Linkedin A3 Capi Enhanced Conversions Compound

Length: ~1,200 words

Six months after Beacon CAPI deployment is where the compound effect becomes visible in the data.

Week 1 of CAPI deployment: the most obvious change is Meta EMQ improvement. In most healthcare accounts we onboard, the EMQ moves from 3.8–4.4 (pixel-only baseline) to 5.5–6.5 within the first 2 weeks as the server-side events begin flowing. By week 6, once the hashed customer data has enough matching volume, EMQ settles at 7.5–8.5.

The immediate impact is CPM reduction: 25–38% lower cost per thousand impressions as Meta's algorithm, now working from higher-quality matching data, finds more efficient audience inventory. This CPM reduction is visible in the account within 4–6 weeks of CAPI deployment. It is not the compound effect. It is the first-order effect.

The compound begins at month 3

By month 3, something more interesting happens. The algorithm has now been learning from consultation attendance events — not just form submission events, not just WhatsApp clicks. Beacon's offline conversion import has been sending "consultation attended" events back to both Meta and Google, matched to the original ad exposure via hashed customer identifiers.

The algorithm's audience model shifts. It has learned what a converting patient looks like — not what someone who clicks a form looks like, which is a different (and less valuable) audience segment.

In IVF accounts, this shift is measurable: the audience the algorithm favours after month 3 of consultation-attendance training has significantly higher consideration cycle indicators. More website page views before conversion. More YouTube video interactions. More time between first click and conversion event. These are the signals that identify genuine consideration-phase patients — the ones who will attend the consultation, not just click a form to see what happens.

What month 6 looks like in the data

ICG's portfolio data across accounts with 6 months of Beacon deployment:

Meta EMQ: 7.5–8.5 (stable, vs 3.8–4.4 at baseline) CPM: 25–38% below baseline Consultation attendance rate: +4–8 percentage points (the algorithm is finding higher-quality audiences who attend at higher rates) CPQL: 30–45% below engagement-start baseline

Google Enhanced Conversions parallel data: Conversion match rate: 85–92% (vs 55–65% with tag-only tracking) Smart Bidding Target CPA achievement rate: +28% (algorithm achieving target CPQL more consistently) Lead quality score (attendance rate indicator): improving month-on-month as the algorithm learns from offline consultation events

The compound continues

At month 12, the accounts with the most complete Beacon deployment — server-side WhatsApp capture, offline consultation attendance import, 180-day attribution window configuration — show further improvement:

The algorithm has now seen 12 months of consultation-quality conversion data. Its audience model is substantially better than it was at month 3. CPQLs in IVF accounts at month 12 average ₹1,040–1,180 — well within ICG's portfolio target, and significantly below the ₹2,120–2,400 baseline at engagement start.

The compound effect is not primarily from campaign optimisation. It is from algorithm quality improvement — the same spend, finding better-quality audiences, generating higher attendance rates, at lower CPQL.

Why this does not happen with pixel-only tracking

A pixel-only account with the same campaigns, the same creative, the same budget, running in parallel, would show improvement from campaign structural changes (better keywords, better landing pages). But it would not show the algorithm quality compound — because the algorithm never receives the consultation-attendance signal that teaches it what a converting patient looks like.

Pixel-only tracking shows Meta and Google that someone submitted a form. Beacon shows them that a specific patient attended a consultation. These are different signals that produce different algorithm behaviour.

The compound effect at month 6 and beyond is only achievable with full Beacon deployment — server-side events, offline consultation import, and the appropriate attribution window per specialty.

→ Healthcare Growth Operating System