SEM & Paid Advertising
Search, Performance Max, Shopping and Microsoft Ads engineered for profit — not clicks, not impressions, not lead counts.
Most Google Ads accounts optimise to the wrong number. Maximize Conversions chases the cheapest lead instead of the qualified one, broad match leaks budget into queries that never buy, Performance Max runs as a black box cannibalising brand search, and Smart Bidding optimises against a conversion signal that's missing 30–40% of the real data thanks to iOS and ad blockers. The dashboard shows a healthy reported ROAS while your margin quietly bleeds. The pattern repeats across accounts: a CPC that doubled in 18 months, a search-terms report nobody reviews, conversion actions that count form-fills instead of booked appointments, no Microsoft Ads layer despite the 8–12% incremental volume at lower CPC, and no negative-keyword discipline at the account level. We rebuild the account on three principles: profit-first bidding (POAS, not ROAS), structural discipline (negative keywords, query mining, intent clustering, fenced PMax) and closed-loop measurement (Enhanced Conversions for Leads, offline conversion import, server-side CAPI) — so Smart Bidding finally optimises on real money and scaling decisions follow facts instead of platform-reported fiction.
- Profit-bidding
- Creative velocity
- Multi-channel scale
The mechanism behind the numbers.
No black box. Here's exactly how we run it.
Account forensic audit + waste map (week 1)
We audit every campaign, ad group, keyword, search term, asset, audience and conversion action. The output is a waste map: queries to negate, broken conversions, mis-attributed PMax spend, overlapping ad groups, redundant assets, and the top 10 highest-ROI fixes ranked. Typical accounts have 15–30% of spend that can be reclaimed in week one without touching budget.
Profit-first bidding — POAS, Enhanced Conversions, value rules
We push margin and lifetime value into Google via Enhanced Conversions, conversion value adjustments and value rules — so Smart Bidding optimises to Profit on Ad Spend, not gross revenue. For lead-gen accounts, offline conversion import from your CRM closes the loop to booked appointments and closed deals. The algorithm stops bidding on cheap, low-value form-fillers.
Search structure: intent clusters, SKAGs where they matter, RSAs
Tight intent clustering by buyer stage (research / comparison / purchase / branded). SKAGs reserved for the top 20 highest-value head terms; everything else runs in themed ad groups for Smart Bidding signal density. RSAs with pinned headlines for compliance-sensitive verticals, full ad-strength optimisation, and asset reporting tied to conversion value — not click-through rate.
Performance Max — fenced, not autopilot
PMax runs with account-level brand negative keywords, asset-group splits by product margin or audience theme, customer-list signals, account-level negative placements, and weekly insight-tab reviews. Brand search runs in a separate dedicated campaign so PMax can't cannibalise it. Search themes used surgically. Page feed used to control coverage.
Shopping feed engineering (for ecom accounts)
Title structure rebuilt to lead with the highest-converting attributes for each category. Custom labels for margin band, seasonality and bestseller status. Supplemental feeds plug missing GTIN, brand and attribute gaps. Item-level exclusions for unprofitable SKUs. Supplemental rules for promo and shipping. Coverage often doubles inside 30 days.
Negative-keyword discipline + standing search-term reviews
Account-level shared negative-keyword lists across irrelevant industries, competitors and informational queries that don't convert. Weekly search-terms review (automated via N-gram script then reviewed by hand) with negation rules. Match-type discipline. Brand vs non-brand campaign separation. This is the single most under-done piece of Google Ads management — and the cheapest source of immediate margin.
Microsoft Ads layer + closed-loop reporting
Microsoft Ads runs as a 15–20% incremental layer on the same query strategy at lower CPC (older, higher-income demo on Bing, Edge, Outlook, LinkedIn placements). Weekly Loom plus Looker Studio dashboard reports blended MER, CAC by channel, query coverage, impression share and waste reclaimed — tied back to closed revenue, not platform-reported.
What this actually looks like at 11pm on a Tuesday.
You manage $120K/month of Google Ads. The platform reports a 4.1x ROAS. Your CFO reports a 2.3x. Both are right — they are measuring different things and your account architecture is hiding which is closer to truth. Inside the account: Maximize Conversions chasing the cheapest form-fill, Performance Max running on autopilot eating brand search at a 12x reported ROAS that is just stealing organic clicks, broad match keywords matching to job-seeker queries, search terms report nobody has reviewed in 6 weeks, 47 conversion actions configured (4 of them duplicates, 11 firing on the wrong event), Smart Bidding optimising against a conversion signal that's missing 38% of events thanks to iOS and ad blockers, no Microsoft Ads layer despite the documented 18% incremental volume at lower CPC on your demographic, no negative-keyword discipline at account level so the same $12 'free download' query has been bleeding $400/week for 8 months. Your media buyer is competent but firefighting. The previous agency burned $34,000 on a PMax 'test' with no brand exclusions that cannibalised $19,000 of brand search at a real net cost. Smart Bidding has 'learned' on bad data for 14 months and the campaigns will not be fixed by tweaking. The account needs to be rebuilt on three pillars: profit-first bidding tied to closed revenue, structural discipline (negatives, intent clusters, fenced PMax), and a measurement stack the algorithms can actually trust.
