The situation
Nirad sells SD-WAN software licences to large Indian enterprises. Search volume in the category is small, sales cycles are long, and one real lead matters a lot. In mid-2025 two things happened at once: a consultant rebuilt the Google Ads account from scratch, and the website was relaunched. Cost per lead climbed, cost per click climbed, and eventually the leads stopped entirely.
By the time the account reached me, the consultant was gone, the account had sat idle for five months, and a restart in early 2026 was performing around ten times worse than before the gap on its biggest campaign. The two simultaneous changes also meant nobody could say which one had broken things.
How I diagnosed it
Instead of tuning bids, I treated it as an audit. I set up direct Python access to the Google Ads API so I could pull the full history rather than click through the interface, then worked through a structured sequence:
- Timeline first. I mapped the cost-per-lead curve against the rebuild date, the website relaunch and the five-month gap. The rebuild had wiped Quality Scores, Smart Bidding history and audience learning in one move, and the gap then reset whatever had started to recover.
- Check what a conversion actually is. The main conversion action wasn't firing on form submissions. It fired on visits to the About Us page, and it had logged 466 conversions over the account's life. Smart Bidding had been optimising toward people reading the About page.
- Read the search terms. Broad-match keywords were matching consumer searches for 4G and 5G routers. That traffic was cheap, plentiful and irrelevant to a company that sells enterprise software licences.
- Follow the spend. Several campaigns, including regional campaigns outside India and a competitor-keyword campaign, were spending steadily with zero conversions.
- Check where clicks land. A key landing page was orphaned with no internal links, most visitors left from the homepage, and the contact page barely registered engagement.
Search term audit
Representative examples of the two query types found, not exact search terms from the account.
What I changed
- Fixed measurement before anything else. I created a new conversion action that fires only on the thank-you page after a real form submission, and retired the one counting About page visits.
- Paused what was bleeding. Every campaign spending with zero conversions was paused. That cost nothing in real leads.
- Restructured to three campaigns, one per job. Search captures buyers already looking for SD-WAN. Display puts Nirad in front of system integrator partners. Remarketing brings back visitors who read the solution pages but didn't enquire. All three run within the existing monthly budget.
- Separated the two audiences. The account had been trying to reach end customers and system integrator partners with the same campaigns. End customers now get high-intent Search and remarketing. Partners get Display, targeted at a custom audience built from the integrator database, running alongside direct outreach.
Before Campaign sprawl
After Three campaigns
The restructure, at a glance.
Execution by channel
With measurement fixed, each campaign got its own job, audience and measure of success.
Search: capture buyers already looking
- Exact and phrase match only, on commercial-intent SD-WAN queries. Broad match came out.
- Consumer router searches blocked with negative keywords, after they had been soaking up cheap, irrelevant clicks.
- Every conversion now counted on a real form submission, so Smart Bidding learns from leads, not page views.
| Search | Before | After |
|---|---|---|
| Click-through rate | 3% | 7% |
| Average cost per click | ₹110 | ₹32 |
| Conversion rate | 0.9% | 1.4% |
Cost per lead is measured across the whole Ads account, not per campaign: total spend on Search, Display and Remarketing divided by the MQLs Ads produced. It fell from ₹30,000 to ₹10,000.
Display: reach integrator partners
- A custom audience built from Nirad's database of system integrators, rather than the open web.
- Three messaging concepts (revenue opportunity, Made in India for public-sector bids, and competitive displacement) A/B tested across three standard sizes: nine banners from a creative brief I wrote.
- The goal was to start partner conversations, not explain the partner programme, since commercial terms are handled by phone.
| Display | Before | After |
|---|---|---|
| Impressions | 140K | 53K |
| Click-through rate | 3.3% | 14% |
| Clicks | 4.6K | 7.4K |
Impressions fell on purpose. Budget moved out of Display and into Remarketing, and the tighter integrator audience still lifted clicks by about 60%. CTR counts link clicks only.
Remarketing: a new campaign for warm visitors
- The account had no remarketing before. I introduced it to bring back visitors who read the solution pages but didn't enquire, with product-specific ads.
- Cheaper, warmer traffic that supports Search rather than competing with it.
| Remarketing New campaign | Result |
|---|---|
| Impressions | 48K |
| Click-through rate | 19% |
| Clicks | 9K |
| Average cost per click | ₹16 |
CTR counts link clicks only.
The decision I pushed back on
The original plan included running ads to recruit integrator partners from a database of about 4,000 companies. I argued against making that a primary channel. If a partner ignores a cold email, they'll ignore a display ad faster. What makes an integrator pay attention is the offer, not the channel. Ads earn their place as a warm-up layer, so that when sales calls, the partner has already seen the name a few times.
Result
Cost per lead fell by two thirds, and the leads are now real. Google Ads produces about 15 marketing-qualified leads a month, part of 23 across all sources, and 10 of those 23 (about 43%) become sales-qualified and go on to product demos. Before the fix, the account's own reporting couldn't say whether it was working at all.
What I'd do next
Give the new conversion action enough volume to make Smart Bidding reliable again, then refresh the ad creative, which hadn't changed in a long time.