Optimising Paid Campaigns

The campaign that went from 0.4× to 8× ROAS in 90 days

In 2022, an online education company launched Google and Meta campaigns for a £297 professional development course. Month 1 ROAS: 0.4× (spending more than they earned). The operations director wanted to shut everything down.

The paid media manager asked for 90 days. She had a methodical approach:

Days 1–30: Identified where the money was going. Found: 60% of Google budget was in one broad-match keyword cluster with a 0% conversion rate. Three of five Meta ad sets had never generated a single purchase. Paused all zero-conversion campaigns. Reallocated budget to what was showing early conversion signals.

Days 31–60: Tested three new landing page versions against the original. Version C (with a testimonial video above the fold) had a 4.2× higher conversion rate. Deployed Version C across all campaigns. Built a retargeting sequence for website visitors.

Days 61–90: Scaled the two Google ad groups converting below target CPA. Built lookalike audiences from the 47 purchasers in the Pixel. Launched a 3-ad retargeting sequence for add-to-cart abandoners.

Month 4 ROAS: 8.1×.

Not new campaigns. Not new creative (mostly). Systematic optimisation of what was already running.

(Illustrative scenario based on patterns common in online education advertising. Specific figures are representative of real-world outcomes — not a verified account of a specific named company.)

The optimisation mindset

Optimisation is a cycle, not a destination:

The most important rule: Don't optimise before you have data. The temptation to make changes in the first week (when results look bad or strange) is powerful — but changes restart the algorithm's learning phase and invalidate the data you've gathered. Allow campaigns to run for minimum 2–4 weeks and minimum 50+ clicks before drawing conclusions.

The diagnostic framework: where to look first

When a campaign underperforms, work through the funnel from top to bottom:

Step 1: Is the ad being shown? (Impression share)

Check impression share in Google Ads: what % of available auctions is your ad appearing in? If impression share is low (< 40%):

• Budget may be too low (bidding is limited by daily spend cap)
• Quality Score may be too low (Google isn't showing your ad because it's less relevant than competitors)
• Targeting may be too narrow (search volume for your keywords is lower than expected)

Step 2: Is the ad being clicked? (CTR)

• Google Search: CTR below 2% suggests the ad copy isn't matching user intent or isn't compelling enough
• Meta feed ads: CTR below 0.8% suggests creative isn't stopping the scroll
• Google Display: CTR below 0.1% is normal; focus on view-through conversions

Fix: improve ad copy or creative. For Google, ensure the headline includes the keyword. For Meta, test a new hook.

Step 3: Are clicks converting? (Conversion rate)

This is the most frequently broken part of the funnel. If you're getting clicks but not conversions:

• Landing page doesn't match the ad promise (continuity failure)
• Landing page loads too slowly (every extra second of load time measurably reduces conversion rates — research from Google/Deloitte and Akamai estimates 1–8% reduction per second depending on baseline speed, industry, and context)
• Landing page isn't mobile-optimised
• The offer isn't compelling (price too high, form too long, trust too low)
• Wrong audience (clicking because of interest, not because they have purchase intent)

Step 4: Is the conversion profitable? (CPA vs. LTV)

Even with a good conversion rate, if the CPA exceeds customer value, the campaign isn't sustainable. Options:

• Reduce CPA (better targeting, better ad, better landing page)
• Increase offer value or AOV (higher price, upsell at checkout)
• Increase LTV (subscription over one-time, referral programme)

The optimisation actions

Pause and reallocate: Any campaign, ad group, or ad that has received sufficient spend (5× target CPA minimum) and generated zero conversions should be paused. The budget should move to what's converting.

This sounds obvious. In practice, people rarely do it quickly enough — they "give it more time" when the data already indicates it's not working.

Keyword pruning (Google Ads): Weekly Search Terms Report review. Add irrelevant searches as negative keywords. Add high-performing searches as new exact-match keywords. Over time this makes the account increasingly precise.

