N-Gram Analysis for Healthcare Google Ads (2026): Cut 15-35% Waste

2026-06 update — n-gram analysis with Responsive Search Ads

Google's full transition to Responsive Search Ads (RSA) in 2024–2025 obscured per-asset performance — a single ad campaign now serves dozens of asset combinations and Google's reporting aggregates them. The n-gram framework has been adapted for this environment.

N-gram analysis layer 1 — Search Term Report. Words and phrases customers actually typed to reach the ad. The baseline n-gram input. The conversion rate per n-gram identifies high-converting phrase patterns and exhaust phrase patterns.

angryturtle/07-top-keywords.png" alt="Angryturtle Top Keywords — every keyword the listing appears for, with impressions, rank, and click share" loading="lazy" decoding="async" style="width:100%;height:auto;display:block;">

N-gram analysis layer 2 — RSA asset combination. Cluster the RSA serving combinations and identify which headline + description pairs co-occur in high-converting sessions. The asset-level n-gram is the input to the next RSA asset refresh.

N-gram analysis layer 3 — negative keyword surface. N-grams that consistently appear in high-spend non-converting sessions become the next negative-keyword expansion list.

The three layers together typically identify 15–35% spend recovery within 60 days of first n-gram audit.

A single-word analysis of your search terms is not enough. The pattern analysis — which two-word or three-word combinations are generating the most spend versus the most conversions — reveals the real waste in healthcare Google Ads accounts.

This is N-Gram Analysis. ICG runs it weekly across every Google Ads account we manage. The patterns it surfaces are not visible from a standard search-term report — but they are typically responsible for 15–35% of healthcare account waste.

What N-Gram Analysis Is

An "n-gram" is a contiguous sequence of n words. For search-term analysis:

• 1-gram: individual words ("clinic", "doctor", "treatment", "hair")
• 2-gram: two-word combinations ("hair transplant", "dermatologist near", "IVF cost")
• 3-gram: three-word combinations ("best IVF clinic", "hair transplant cost", "skin doctor delhi")

Standard search-term analysis looks at full search queries. N-Gram analysis looks at the patterns inside those queries — and identifies which sub-patterns drive spend, which drive conversions, and where the gap between the two creates waste.

Why This Matters for Healthcare

Healthcare searches have a specific intent pattern:

• The highest-intent searches are typically 3–4 word phrases ("best IVF clinic in Gurgaon", "cost of hair transplant Delhi", "PRP treatment for hair loss")
• The highest-spend terms in most accounts are 1–2 word broad queries triggered by phrase or broad match
• The gap between these two — high-spend 1-grams that do not convert; low-spend 3-grams that do — is the structural waste

N-Gram Analysis reveals this gap precisely. Without it, the search-term report shows you "what is being searched" but not "which sub-patterns are bleeding".

Running Your Own N-Gram Analysis

The procedure:

• Export 90 days of search-term data from Google Ads (Reports → Search terms → Download CSV)
• In Excel or Google Sheets: split each search term into n-grams (use SPLIT and array functions or run through a Python script)
• Pivot by n-gram: aggregate spend, clicks, conversions, conversion value per n-gram
• Sort by spend descending. Then re-sort by conversion rate ascending. The top 20 of each list are your priority action items.

The Four Patterns You Will Find

A Real Example From an ICG Account

A hair transplant centre in Delhi was spending ₹4.2L/month on Google Ads. Standard search-term review showed reasonable performance. N-Gram Analysis revealed:

• "hair fall" 2-gram: ₹68,000 spent, 11 conversions. CPL ₹6,180.
• "hair transplant cost" 3-gram: ₹14,000 spent, 19 conversions. CPL ₹737.

"Hair fall" was being triggered by broad-match keywords intended to capture "hair transplant" patterns. The cost 3-gram pattern was underfunded by a factor of 5. The restructure: "hair fall" added as negative phrase; "hair transplant cost" 3-gram pattern given dedicated ad groups, exact match, raised bids. CPL dropped from ₹2,140 to ₹920 in 6 weeks. Conversion volume increased 31%.

Action Framework

• Run N-Gram weekly — search-term patterns shift faster than monthly cadence catches
• Use 1-gram for negative-keyword discipline; 2-gram for match-type discipline; 3-gram for opportunity expansion
• Restructure when 3-gram conversion patterns diverge from keyword groupings — the existing structure is hiding the actual intent

Healthcare N-Gram Benchmarks We See Across ICG Accounts (2026)

After running n-gram sweeps across 40+ healthcare Google Ads accounts in the last 12 months — IVF, dental, aesthetic, ophthalmology, multi-specialty hospitals — the waste patterns cluster in a way that’s hard to ignore. Standard search-term reports flag single tokens. N-gram analysis catches the 2-word and 3-word phrases that quietly drain 15–35% of monthly spend before anyone notices.

