A Search Console Only Shows You Your Own Half: Counting 380 Naver SERP Slots Over Four Weeks

For: marketers who want to know where they stand in Naver  ·  Format: measurement report  ·  2026-08-04

TL;DR

Q. Isn't Search Advisor enough to know how much we show up in Naver?

Search Advisor does report impressions, clicks and CTR. But it lists search keywords only for the top 30 by clicks, does not provide the exposure ranking, and no console answers how competing domains divide those slots between them. So I opened the search results for 77 keywords every week and counted the slots one by one, and out of 380 slots exactly 4 were ours, all of them near the top.

Three things to do today

• Check whether the question your console answers is the question you actually asked
• Store one row per slot instead of a single share number, it changes what you can ask later
• Never read exposure numbers from a period when indexing was blocked as a content result

The console answered a different question than the one I asked

I thought this would be simple. Google gives you impressions, clicks, CTR and average position per keyword in Search Console, so Naver's Search Advisor would be similar.

It does report. Naver Webmaster Tools has a "content exposure and clicks" report providing impressions, clicks and CTR. That part matched expectations.

The trouble started once I opened it. What I wanted to know was not "did our exposure grow."

How many slots does this keyword's search result have for outside sites, who holds them right now, and where do we sit?

A console cannot answer this by construction. Lay the documented scope against the question and the reason is plain.

A console reports on your site. It is not a tool for mapping the whole set of slots.

Naver's constraint bit hardest. The documentation states that search keywords are provided for the top 30 by click count. We were tracking 77 keywords, and plenty of them still had zero clicks. No clicks means never appearing in the list at all, so the keywords you are trying to break into are structurally the ones that vanish from the report.

The missing ranking mattered too. The documentation says the exposure position "could be the first page of results or the last." An impression cannot be distinguished between rank 1 and rank 50, which is the opposite of Google Search Console giving you average position.

And the shared limitation settled it. Both are your own site's data. How competing domains divide those slots cannot come out of either, by construction. The map of which tool covers what is in the Search Console and DataLab piece, and this article takes the part outside the console.

One more thing worth noting. Naver results contain several blocks, and most of them are Naver's own services. The places an outside site can actually appear are effectively the web area and ads, so the web area is what I measured. Which block to aim for is covered in the Naver search authority piece.

I wrote the hypotheses down before starting

Before building the crawler I wrote out what I expected, so that I could not later claim I knew all along.

• Hypothesis 1. Large corporations and big comparison sites hold nearly all of the web area.
• Hypothesis 2. Our share will be very low, since the domain is new.
• Hypothesis 3. Rankings will barely move week to week.

To spoil it: only number 2 held. The numbers betrayed 1 and 3, and that is where this article gets interesting.

What was counted and how

Writing down the design so it can be reproduced. I crawled the 77 to 78 target keywords of a service whose SEO I run (a comparison and sign-up business, call it Company A) once a week, from 7 July to 20 July 2026, four rounds in total.

For each keyword I found the web area block, walked its slots from the top, and stored one slot as one row.

crawl_date, keyword, slot, domain, is_brand
2026-07-20, KW-A,   1,    (carrier official store), 0
2026-07-20, KW-A,   2,    (specialist reseller),    0
2026-07-20, KW-A,   3,    (marketplace),            0
...

is_brand marks whether that domain is one of ours. Rather than hardcoding it, the crawler reads an asset list from a spreadsheet, so that partner bloggers or affiliate domains joining later would not require a code change. Add a row to the list and the next crawl classifies differently.

Why I did not store just the share number

My first instinct was to store only "share percentage" per keyword, since that is the number that goes in a report. But doing that sharply narrows what you can ask later.

Aggregates can always be rebuilt from raw rows. Raw rows cannot be rebuilt from aggregates.

The choice paid off in practice. A few rounds in, the question "how much are the competing domains taking slots from each other" came up, and because everything was stored per slot the answer came out of existing data with no new crawl. Had I stored only the share, that question would have needed a fresh month of collection.

Generalising it. In measurement design, "what do we store" matters more than "what do we look at." The metric you want today can always be derived from raw data, but skip the raw data and questions invented later can never be answered about the past. The same holds in web analytics, where recording fine-grained events and assembling reports afterwards always wins.

Result 1. Four of the 380 slots were ours

The four rounds, as recorded.

Crawl date	Keywords	Total web-area slots	Our slots	Share
2026-07-07	77	376	6	1.6%
2026-07-08	78	405	4	1.0%
2026-07-13	78	417	4	1.0%
2026-07-20	77	380	4	1.1%

Hypothesis 2 held. Around 1 percent is close to absent. For reference, a single keyword had between 1 and 10 web-area slots with a median of 5. That the slot count differs per search term was itself something I learned here.

But every slot we held was near the top

There was one more number I should have looked at sooner. Among the keywords where we did appear, what position within the web area were we?

As of 20 July 2026 the four keywords sat at 1st, 1st, 2nd and 3rd. On 7 July, across six keywords, 1st, 1st, 1st, 2nd, 3rd and 4th. In other words, where we got in, we were nearly at the top.

