01 — The finding everything else follows from
Most of our visibility is below the top three
Where our non-brand impressions actually sit, last 28 days. This is the finding the click-through numbers were hiding.
| Where we appear | Queries | Share of impressions | Share of clicks | CTR |
|---|---|---|---|---|
| Positions 1–3 | 8,652 | 11.7% | 42.2% | 0.51% |
| Positions 4–10 | 21,486 | 47.4% | 43.0% | 0.13% |
| Positions 11–20 | 10,134 | 27.7% | 10.1% | 0.05% |
| Positions 21+ | 10,335 | 13.2% | 4.7% | 0.04% |
customer service, ai call center, call recording, cx — all sit at position 9–12.
A position-1 click-through rate computed from a handful of tiny queries is not a stable number and should not have been the headline. The stable finding is the distribution above: we are visible mostly where visibility does not pay.
The click-through rate is still low — and here is the control that proves it is real
Same domain, same brand, same search results page mechanics. The only difference is what kind of page it is and what it ranks for.
| Page type | CTR at positions 5–10 | Impressions |
|---|---|---|
| /careers | 4.73% | 204k |
| /products, /solutions | 1.24% | 60k |
| /company | 0.80% | 498k |
| /glossary | 0.18% | 1,845k |
| /info guides | 0.15% | 1,979k |
The problem is concentrated in the glossary and guide content — which is also where 3.8 million of our impressions are. Those pages rank for questions the search result itself answers, so the reader never needs to arrive.
Why this curve excludes brand queries, and why that matters
The single query nice carries roughly two thirds of all impressions at position 5. Include it and "expected CTR at position 5" becomes, in effect, the click-through rate of the word nice — a navigational query behaving nothing like a commercial one.
An earlier analysis did exactly that and consequently overstated every opportunity estimate by around 4×. Every figure on this page uses the non-brand curve.
Brand queries earn 3.68% overall against 0.14% non-brand — a 26× gap that mirrors the careers-versus-glossary gap almost exactly. People looking for NiCE find NiCE and click. The difficulty is everyone else.
02 — Where the clicks are going
Positions we hold that return nothing at all
Queries with at least 1,000 impressions where actual click-through falls below our own measured curve. "Clicks lost" is against our curve, not an industry one — so it is a conservative floor.
| Query | Position | Impressions | Clicks | Clicks lost |
|---|---|---|---|---|
| seamless customer service | 2.3 | 11,775 | 0 | 60 |
| ai call center | 10.9 | 54,342 | 5 | 24 |
| what is live chat customer service | 1.1 | 1,103 | 2 | 24 |
| which platforms integrate custom ai chatbots… | 4.1 | 7,584 | 0 | 23 |
| unified | 8.0 | 14,994 | 7 | 20 |
| call center handle time | 1.6 | 3,571 | 0 | 18 |
seamless customer service — position 2.3, 11,930 impressions, zero clicks in 28 days. Three things about it do not fit normal search behaviour:
- 99% desktop, against a site baseline of 70/30 desktop to mobile.
- SEMrush shows the search volume appearing from nothing — flat near zero for most of the year, then vertical in the last two or three months. Control terms (
no caller id,call recording) show ordinary fluctuation with no such step. - The page ranking for it is
/glossary/seamless-omnichannel-communication.
seamless customer service is not a customer-experience query at all. The results page belongs to Seamless, the food-delivery brand: position 1 is seamless.com’s terms of use, the ads read “Order the Food You Love”, and the rest is complaint sites, seamless.ai and Reddit. Our glossary page ranks second for a navigational query about ordering dinner.
Zero clicks is then exactly what you would expect. We are visible for a query whose intent belongs to another company — and that is a category of wasted impression worth looking for deliberately.
Two claims this page originally made were wrong. Corrected in place:
- It said
unifiedshowed “the identical shape”. It does not.unifiedhas volume across the whole timeline, rising gradually — a growth or seasonality curve, not flat-then-vertical. Two series that both end high are not the same series, and pairing them overstated the evidence. - It floated that some impressions may not be human. For this query that is unnecessary, and it was stated more confidently than the evidence supported. Brand ambiguity explains it without inventing bots.
