Context: The Midrand B2B Firm Running Google Search and Meta in Parallel - and Why Its Two Dashboards Disagree
Picture a hypothetical Midrand software firm with a R150,000 monthly media budget split between Google Search and Meta. On the first Monday of the month, the Google Ads dashboard reports 84 conversions; Meta reports 61; GA4 reports 97; and the CRM shows 43 qualified sales-qualified leads. Nobody is lying. Each system is counting a different event at a different point in the journey, across different attribution windows, with different consent states applied. This is the everyday context in which a Gauteng PPC agency earns or loses its keep - not in the keyword research deck, but in whether the numbers you act on reflect reality.
This guide is operational. It walks through what a multi-channel performance team actually does week to week across Google, Meta, LinkedIn, Microsoft Advertising and Amazon, why platform-reported conversions diverge from your warehouse, and how to translate all of it into decisions a Gauteng marketing director can defend to a board.
What a Gauteng PPC Agency Is Actually Accountable For (Beyond Running Ads)
A credible performance partner is accountable for four things that sit outside the ad platforms: a clean measurement foundation, channel-role clarity, bid and budget discipline, and reporting that reconciles to revenue. Running campaigns is table stakes. The differentiation is in the plumbing and the reporting reconciliation.
- Measurement foundation: GA4 key events configured to business outcomes, Consent Mode v2 implemented, server-side tagging for Meta and Google, offline conversion imports from CRM.
- Channel-role clarity: each platform assigned a funnel job with its own KPI, so budgets are defended on role rather than last-click credit.
- Bid discipline: gradual target moves, not weekly resets that restart the learning phase.
- Reconciliation: a monthly view that maps platform numbers to GA4, warehouse data and blended MER.
Prebo Digital is Johannesburg-based, founded in May 2016 by an ex-Google employee, with 11-50 staff and an average client tenure of around four years (brand-stated). That tenure is a useful signal: PPC partnerships that survive are usually the ones where the reporting reconciled and the tracking did not quietly decay.
Does a PPC Agency Need to Be Physically in Gauteng? Proximity, Time Zone and Escalation Paths
Proximity matters for three things and almost nothing else: workshop sessions where your team and the agency sit in the same room to align on funnel roles; escalation when a campaign breaks on a Friday afternoon; and the occasional face-to-face reporting walkthrough. For everything else - build, optimisation, testing - a competent team works asynchronously across SAST, which is within an hour or two of UK and Europe business hours and offset favourably for early-morning Middle East calls.
What you should actually test is escalation speed and decision rights. Ask a prospective partner: who has login access, who can pause a campaign without approval, what the escalation path looks like, and when the weekly performance read happens. A local suburb in a page title proves nothing; a named account lead with a documented escalation path proves a lot.
The Tracking Foundation That Comes Before Any Campaign Build: GA4 Key Events, Consent Mode v2 and POPIA Alignment
Before a single campaign is rebuilt, the foundation should be audited. GA4 key events must map to commercial events - not page views. Consent Mode v2 must be implemented, which is a shift from the simpler v1 model: v2 sends consent signals for ad personalisation and remarketing alongside analytics, and consent-denied users contribute modelled conversions rather than observed ones. That modelling is useful, but it shrinks observable remarketing list sizes, which changes how you plan audience-based campaigns. The effect on list volume after implementation is a real operational consideration that should be projected, not discovered.
POPIA alignment runs alongside this. The Information Regulator of South Africa governs how personal information is processed; consent banners, CRM lead data and customer-list uploads for Meta or Google Customer Match must be lawful, documented and purpose-limited. Where data is used for advertising, the consent state should travel with it.
A common failure: an agency rebuilds campaigns on top of a broken measurement layer. Bidding then optimises toward the wrong signal, and every subsequent optimisation compounds the error. Fix tracking first, then rebuild.
Why Platform-Reported Conversions Diverge From Your Warehouse: Enhanced Conversions, Meta CAPI and Offline Imports
Google, Meta and GA4 each use their own attribution model and lookback window. On top of that, consent state reduces what each platform can observe, and modelling fills some - but not all - of the gap. Enhanced Conversions improves match rates for web leads by hashing first-party identifiers such as email; Meta's Conversions API (CAPI) pushes server-side events that are less susceptible to browser restrictions. The two require careful deduplication: if you fire both the browser pixel and the server-side CAPI event without a shared event_id, you will double-count.