Read 1.1 percent alone and it says "we got in nowhere." Read it together with position and the interpretation flips. We lack breadth, not depth. The two call for opposite fixes. A breadth problem means finding more topics to cover, a depth problem means rewriting the pages you already have. Ours was the former.

Never conclude from one metric. Share and position have to be read together before a direction appears. Share alone leads to "we must write better content," when what was actually needed was "find more search terms we can get into."

Result 2. I expected big players to own it, and they did not

This is where hypothesis 1 broke. Counting the 380 slots of 20 July 2026 by domain turned up 129 distinct domains.

Rather than one domain owning it, 129 of them were holding a little each.

The largest holder had 32 slots, only 8.4 percent of the total. The top five reach 26.1 percent, the top twenty 59.5 percent. And 79 of the 129 hold exactly one slot, which is to say the median slots per domain is one.

The most unexpected part was that domains visibly built on no-code builders or CDN hosting held 30 slots between them. Small sites made without a dedicated engineering team were sitting in the web area, so the premise that this was big-company territory was wrong.

Why that matters in practice: if hypothesis 1 had held, the conclusion would have been "give up on the web area and buy ads." What was actually true was closer to there is a way in and we had not found it. The same 1.1 percent leads to opposite prescriptions.

Result 3. The total does not move while the composition does

Hypothesis 3 broke too. In the table above our slot count sits at 4 from 8 July onward, which looks static.

But look at which keywords those slots came from and it differs every week.

The total stayed at four while turnover ran underneath. Storing only the total would have erased this picture.

The two keywords containing the brand name held their slots all four rounds, which is unsurprising since branded terms are barely contested. The rest is the problem. Of the four long-tail benefit keywords, two dropped out and one left and came back.

Between 7 and 8 July, two slots vanished in a single day, without us touching the site. This looks like a competing domain moving in and pushing us out, and because the records are per slot it was possible to check who took the position.

The practical conclusion: branded keyword share is close to useless as a metric. It is always captured so it never moves, and mixed into the total it actually masks long-tail movement. Splitting branded from non-branded is correct, and it shares a root with the problem covered in the piece on who owns branded search conversions.

What these numbers must not be used to claim

This is the most important section in the article. Every number above carries one condition that shakes the whole result.

Image: A reading taken while the door was shut tells you the door was shut, not what lies beyond it.

Throughout the measurement period that site was deliberately blocking indexing. It was pre-launch, so robots.txt blocked everything and every page carried noindex. That is a period in which search engines were never given the right to list us at all.

So what is 1.1 percent? It is not evidence that the content was too weak to get in. It is closer to residual exposure surviving while indexing was blocked, and real capability can only be measured once the gate opens. Leave this out and report "share 1.1%" and the next person goes off to fix the content, when what needed fixing was one environment variable.

The remaining limits, recorded too.

• It is a weekly snapshot. Anything that rose and fell in between is invisible, and four rounds is short for claiming a trend.
• Personalisation and localisation were not controlled. Results can differ by viewer and location, so this is the screen the crawler saw, not the screen every user saw.
• Absence was not treated as proof of absence. Something missing from the parse may still have been there, so the raw HTML was kept whole for re-parsing later. That discipline was adopted after repeating the same mistake four times in this project.
• Only the web area was counted. Sponsored content in blog and community areas is not in this tally.

What actually changed as a result

Three things changed because of these numbers.

First, the keyword selection criterion changed. We used to rank by search volume, but the slot data showed that the higher the volume, the stronger the domains holding those slots. So the ranking was rebuilt as volume times winnability. Winnability works as ranking material rather than as a filter that excludes.

Second, the monitoring metric moved from total to composition. Instead of "share is N percent" we watch "keywords gained and lost this week." The total sits still while this moves every week.

Third, branded and non-branded are read separately. Mixed together, branded masks long-tail movement.

If you take only one thing from this, take this.

On a channel that gives you no data, building the instrument is the first half of the job.

This is an advanced entry in the Digital Marketing Analytics series. The layer this sits inside is laid out in measuring off-site exposure channel by channel, and off-site clicks are covered in tracking links and MMPs. The full map is in the five-layer measurement audit.

Sources

• Search Console's four metrics (impressions, clicks, CTR, position) and query-level breakdown: Google Search Console Help, Performance report
• Naver exposure and click report scope, top 30 keywords by clicks, no exposure ranking, 90-day retention, roughly one week lag, ads and VIEW excluded: Naver Search Advisor, Content Exposure and Clicks
• How robots.txt and noindex block indexing: Google, Block indexing with noindex
• All figures in the body: own SERP tracker measurements (77 to 78 keywords, four rounds on 2026-07-07, 07-08, 07-13 and 07-20, stored per slot)

Service names, competitor domains and search terms have all been replaced with pseudonyms or category labels. The figures are measured but carry the conditions in the section above, and in particular they come from a period when indexing was blocked, so they must not be read as content performance. Search results can be personalised and localised, so these numbers are the screen the crawler saw.