03 — We are competing with ourselves
Six of our pages are fighting over our single biggest opportunity
Distinct nice.com pages ranking for the same query in one week. Brand queries excluded — nice cxone legitimately returns 78 of our pages, which is Google showing sitelinks, not a conflict. Locale variants are treated as hreflang, not competition.
| Query | Our pages | Impressions | Clicks | Best vs worst position |
|---|---|---|---|---|
| ai call center | 6 | 40,557 | 1 | 11.0 → 88.5 |
| call recording | 2 | 14,957 | 12 | 9.9 → 10.8 |
| customer service | 7 | 14,542 | 5 | 1.0 → 71.7 |
| cx | 8 | 12,812 | 18 | 7.5 → 64.9 |
| tts | 2 | 5,435 | 24 | 8.5 → 9.5 |
| rpa | 5 | 3,380 | 0 | 2.5 → 8.4 |
ai call center — our largest non-brand opportunity at 54,342 impressions — the page Google ranks highest is /glossary/what-is-call-center-ai at position 11. Our actual commercial page, /call-center-ai, sits at 23.9. The glossary is outranking the product page for the product's own term, and between the six of them they earned one click.
The same shape repeats: call recording puts the glossary within one position of /products/recording-management; cx has eight glossary entries competing; tts has two glossary pages for one term.
This is a consolidation problem, not a content-creation problem — the opposite of what a "publish more" recommendation would produce. The glossary is winning slots the commercial pages should hold.
04 — What is dying
Pages losing traffic fast enough to be worth a decision
Clicks in the last 28 days against the 28 days before, pages with at least 150 prior clicks. Measured at page level, the grain whose totals reconcile with Google's own property total.
| Page | Before | Now | Change | Position |
|---|---|---|---|---|
/glossary/live-chat | 153 | 14 | −90.8% | 7.5 → 16.7 |
/websites/nice-world-london | 2,036 | 219 | −89.2% | 5.6 → 10.2 |
/websites/nice-world | 1,842 | 323 | −82.5% | 7.6 → 8.7 |
/glossary/live-chat-support | 1,091 | 401 | −63.2% | 7.5 → 10.4 |
/workforce-management | 998 | 392 | −60.7% | 9.2 → 9.9 |
/resources/teleperformance | 317 | 144 | −54.6% | 4.0 → 4.3 |
/websites/nice-world pages are event sites; a post-event decline is expected and probably needs no action. /workforce-management is not — it is a core product page that lost 61% of its clicks while barely moving in position (9.2 → 9.9). Losing traffic without losing rank means demand moved or the result stopped being clicked, and that is the more troubling of the two.
05 — What to do about it
Three things, in the order the data supports
1 · Check the SERP features
- Look up
seamless customer servicein SEMrush and read the SERP Features column. If an AI Overview is present, that is the answer for this class of query — and Search Console can never tell us. - Then cross-reference the 848 AI-Overview-triggering keywords the earlier SEMrush audit found against our zero-click list, to size the problem instead of guessing at it.
2 · Decide what the glossary is for
- The glossary and guides hold the majority of our impressions and almost none of our clicks. Careers proves the site can convert when the page matches the intent.
- Start with
ai call center, where six of our pages compete and the glossary outranks the commercial page. Consolidate rather than publish.
3 · Stop optimising for rank alone
- 88% of our non-brand visibility is below the top three, and moving it up is expensive. The 26× gap between careers and glossary is not a ranking gap.
- Re-measure monthly — this is now tracked automatically, so a change is verifiable rather than asserted.