The hidden wrinkles in this work.
The unglamorous edge cases that decide whether the engagement works or not.
- PMax brand exclusions require Google rep request via Smart Bidding Insights form — not in UI — and most reps don't know how to file it
- Maximize Conversion Value with target ROAS needs 30 days of stable conversion data — switching bid strategy resets the learning phase and costs 2 weeks of performance
- Microsoft Ads import from Google duplicates ad copy that may violate Microsoft's policy (different disapproval rules) — needs manual cleanup
- Negative-keyword lists in Google have a 5,000 keyword limit per list and 20 lists per account — large accounts hit this and need consolidation
- Performance Max audience signals are suggestions, not targeting — most managers think they constrain delivery, they don't; account-level negatives are the only real lever
- Enhanced Conversions for Leads requires GCLID stored in CRM AND uploaded within 90 days OR matched via hashed PII — most clients don't have GCLID capture
The step-by-step we actually run.
No black box. Real tools (sGTM, Klaviyo, GHL, BigQuery, Looker), real sequencing.
01 — Account forensic audit + waste map
Audit every campaign, ad group, keyword, search term, asset, audience, conversion action. Output: ranked waste map of queries to negate, duplicate conversions, mis-attributed PMax spend, redundant ad groups. Typical week-1 reclaim: 15-30% of spend without touching budgets.
02 — Conversion action cleanup + Enhanced Conversions
Consolidate to 2-4 primary conversion actions per business model. Implement Enhanced Conversions for Web (hashed user-provided data) and Leads (GCLID capture at form submit + offline upload). Verify dedup against GTM/GA4. EMQ-equivalent match rate >70%.
03 — Profit-first bidding (POAS via value adjustments)
Push margin-weighted conversion values via supplemental feed (custom_label for margin band) or value rules (audience-level adjustments). Switch bid strategy to Maximize Conversion Value with tROAS calibrated to margin. Smart Bidding now optimises to profit per order.
04 — Search structure: intent clusters
Tight clustering by buyer stage: research / comparison / purchase / branded. SKAGs reserved for top 20 head terms; everything else in themed ad groups for signal density. RSAs with pinned headlines for compliance, full asset reporting tied to conversion value.
05 — PMax fencing + asset-group split
Account-level brand negative keywords filed via Google rep request. Asset-group splits by product margin band, audience theme or customer-list signal. Customer-match signals uploaded. Account-level negative placements applied. Brand search in a separate dedicated campaign so PMax can't cannibalise.
06 — Shopping feed engineering (ecom)
Title rewrite leading with highest-converting attributes per category. Custom labels 0-4 for margin band, bestseller, seasonality, return rate, COGS band. Supplemental feeds via Feedonomics/DataFeedWatch for missing GTIN/brand/attributes. Item-level exclusions for unprofitable SKUs. Coverage typically doubles in 30 days.
07 — Negative-keyword discipline + N-gram analysis
Account-level shared lists for irrelevant industries, competitor terms, informational queries that don't convert. Weekly N-gram script (Python + Google Ads API) surfaces wasteful 2-3 word patterns; reviewed by hand for negation rules. Brand vs non-brand campaign separation.
08 — Microsoft Ads incremental layer
Import from Google, clean policy violations, run on the same query strategy at 30-50% lower CPC. Bing, Edge, Outlook, Yahoo, LinkedIn placements. Older, higher-income demo. Typically delivers 15-20% incremental volume.
09 — Offline conversion import (lead-gen)
Booked appointments + Qualified opportunities + Closed-won deals pushed back to Google via OCI or Zapier. Hashed PII match + GCLID match where available. Smart Bidding now optimises to closed revenue, not form-fills. CAC typically -25-41% over 90 days.
10 — Weekly Loom + monthly strategy
Friday 15-min Loom: what scaled, what got cut, what's next. Monthly 60-min strategy with founder/CMO + media buyer reviewing channel allocation, CAC by source, and 90-day pacing. Looker Studio dashboard live and shared.
What this typically produces.
Week 1: 15-30% waste reclaimed via negatives + structural fixes without budget change. Month 2-3: CAC -25-41%, blended ROAS up but importantly contribution margin up materially. Month 3-6: Microsoft Ads layer adding 15-20% incremental volume. Lead-gen accounts see SQL conversion rate +30-90% as Smart Bidding learns on closed revenue, not form-fills. Account-level CPC typically down 12-22% as quality score lifts.
A day in the life of your team.
What changes Monday morning when this engine is live.
Media buyer opens Google Ads Monday: PMax campaign at '14.7x ROAS' that's just stealing brand search. Search-terms report has 4,200 unreviewed queries from 6 weeks. Maximize Conversions chasing $9 form-fills from job seekers. CFO asks 'why is CAC up 40%?' — buyer has no answer. Account has 47 conversion actions, 11 broken, nobody knows which Smart Bidding is using.