Bid adjustments:

• Device: If mobile converts at half the CPC of desktop, reduce mobile bid modifier by 30–50%
• Time of day: If data shows conversions cluster between 9am–5pm, reduce bids overnight
• Location: If certain postcodes or cities convert better, increase bids for those locations

Audience refinements (Meta):

• Expand cold audiences that are converting well (1% lookalike → 2–3% lookalike)
• Introduce new interest combinations based on who's converting
• Test Advantage+ audience (let Meta broaden beyond your defined audience) once you have sufficient conversion data

Landing page testing: Landing page conversion rate improvements compound across all ad spend. A landing page improvement from 2% to 4% conversion rate halves your effective CPA across all campaigns — more impact than any single ad change.

Test: headline (most impactful), social proof placement, form length, CTA button copy, above-fold content.

Scaling: when and how to increase budget

Scaling a campaign means increasing spend while maintaining or improving efficiency. It's not simply "add more budget."

The scaling signals (when it's safe to scale):

1. Consistent ROAS above target — for at least 3–4 weeks
2. Sufficient conversion volume — the algorithm's smart bidding has data to maintain performance
3. Proven creative — you know which ads are driving results, not guessing

How to scale (gradually):

• Increase daily budget by no more than 20–30% at a time
• Allow 7–10 days after each budget increase before evaluating performance
• Large budget increases (doubling or tripling) destabilise smart bidding and often cause temporary CPA spikes

Google Ads scaling options:

• Increase daily budget on proven campaigns
• Add new keyword themes (new ad groups for related searches)
• Expand geographic targeting (add new cities or regions)
• Bid for position 1 on winning keywords

Meta Ads scaling options:

• Increase budget on winning ad sets (20–30% at a time)
• Expand cold audience lookalike from 1% to 2–3%
• Duplicate winning ad sets and test new audiences with the same creative
• Launch new creatives inspired by winning ads to address creative fatigue

Creative fatigue: Every ad eventually exhausts its audience. Frequency (the average number of times someone in your target audience has seen your ad) is the key signal. When frequency exceeds 3–5 for a Meta ad and performance drops, the audience has seen the ad enough — introduce new creative.

There Are No Dumb Questions

"How do I know if poor performance is caused by bad creative, bad targeting, or a bad offer?"

Isolate each variable: If CTR is low (people aren't clicking), the problem is creative or targeting. If CTR is high but conversion rate is low, the problem is landing page or offer. If conversion rate is fine but CPA is too high, the problem is bidding efficiency or audience cost. The funnel structure tells you where to look — start at the top (impressions and CTR) and work down (clicks, sessions, conversions, CPA).

"When should I give up on a campaign vs. keep optimising?"

Run the maths first. Calculate: if your conversion rate improved to your desired level, would the campaign be profitable? If yes — keep optimising. If no (the economics don't work at any realistic conversion rate) — the campaign concept needs rethinking, not just the execution. A campaign with a £15 product and £40 CPC economics cannot be optimised to profitability; no amount of ad copy testing changes the unit economics.

The optimisation calendar

Structure your optimisation work to avoid both paralysis (making no changes) and chaos (changing everything constantly):

Back to the education company

The 8× ROAS wasn't a creative breakthrough or a lucky audience. It was 90 days of weekly decisions: pause the broad-match keyword cluster burning 60% of budget with zero conversions, test three landing page variations, deploy the one that converted 4.2× better, build retargeting sequences from real purchaser data. None of those actions were dramatic — but each one compounded. Month 1 ROAS of 0.4× became Month 4 ROAS of 8.1× because a methodical practitioner treated the data as a diagnostic, not a verdict. Optimisation isn't about finding the perfect campaign — it's about 90 days of not ignoring what the data is telling you.

Key takeaways

• Don't optimise before you have data. Allow 2–4 weeks minimum and 50+ clicks before drawing conclusions or making changes. Early interventions restart the learning phase.
• Work through the funnel diagnostically. Impression → CTR → click-to-session rate → conversion rate → CPA. Each stage has different causes and fixes.
• Pause losers fast; scale winners carefully. Budget reallocated from non-converters to converters improves ROAS without increasing spend. Scale budgets 20–30% at a time to maintain algorithm stability.
• Landing page optimisation has the highest leverage. A conversion rate improvement applies across all ad spend simultaneously — more impact than any single ad change.
• Creative fatigue is real. Monitor Meta ad frequency; refresh creative when frequency exceeds 3–5 and performance drops.