01 — The collapse
Ten million times seen. Fifty-one thousand clicks.
One 28-day window, taken from Google's own property total — the only figure that counts each search once. Bars are proportional.
0.51% overall. The bar is not a rendering error — at this scale the click bar is 5% of the impression bar, rounded up to stay visible.
Corrected 2026-08-04: an earlier version of this page said 17.5M impressions and 0.31%. That used our per-page table as the denominator, which counts one impression per URL shown — so a search where two of our pages appear counted twice, and jump-to-section anchors counted again on top. Google's property total is the honest denominator.
What is working
Visibility. Google shows us 17.5 million times a month. We are not invisible, we are not de-indexed, and we rank on the terms we sell into. Anything framed as "we need more rankings" is solving a problem we do not have.
What is not
Conversion of that visibility into visits. Three tenths of one percent of the people who see us click. The next two panels show that this holds at every position, which is what makes it a systemic problem rather than a ranking one.
02 — The gap
Same site, same positions, 26× apart
Click-through by page type, all measured at positions 5–10 so position is held constant. Bars scaled to the highest.
Glossary and guides carry 3.8 million impressions between them. Careers carries 204,000.
What this rules out
Not the domain. Not brand recognition. Not a site-wide title template or a technical fault. When a nice.com page matches what the searcher wants, people click it at a perfectly ordinary rate — careers proves that every month.
What it points at
The glossary and guides rank for questions the search result already answers. The reader gets what they came for without arriving. That is most of our visibility and almost none of our traffic.
customer service, ai call center, call recording, cx — all sit at position 9–12. An earlier version of this page said “we rank #1 and nobody clicks.” We mostly do not rank #1 at all.
03 — The pile-up
Six of our pages, one query, one click
ai call center — our largest non-brand opportunity at 54,342 impressions. Every nice.com page Google ranked for it in one week, ordered by position.
40,557 impressions across the six. One click.
The same shape, elsewhere
cx — eight glossary entries competing. customer service — seven pages, positions 1 to 71. call recording — the glossary sits one position behind /products/recording-management. tts — two glossary pages for one term.
Why this inverts the usual advice
The instinct on seeing weak search performance is to publish more. Here, more pages is the mechanism of the failure: the glossary is absorbing the slots the commercial pages should hold. The move is consolidation — fewer, stronger pages.
04 — The slide
Clicks lost in a month
Last 28 days against the 28 before. Position shown underneath, because a page can lose traffic with or without losing rank — and those are different problems.
position 7.5 → 16.7 · lost rank
position 5.6 → 10.2 · event site, decline expected
position 7.6 → 8.7 · event site, decline expected
position 7.5 → 10.4 · lost rank
position 9.2 → 9.9 · rank barely moved — this is the worrying one
position 4.0 → 4.3 · rank held
/workforce-management is a core product page that lost 61% of its clicks while holding its position. Losing traffic without losing rank means either demand moved or the result stopped being clicked — and it is the second one that connects back to panels 01 and 02.
05 — The fork
Three explanations. One afternoon in SEMrush separates them.
seamless customer service — position 2.3, 11,930 impressions in 28 days, zero clicks. Not few. Zero.Resolved 2026-08-04 — branch C was right. The results page belongs to Seamless, the food-delivery brand. We rank second for someone else’s navigational query, so the zero is expected rather than mysterious. A was ruled out; B was overstated.
Content-format problem
The click is satisfied before it reaches us. Neither rankings nor titles recover it; the response is content shaped to be cited, and a success measure that is not sessions.
Checked: this query has Featured snippet, Video and People Also Ask — no AI Overview. But unified does have one, so this must be read per keyword and never assumed across a class.
Measurement problem
Automated queries — rank trackers, scrapers, AI agents — generate impressions and never click. If a meaningful share of our top-3 visibility is this, part of the gap is noise rather than failure.
Weakened 2026-08-04. The desktop skew and the volume step are real, but a second query we cited as showing “the identical shape” turned out to be an ordinary growth curve. Brand ambiguity explains this case without needing bots.
Intent problem — confirmed
The page ranking is /glossary/seamless-omnichannel-communication — a definition, ranking against a food-delivery brand. Not merely the wrong format for the intent: the wrong industry. Same family as the 26× careers-versus-glossary gap in panel 02, and now the confirmed explanation.
Check: GSC → filter the query → Pages tab.
What no dataset here could do was tell us what was on the screen above our result. A person looked, and it took under a minute. That is the method, not the exception — every anomaly on this page deserves the same thirty seconds before it earns a theory.
→ Full detail, method and caveats (007) → All documents (index)