// Minimal deduplication pattern (illustrative)const eventId = crypto.randomUUID();// Browser pixelfbq('track', 'Lead', {}, {eventID: eventId});// Server-side CAPI payloadawait fetch('/api/capi', { method: 'POST', body: JSON.stringify({ event_name: 'Lead', event_id: eventId })});For B2B, the bigger lever is offline conversion imports. Import sales-qualified lead and closed-won events from the CRM back into Google Ads so bidding learns from lead quality rather than form fills. Do this before scaling LinkedIn lead-gen - otherwise the algorithm optimises toward cheap leads that never close.
Tip: import offline conversions from CRM before scaling LinkedIn. Lead-quality signal materially changes the algorithm's targeting, and scaling without it builds a bigger funnel of the wrong leads.
Mapping Channel Roles Across the Funnel: Where Google Search, Shopping/PMax, Demand Gen, Meta, LinkedIn, Microsoft and Amazon Each Earn Their Budget
Each channel has a job. Budget should be defended against that job's KPI, not against a blended last-click view that conflates prospecting and demand capture.
Google Search captures existing demand - its KPI is usually CPA or lead quality. Shopping and Performance Max sit closer to product-level demand capture for e-commerce. Demand Gen and Meta prospecting create demand - measured on incremental reach, assisted conversions or new-customer share. LinkedIn serves B2B consideration and lead-gen, judged on SQL quality. Microsoft Advertising picks up desktop and B2B search where relevant. Amazon is a marketplace channel where advertising competes on the shelf against organic ranking, so the KPI is often share of voice and profitable TACoS rather than raw ROAS. Where each channel maps to the funnel is the budget argument.
Bid Strategy Discipline: Moving tCPA and tROAS Incrementally Instead of Resetting Weekly
Smart bidding learns from volume. Resetting tROAS or tCPA every Monday restarts the learning phase and produces erratic performance. Move targets in small increments - typically 5-10% at a time - and hold for at least two to three weeks before evaluating. Use seasonality adjustments rather than target resets for short promotional windows. Watch impression share alongside CPA: an aggressive target that shrinks spend is a signal, not a win.
South African Operating Realities: Load-Shedding, ZAR Billing Volatility, Payment Rails and WhatsApp as a Conversion Channel
Gauteng teams plan around realities that global playbooks ignore. Load-shedding can interrupt reporting reviews - keep offline dashboard snapshots and schedule creative reviews outside known load-shedding windows. Accounts billed by Google and Meta in USD are exposed to ZAR volatility; build a billing buffer into budget planning and review FX-sensitive spend monthly. Payment rails matter for measurement: PayFast, Yoco, Ozow, Peach, Paygate and Stitch each integrate differently, and where a purchase completes inside an iframe or a third-party checkout, your server-side event coverage changes. Audit these per gateway. Finally, WhatsApp is a genuine conversion channel for South African service businesses - measure the click that initiates the conversation as a key event, but do not process message content. Track the entry point, not the content.
In-House Hire vs Freelancer vs Managed PPC Agency: A Comparison Table Built for South African Market Realities
The decision is not which model is objectively better; it is which model matches your spend level, channel breadth and internal capacity. The table below maps the three options against the dimensions that matter once you are past entry-level spend.
| Dimension | In-house hire | Freelancer | Managed PPC agency |
|---|---|---|---|
| Cost model | Full-time salary plus tools and training; highest fixed cost | Hourly or monthly retainer; lowest fixed cost | Monthly retainer, often tiered by spend or channel count |
| Channel breadth | Deep in one platform, thin elsewhere | Varies widely; verify across Google, Meta and Amazon | Broad by design - Google, Meta, LinkedIn, Microsoft, Amazon under one team |
| Accountability | Internal; depends on manager oversight | Contractual; often informal escalation | Contracted scope, SLA and named account lead |
| Tracking capability | Depends on individual skill; server-side rare | Mixed; usually platform-side only | Server-side GA4, CAPI, offline imports as standard scope |
| Time to value | Slow to hire, then ramp | Fast start, variable quality | Fast audit lead time; builds on existing team knowledge |
Who Is This For? Matching Three Gauteng Buyer Profiles to the Right Engagement Model
Profile 1 - the DTC Shopify or WooCommerce store scaling past roughly R100,000 per month in media spend. At this level you need feed discipline, margin-aware bidding and server-side conversion tracking. A managed agency with e-commerce tracking capability is usually the right fit; an in-house hire alone often cannot cover feed, server-side, Meta CAPI and Amazon in the same week. See the CRO and tracking services as a starting reference.
