Services 02 · Algorithm Training

Teach the machine to find your best customers.

Meta and Google optimise toward the events you send them. We rebuild your tracking, event taxonomy and value signals so the algorithm hunts for buyers — not clicks.

Check my signals See the signal stack

The algorithm isn't working against you. It's learning from the wrong data.

Five layers between your data and cheaper customers.

Each layer feeds the one above it. Most accounts stop at the first — then wonder why the platform keeps buying the wrong people.

05

Optimisation events

Campaigns bid toward the event that actually predicts revenue — not the one that's easiest to fire.

04

Value signals

Purchase values, predicted LTV and lead scores attached to events, so the algorithm learns who is worth more.

03

Identity & match quality

Hashed customer parameters raise event match quality, so more conversions connect back to real profiles.

02

Server-side tracking

A CAPI / server pipeline that survives ad blockers and iOS, deduplicated against the browser pixel.

01

Pixel & event taxonomy

A clean base layer: the right events, named consistently, firing once, mapped to real funnel stages.

The work, line by line.

T-01

Event taxonomy

Audit every event against the funnel it claims to measure. Kill duplicates, rename inconsistencies, add what's missing.

T-02

Server-side pipeline

Implement or repair CAPI and server containers, with deduplication verified event by event — not assumed.

T-03

Match quality

Enrich events with hashed customer data to push EMQ scores up — every point compounds into cheaper delivery.

T-04

Value-based bidding

Send purchase values and lead scores so campaigns can optimise for revenue, not conversion counts.

T-05

Audience signals

Seed lookalikes and custom audiences from your best customers — and exclude the traffic that poisons learning.

T-06

Learning-phase management

Structure and edit discipline that keeps campaigns exiting learning fast — and stops resets from burning budget.

Ninety days of clean signals.

Before
4.1 EMQ score
$92 Cost per acquisition
1.9× Blended ROAS
Stuck Learning phase
After
8.7 EMQ score
$57 Cost per acquisition
3.4× Blended ROAS
5 days Learning-phase exit

Illustrative account figures. Signal quality compounds — the gains arrive over weeks, not overnight.

Real account · B2B lead-gen startup

Then it happened for real.

We took over the account, rebuilt tracking server-side and fed real, offline conversions back to Google from the CRM. Once the algorithm could see which leads actually became pipeline, it started buying buyers — and the account scaled without efficiency breaking.

Google Ads account overview from Aug 2024 to Jun 2026 — cost and conversions climb together while cost efficiency stays flat, after server-side tracking and offline conversion import were implemented.
Google Ads account overview, Aug 2024 → Jun 2026. Spend and conversions climb together while cost efficiency stays flat — the signature of an algorithm optimising toward real customers, not cheap clicks.
~4× Conversions +858 vs. baseline
£3.14 Avg. CPC held flat (−£0.02)
6×+ Spend scaled efficiency intact
25.85% Search impr. share +15.9 pts
Ready when you are

Find out what your account taught the algorithm.

One call to review your tracking, events and signal quality. We'll tell you what the platforms currently believe about your customers — and what to fix first.

Book a call
Mor Sofer, founder of Performore
Tracking CAPI Signals Bidding