06 — Method and limits
Where these numbers come from, and where they are soft
Stated plainly, because the uncomfortable figures deserve the same scrutiny as the comfortable ones.
| Item | Status | Note |
|---|---|---|
| Page-level clicks (§04) | reconciled | Re-summing our stored data returns 52,458 clicks against Google's own property figure of 51,421 — 102%, the small excess being clicks attributed to more than one URL. Click figures are trustworthy. |
| Page-level impressions | do not sum | Corrected 2026-08-04. Summing impressions per page gives 16.4M against a true property total of 10.0M — 163%. Google counts one impression per URL shown, so a search displaying two of our pages counts twice, and jump-to-section anchors (/#industries alone: 702,000 impressions, zero clicks) count again on top. An earlier version of this briefing used that inflated denominator and reported a site CTR of 0.31%. The true figure is 0.51%. |
| Query-level figures (§01, §02) | ~84% coverage | Google withholds low-volume queries, and we additionally drop anything under 3 impressions a day to control storage. Ratios and rankings hold; absolute totals understate. |
| Cannibalization (§03) | measured | One full week, brand-excluded, fragment- and locale-normalised. Anchor rows re-report the parent page's impressions, so they are merged rather than summed — counting them raw inflated one query from 14,500 to 40,599. |
| The industry curve (§01) | external reference | A widely-cited public benchmark, not a measurement of our competitors. The gap is an order of magnitude, so the conclusion survives a generous margin of error on the reference. |
| Topic relevance filter | unreviewed | The list deciding which queries count as "ours" was written from the data, not agreed with the SEO/content owner. It currently keeps 55 of the top 60 opportunities. It needs review before it is trusted. |
Where to confirm each of these by hand
| Question | Where | Exactly what to do |
|---|---|---|
| Is the impression total right? | GSC | Performance → Search results. Read Total impressions off the property view. Any report quoting a larger number is summing pages and double-counting — that is the error corrected above. |
| Is an AI Overview taking the click? | SEMrush | Checked 2026-08-04, and it varies by keyword — never assume it across a group. seamless customer service shows Featured snippet, Video and People Also Ask but no AI Overview. unified does show one. Read SERP Features per keyword; GSC cannot answer this at all. |
| Whose brand does this query belong to? | any browser | Search it and look at the page. The highest-value check here, and the one that resolved seamless customer service: if another company’s brand owns the results, our impressions there are noise however good the position looks. GSC reports the query and never the competition, so it cannot tell you this. |
| Which page ranks for it? | GSC | Filter by the query → Pages tab. For this one it is a glossary entry answering a commercial phrase. |
| Why is review markup firing? | GSC | Search Appearance tab: REVIEW_SNIPPET carries 505k impressions at average position 17.2. Worth asking whether that markup belongs there. |
One thing this data cannot settle, and a question it raises
Around 37% of our search impressions are off-topic — no caller id, unknown caller, reverse phone lookup, and similar. That traffic earns more clicks than our entire contact-centre footprint combined, and it converts at a click-through rate our commercial pages would envy.
Whether that is a deliberate top-of-funnel play or an accident of old content, nobody on the technical side knows. It is excluded from every figure above, and it should be a conscious decision rather than a filter setting.
07 — Questions
Asked and answered
Is the benchmark wrong? These comparisons always flatter the person making them.
Do not lean on the external benchmark at all — it was the weakest part of the first version of this page. Two internal checks carry the argument without it.
The control. Careers pages earn 4.73% at positions 5–10; glossary pages earn 0.18% at the same positions. Same site, same brand, 26× apart. Whatever is suppressing clicks is not site-wide.
The inversion. Position 8 earns 0.18% while position 7 earns 0.12%. In a curve where rank drives clicks, a worse position cannot out-earn a better one. Something other than rank is deciding.
Could this just be that we rank for the wrong queries?
Partly, and that is why the off-topic traffic is excluded from these figures. But the queries in §02 are squarely ours — seamless customer service, ai call center, call center handle time. We hold positions 1–4 on several of them and take zero clicks. Wrong-query mix does not explain a zero.
What did this cost?
No new vendor. Search Console has no per-call charge — the entire 500-day capture used about a tenth of one day's free quota. The only new spend is a Supabase tier upgrade, roughly $25/month, needed because the history is 8.4 million rows.
The urgency was real: Google keeps about 16 months and deletes the oldest day every day. Five days were lost between deciding to capture it and capturing it. That data cannot be bought back from anyone.
What's the single next step?
Search Google for “seamless customer service” and look at what is above us. Everything in §05 is sequenced behind that one answer.
→ The same findings as diagrams (008)→ What is still blocked (006)