Here is the median waste share we typically surface on the first n-gram pass of a mature account (90+ days of data, ₹3L+ monthly spend):

Common Mistakes We See in Healthcare N-Gram Audits

• Negating single words too aggressively. Blocking "free" without checking 2-grams like "free consultation" kills legitimate lead intent in dental and aesthetic accounts.
• Ignoring 3-grams. "cost of IVF" behaves nothing like "IVF cost calculator" — the first converts, the second window-shops. Only trigram analysis separates them.
• Running n-grams once and forgetting. RSAs mutate weekly. A monthly cadence is the floor; we run it fortnightly for accounts above ₹10L/month.
• Skipping the compliance filter. Healthcare queries containing "cure", "guaranteed", or "100%" are DCI/MCI grey-zone triggers — flag them for creative review, not just negation.

If Google Ads is one leg of your funnel and Meta is the other, our Meta Catalyst IQ team runs the same n-gram discipline on Meta search terms and ad-copy tokens. For competitor-side intelligence — what phrases they’re winning on — Prism Spy pulls creative-level Meta Ads data. And if you want the whole paid-media operating layer stitched together, that’s what our Client Elevation Programme exists for. Prefer to just talk it through? WhatsApp Rohit, our Business & Growth Co-Founder, for a 15-minute account read.

Which N-Grams Should Healthcare Advertisers in India Negative-Match for NMC and DPDP Compliance?

Meta Catalyst IQ Audience Size analysis showing the fatigue and saturation curves for each audience segment in a Meta Ads account" width="1200" height="675" loading="lazy" decoding="async" style="width:100%;height:auto;display:block;">

Three families of n-grams need to leave a healthcare Google Ads account the day you find them: NMC-flagged promise terms, DPDP-sensitive disclosure phrases, and off-formulary drug-name queries. Across 42 healthcare accounts ICG audited in 2026, these three buckets together account for 18-27% of unqualified spend roughly Rs. 8-14 per click that never converts and, worse, invites regulatory attention before it converts.

NMC-flagged promise n-grams

The NMC Professional Conduct Regulations and the Drugs and Magic Remedies (Objectionable Advertisements) Act, 1954 restrict guarantees, superlatives, and cure claims in medical advertising. Search queries triggering your ads that contain "guaranteed cure", "100% success", "best doctor in [city]", "no.1 hospital", "permanent solution", or "miracle treatment" pull the responsive search ad engine into auto-stitched headlines that can invert your compliance posture in a single auction. Add these unigrams and bigrams guaranteed, permanent, miracle, painless, no.1, best, cheapest as campaign-level negatives. In a Delhi multi-specialty account we audited last month, that single move cut Rs. 42,000/month of wasted spend and killed 11 auto-generated headline variants the algorithm had assembled from user queries.

DPDP-sensitive disclosure n-grams

The DPDP Act, 2023 classifies health information as sensitive personal data. When search queries carry self-disclosed conditions "diagnosed with", "my HbA1c", "stage 3", "positive report", "biopsy result" and those queries fire your call-tracking or lead-form ads, you inherit consent-and-purpose obligations most intake stacks have not scoped. Negative-match these n-grams unless your consent flow captures granular, purpose-linked consent under Section 6. It is cheaper to lose the click than to explain the pipeline to a Data Protection Board notice.

Specialty waste benchmarks we see in Indian healthcare accounts

ICG's 70-30 Google Ads engagements run this compliance sweep as a fixed, monthly deliverable the audit sits alongside the performance n-gram sweep but on a separate calendar, because compliance risk is binary while performance is a threshold.

Mini-FAQ

Q: How often should we sweep for compliance n-grams versus performance n-grams?
Compliance sweeps monthly, performance sweeps fortnightly. NMC and DPDP terms are binary one auto-generated headline built from a flagged query is enough so they belong on a fixed calendar. ICG's 70-30 managed accounts run the compliance sweep on the first working day of every month; performance n-grams get a lighter mid-month pass keyed to CPQL drift.

Q: Does ABDM or ABHA query data flow into our Google Ads keyword report?
No. ABDM traffic sits behind the ABHA login and does not surface as search-term data in your Google Ads account. What you should watch is the language overlap: patients who searched "book abha hospital" or "ayushman card hospital" on Google will land with the same intent as ABDM users, so treat "abha", "abdm", and "ayushman" as their own n-gram cluster when you segment reports the CPQL, close rate, and payor mix behave differently from cash-pay queries.