Profile 2 - B2B SaaS or professional services using LinkedIn plus Google Search. The critical dependency is offline conversion imports from the CRM so bidding optimises toward qualified pipeline, not raw form fills. A managed engagement that owns the CRM-to-platform data loop is usually correct; a freelancer running LinkedIn in isolation tends to optimise toward cheap leads that never close.
Profile 3 - marketplace or FMCG sellers on Amazon with Google and Meta halo demand. The decision here is whether the partner can measure Amazon advertising in context of marketplace organic ranking while also running Google and Meta as demand-creation support. That is a multi-platform remit; a managed team is usually the fit, provided it can prove verified Amazon Ads Partner status.
The Weekly and Monthly Operating Rhythm of a Managed PPC Engagement: Audit, Build, Test, Scale, Report
A structured engagement has a predictable cadence. Weekly: search-term and placement review, budget pacing check, anomaly flags in GA4. Fortnightly: creative or ad copy refresh, test read-outs, bid-target review. Monthly: full reconciliation, cohort and MER review, roadmap adjustment. Quarterly: incrementality experiment design and channel-role reassessment. The shape is audit, build, test, scale, report - and the same cycle repeats. Prebo Digital publishes its custom reporting approach as a reference for what that cadence should look like.
How to Reconcile Platform Numbers to GA4, Your Warehouse and MER (With a Worked ZAR Example)
Reconciliation is arithmetic plus judgment. Start with blended MER - total revenue divided by total marketing spend - then decompose it. New-customer CAC should be tracked separately from blended CAC; reconciling only the blended figure lets a repeat-heavy month flatter acquisition performance.
MER = total revenue / total marketing spendNew-customer CAC = total acquisition spend / new customers acquiredIllustrative example (hypothetical, not a client result): a Gauteng store spends R140,000 across Google and Meta in a month. Google reports R360,000 revenue, Meta reports R290,000, and GA4 last-non-direct reports R510,000. The store's real revenue in the period is R540,000. MER is R540,000 ÷ R140,000 ≈ 3.86. The platform sums overstate because of overlapping attribution. The reconciliation question is not which platform is right, but which delta reflects genuine incremental return - and that requires an experiment, not a debate.
Warning: resetting a tROAS target weekly destroys the learning phase and inflates volatility. Adjust in small increments, hold for two to three weeks, and judge against MER and new-customer CAC, not platform ROAS alone.
Incrementality When Attribution Is Probabilistic: Geo Holdouts and Practical Experiment Design
Attribution estimates; experiments measure. A practical geo holdout for a Gauteng advertiser splits a primary metro (for example, Johannesburg or Pretoria) from secondary Gauteng geographies, then turns spend off in one cell for four to six weeks and compares. The design needs a minimum spend per cell to produce a readable signal - usually enough that a 10-15% lift would be statistically detectable - and the two cells need similar baseline demand. Run the experiment while everything else stays constant; a simultaneous creative overhaul contaminates the read. Repeat the experiment quarterly to rebuild confidence.
How to Verify an Agency's Claims in Under an Hour: Google Premier Partner Directory, Amazon Ads Partner Listing, Named Account Team and Reporting Walkthrough
Verification is quick and mostly public. Check the Google Ads Partner directory for Premier Partner status, and the Amazon Ads partner listing for Amazon status. Ask for a named account lead and the reporting walkthrough - the same six-page document should be reconcilable to GA4 and the warehouse. Prebo Digital holds Google Premier Partner status from 2022 to 2026 and verified Amazon Ads Partner status, and its Johannesburg base means Gauteng buyers have a direct escalation path (see contact).
