How to Do SERP Analysis, Step by Step

Full transcript

1,043 words · script draft, under review.

Intro

Everyone says "check the SERP" before you target a keyword. Few people say what that actually means.

So you search the keyword, scroll down, see a few familiar names, and decide it looks hard. Or easy. Ten seconds, a gut feeling, and the decision is made.

The trouble is that a gut feeling can't be compared, repeated, or explained to anyone else. Next month you won't remember why you said yes.

In this episode I'll show you a manual SERP analysis, step by step, for ten keywords, with a worksheet you fill in as you go. No API yet. The point is to learn what to look at. Automation comes in the next episode.

Before you search: a clean view

SERP means search engine results page: what Google shows for one search. Before you look at it, make sure you're seeing roughly what your audience sees.

Use a private browser window, so your search history doesn't shape the results. Check that the country and language match your market. If you sit in one country and target another, your own results are the wrong evidence. And decide on device. For most consumer topics, mobile is where most searches happen, so look at mobile at least once.

Then pick your ten keywords. Not random ones. Take the main keyword from each of your top clusters from episode twenty-four, or the strongest buying phrases from episode twenty-five. Ten is enough to learn the method and small enough to finish in an hour.

At the top of the worksheet, write down the date, the country, the language and the device. It feels like bureaucracy. It isn't. Results pages change, and in three months you'll want to know whether a difference comes from Google or from the fact that you checked on a different device. Without those four fields, two worksheets can't be compared.

The six columns

The worksheet has one row per keyword and six columns. You fill them in the same order every time.

Column one: who ranks. Go through the top ten organic results and note what kind of site each one is. A big brand or retailer, a marketplace, a forum, a niche specialist site, a publisher. Just a short label. Episode twenty-nine turns this into a proper taxonomy.

Column two: page type. Is the result a product page, a category page, a guide, a "best of" list, a tool, a video, a forum thread? Note the dominant type, and note if it's mixed.

Column three: features. Everything that isn't a normal blue link. A featured snippet, "People also ask", a map with local results, videos, shopping results, an AI-generated summary at the top. Write down where the first organic result actually starts. Episode twenty-eight goes through each feature.

Column four: intent. Based on columns two and three, what does the searcher want? To learn, to compare, to buy, or to find a place? Episode thirty covers this in depth.

Column five: title match. How many of the top ten titles contain the keyword, or something very close? If only two or three do, Google is struggling to find an exact answer. That's often an opening.

Column six: the verdict. Yes, maybe, or no, and one sentence of why.

Worked example

Here's one row, filled in for a fictional store, example-shop.test, and the keyword "manual coffee grinder". This is an illustrative sketch, not a real results page.

Who ranks: two large retailers, one marketplace, two forum threads, three niche coffee sites, and two publishers with buying guides.

Page type: mixed. Four buying guides, three product or category pages, two forum threads, one video.

Features: a "People also ask" box after the third result, and a shopping row at the top. The first organic result starts below the shopping row.

Intent: mostly commercial. People are comparing before they buy. The product pages show that some are ready to buy now.

Title match: six of ten titles contain the phrase.

Verdict: maybe, leaning yes. Two forum threads in the top ten means Google doesn't have enough good dedicated pages. A strong buying guide could compete. A product page alone probably wouldn't.

Notice that the verdict includes a page type. That's the most useful thing a SERP analysis gives you. Not just "can I rank", but "with what kind of page".

Now compare with a second keyword, "buy coffee grinder". Same store, very different page. Shopping results at the top, then eight product and category pages from big retailers and one marketplace. No forums, no guides. Every title contains the phrase. The verdict here is no, at least for now. Not because the volume is wrong, but because every slot is held by exactly the kind of site Google wants, and they all match the query perfectly.

Common mistakes

A few mistakes to avoid.

Looking only at the top three. The top three tell you who's winning. Positions four to ten tell you who's getting in, and that's where you'll start.

Judging by brand size alone. A big name in the results doesn't mean the page is strong. Often it's a thin category page that ranks on the domain's reputation. Open the page and look.

Mixing settings. If you check five keywords on desktop and five on mobile, your worksheet compares two different things.

And not writing it down. The whole point of the worksheet is that you can repeat the analysis in three months, compare, and see what changed.

From worksheet to decision

When all ten rows are done, sort by verdict. The greens are your candidates. The yellows need a second look, often with a different page type than you first planned. The reds you park, with the reason written down.

Ten keywords by hand is about an hour. It's worth doing at least once, even if you'll automate it later, because you'll know exactly what the script should look for. Later in this module we'll turn these columns into rules: difficulty in episode thirty-one, weak results pages in episode thirty-two, and a formal go or no-go gate in episode thirty-five.

Wrap-up

So: same settings, same six columns, same order, and a verdict that names a page type. The SERP analysis worksheet is linked in the description. Next episode: how to get these results pages from an API, so you can do this for hundreds of keywords instead of ten. See you there.