Most keyword research groups terms by volume. That is the mistake. Volume tells you how many people search something. It tells you nothing about whether those people are close to buying — and in a city-specific category, those are wildly different populations.
Why city intent is its own dimension
Someone searching “lab grown diamond” is reading. Someone searching “lab grown diamond ring price Mumbai” is shopping. Same category, completely different distance from a purchase, and only one of them is worth a high bid.
The city modifier is not just a location filter. It is an intent signal. People add a city name when they have moved from learning about a thing to working out where to get it. That shift is the most valuable moment in the funnel, and volume-based grouping buries it.
The four intent tiers
We sort every keyword into one of four groups before deciding anything about budget:
Tier 1 — Transactional, city-specific
“buy lab grown diamond ring Mumbai”, “engagement ring showroom Bandra”. Lowest volume, highest conversion rate, deserves the highest bid. This is where the money is.
Tier 2 — Commercial, city-specific
“best lab grown diamond Mumbai”, “lab grown vs natural diamond Bengaluru”. Comparing options, still local. Worth real budget; expect a longer path to conversion.
Tier 3 — Commercial, no city
“lab grown diamond price”, “is lab grown diamond real”. Genuine interest, no location signal. Worth capturing, worth bidding cautiously.
Tier 4 — Informational
“how are lab grown diamonds made”. This is content territory, not paid search. Write for it, do not bid on it.
Map clusters to structure, not the other way round
Here is where most accounts fall apart. The research gets done properly, then everything is dumped into one campaign and the structure erases the distinction you just spent a week building.
The clusters should determine the architecture:
- One campaign per city, so budget can be controlled by market. Mumbai and Bengaluru do not behave alike and should not share a budget.
- One ad group per intent tier within that city, so bids reflect distance from purchase.
- One landing page per cluster, matching the language of the cluster. A Tier 1 Mumbai searcher should land on a page that mentions Mumbai and lets them buy. Sending them to a general category page wastes the intent you paid for.
What this changed on a jewellery account
Engagement-ring intent is fiercely competitive and mostly city-specific. We built keyword clusters around lab-grown diamond and engagement-ring searches, split them by Mumbai and Bengaluru intent, and mapped each cluster to campaign structure plus landing pages, so budget followed genuine purchase intent instead of broad research traffic.
The outcome was 38% of ad waste removed and 2.4x more qualified enquiries. Spend did not increase. It stopped going to Tier 4 traffic that was never going to convert.
Negative keywords are half the job
Clustering only works if you stop traffic crossing between clusters. A few that earn their place immediately in this category:
- Job-seeker terms: “jobs”, “salary”, “career”, “vacancy”
- Free intent: “free”, “cheap”, “wholesale” (unless you are wholesale)
- DIY and research: “how to make”, “meaning”, “wikipedia”
- Competitor city names you do not serve
Review search terms weekly for the first month, then monthly. Every account we inherit has money leaking into queries nobody ever looked at.
A simple way to start
Export your last ninety days of search terms. Tag each one with a city (or none) and an intent tier. Sum spend and conversions by tier. In almost every account we have audited, Tier 4 is consuming 30–40% of budget and producing almost nothing.
That single table usually pays for the exercise.
If you would like us to run it against your account, send it over — or read more about how we approach technical and local SEO.