Media buyer opens Google Ads Monday: PMax fenced and producing real incremental revenue (verified via brand-search holdout test), search terms reviewed weekly via N-gram script, Smart Bidding optimising to Maximize Conversion Value with margin-weighted values, 3 primary conversion actions tied to closed-won via offline import. CFO opens Looker dashboard: blended MER, CAC by channel, contribution margin — answers itself.
The worst-case scenarios we plan for.
Every engagement has a worst day. Here is how we handle the ones we have seen.
Conversion volume drops 60% in week 2 after consolidation
Cleanup removed 8 duplicate conversion actions Smart Bidding was double-counting. Reported conversions halve overnight; CFO panics. We show: actual closed-revenue tracking is unchanged, just no longer inflated. Smart Bidding re-learns on clean signal in 14-21 days; CPA stabilises at correct level.
PMax brand exclusion request takes 6 weeks
Google rep slow to process the form. During the wait we run brand search in a dedicated campaign with high tCPA priority so PMax loses the auction on brand queries. PMax non-brand revenue isolates in reporting. Exclusion applied at week 6 with zero performance loss.
Microsoft Ads import flags 40 disapprovals
Different policy on certain claim language and superlatives. We rewrite copy in 2 hours, resubmit, all approved within 24h. Account goes live week 2 instead of week 1.
Real numbers from real accounts.
Outcomes, not activity reports.
Most agencies sell activity. We sell outcomes. The difference is measurement.
- Bidding optimised to POAS / closed revenue, not platform-reported ROAS
- Wasted spend cut 15–30% in week one via negative-keyword and structural fixes
- Performance Max fenced with brand exclusions and asset-group control
- Shopping feed engineered for margin-weighted query coverage
- Microsoft Ads layer adding 15–20% incremental volume at lower CPC
- Closed-loop attribution to booked appointments and paid revenue via Enhanced Conversions
Connected reading & services.
- Data Analytics & Server-Side TrackingEnhanced Conversions, sGTM and offline-sync — the measurement layer SEM runs on.
- DTC Ecommerce GrowthShopping/PMax + Klaviyo + CRO — the full DTC stack SEM plugs into.
- High-Ticket B2B / B2C Lead GenerationGoogle Search + PMax feeding a scored, attributed B2B pipeline.
- SEO & AEOOrganic dominance reduces brand-defence spend on paid and lowers blended CAC.
- POAS vs ROAS — why profit should run your buyingThe measurement principle behind every SEM scaling decision.
- How to audit your Google Ads account in 7 stepsThe exact audit framework we run in week one.
- Case study: -41% CAC in 90 daysPMax fencing + POAS bidding + offline conversion import.
Questions, answered straight.
What is POAS and why does DigiMinds optimise to it instead of ROAS?
POAS (Profit on Ad Spend) measures the contribution margin a campaign produces, not just gross revenue. ROAS can look healthy while you scale unprofitable SKUs or low-quality leads. We push margin and value data into Google Enhanced Conversions and value rules so Smart Bidding optimises to profit — and scaling actually grows the bank balance.
How fast do Google Ads results show up?
Account hygiene and waste reduction land in week one — typically 15–30% of spend reclaimed without touching budget. Structural POAS gains compound over the first 60–90 days; this is where we average a 41% CAC reduction across the book.
Do you manage Performance Max campaigns?
Yes — but properly. PMax runs with account-level brand negative keywords (via Google rep request), asset-group splits by margin or audience, customer-match signals, separate dedicated brand campaign so it doesn't cannibalise, and weekly insight-tab reviews. We do not leave PMax on autopilot — that's how budgets disappear into Display placements.
Do you also run Microsoft / Bing Ads?
Yes, as a 15–20% incremental layer on the same query strategy. Microsoft Ads (Bing, Edge, Outlook, Yahoo, LinkedIn placements) typically delivers 30–50% lower CPC than Google with a higher-income, older demographic. For most accounts it's free volume nobody is bidding on.
What does the SEM engagement cost?
Google Ads management starts at $2,500/month for accounts under $25K/month spend and scales with managed spend. Includes account rebuild, ongoing optimisation, Microsoft Ads layer, Enhanced Conversions setup, weekly reporting and monthly strategy. Book a free 30-minute Profit Audit for a custom quote.
Will you work with my existing tracking or rebuild it?
We audit first. If tracking is sound we work with it. If conversions are missing 20%+ of events, double-counting across platforms, or firing on the wrong actions (form-fill vs booked appointment), we rebuild — usually as part of the Data Analytics & Server-Side Tracking engagement, which pairs naturally with SEM.
Related work & reading.
See SEM & Paid Advertising in action and the thinking behind it.
Your current agency is comfortable with your results. We're not.
48-hour turnaround · info@digiminds.org · +92 371 4232502