will have for the next year is settled by whether this file exists.

Reading what you found: visibility, relevance, or conversion

Three patterns cover most sites, and the remedy differs completely.

Few impressions. The pages are not appearing, which means they are not stored, or they are stored and judged irrelevant to everything you hoped for. Check indexation first because it is binary and quick. If the pages are indexed and still invisible, this is a relevance and competition problem, and it is the slowest of the three to move.

Impressions but few clicks. You are appearing and not being chosen. Look at what your average position actually is, since an average built from the topmost result per query flatters a site that ranks eleventh for everything. Then look at what else is on the result page. When a result page carries an AI Overview, Ahrefs measured a 34.5% lower average click-through rate for the top-ranking page on informational keywords compared with similar keywords without one — 300,000 keywords, desktop Search Console data, March 2024 against March 2025. That is a structural tax on informational queries, not a fault in your title tags.

Clicks but nothing after them. This is the cheapest problem to fix and the one most often outsourced to the wrong specialist. If organic sessions arrive and do not convert while paid sessions on the same terms do, the issue is the landing page, the offer, or the fact that the queries you rank for belong to people who were never going to buy.

Where paid search belongs in a first plan

Paid search is usually sold as the fast alternative to organic. Its more valuable early role is as an instrument. A small, tightly controlled campaign on the terms you believe matter returns query-level evidence in days: which phrasings people actually use, which of them convert, what a lead is worth, and whether the landing page holds up under traffic you chose. Organic work answers those same questions in months.

The discipline that makes this work is the search terms report — the record of what people actually typed, as distinct from the keywords you bought. Read it weekly and it will reshape your organic priorities faster than any keyword tool, because it is your own conversion data rather than a national estimate.

The trap is starting paid before conversion tracking is correct. Modern bidding optimizes toward the conversions it is told about; if those are miscounted, duplicated, or firing on the wrong event, the system will spend efficiently toward the wrong outcome. Verify tracking before the first campaign, not after the first disappointing month.

What not to buy in month one

Saying no to the wrong purchase early is worth more than any single fix.

A crawl budget project, unless you are genuinely large. Google's own guidance says crawl budget management matters for sites with over a million pages that change roughly weekly, sites with 10,000-plus pages changing daily, or sites with a large share of URLs sitting in Discovered — currently not indexed. Below that, it is not your problem.
A routine disavow file. Google's position is that in most cases it can assess which links to trust without additional guidance, and that most sites will not need the tool. It is a remedy for a documented manual action, not hygiene.
AI-specific markup or files. Google states plainly that you do not need to create new machine readable files, AI text files, or markup to appear in AI features, and that there is no special structured data to add. Anything sold on that premise is selling a product Google says does not exist.
A guarantee. Google's own hiring guidance says no one can guarantee a number one ranking and warns against providers who claim a special relationship with it.
A redesign, until you know what is currently earning. Redesigns performed without a URL inventory and a baseline are how sites lose traffic they never knew they had.

How to tell, three months in, whether any of this worked

Decide the measures before the work starts, because measures chosen afterward are chosen to flatter whatever happened. For a first engagement, three hold up.

Coverage. The number of URLs you intended to have indexed that are actually indexed, and the direction that number is moving. This is unambiguous and it moves first, often within days of a fix being recrawled.

Non-brand impressions by template. Impressions are a leading indicator: they move when the stored set or its perceived relevance changes, before clicks follow. Excluding brand terms keeps the number from being a measure of your marketing elsewhere.

Revenue or qualified leads from organic landing pages, compared with the dated baseline rather than with last month. Attribution will understate this and you should say so in the report rather than quietly benefiting from it.

Set the expectation honestly at the same time. Google says some changes can take effect in a few days while others could take several months, and it explicitly declines to promise that changes will produce noticeable impact — if there is more deserving content, that content will keep ranking. Anyone who replaces that with a date and a percentage is not better informed than Google.

Frequently Asked Questions

What should I do first if my business has never done any SEO?

Verify Search Console as a domain property and export sixteen months of performance data before any of it ages out. Then crawl the site and compare the number of URLs returning a 200 status with the number the Page Indexing report says are indexed. That single comparison tells you whether you have a storage problem or a relevance problem, and those have completely different remedies. Only after that is it worth deciding what to buy. The order matters because almost every expensive mistake in this channel comes from purchasing a solution before establishing which link in the chain was broken.

Do I need an audit before I start anything else?

Not always. If the site is small, recently built on a mainstream platform, and already appearing for its own brand name, a full technical audit will mostly confirm that things are fine. What you need first is the baseline and the indexed-versus-existing comparison, which takes an afternoon. Buy an audit when that comparison shows a gap you cannot explain, when the site is large enough that findings will be template-level rather than page-level, or when a migration, platform change or unexplained traffic drop is involved. An audit is a diagnostic instrument, not a starting ritual.

Should I start with paid search or organic search?

Run both, but for different reasons. Paid answers demand and conversion questions in days: which phrasings people use, what converts, what a lead costs you. Organic changes the structural position and takes months. In practice a small, tightly scoped paid campaign early is the cheapest research available, and its search terms report will reshape organic priorities faster than any keyword tool because it is your own conversion data. The prerequisite is correct conversion tracking. Automated bidding optimizes toward whatever conversions it is fed, so miscounted ones cause it to spend efficiently in the wrong direction.

How much of this can I do myself without hiring anyone?

More than most providers will tell you. Access setup, the data export, a crawl, reading robots.txt, segmenting Search Console by brand and template, and recording a dated baseline are all within reach of a competent marketer with a free afternoon and a crawler. What is harder to do alone is judgment about ordering — deciding which of nine real findings actually moves the number, and which are true but irrelevant. That is the part worth paying for, and you will buy it far more cheaply once the diagnostic groundwork above is already done.

How long before I see anything change?

It depends entirely on which link in the chain you fixed. An indexing blocker removed can show effect within days of the affected pages being recrawled. A page speed improvement takes time to appear in field data because that data reflects real visits accumulating over a period. Relevance and authority work runs in months. Google's own statement is that some changes take effect in a few days while others could take several months, and it declines to guarantee that any change produces noticeable impact. Treat anyone offering a specific date with suspicion.

My competitor outranks me. Where do I start with that?

Start by checking whether you are in the same competition at all. Compare the actual queries each of you appears for in your own Search Console data rather than assuming symmetry, and separate branded from non-branded terms. A competitor with far greater brand demand is not beating you on technique. Then check whether their winning pages answer a different intent than yours — informational versus transactional — because that is a content decision, not a ranking one. Copying a competitor's tactics without establishing which of these applies is the most common way early budget disappears.

Published 2026-08-11