Skip to content
SearchSynth.
Analytics & Conversion

Half your marketing works. Most reporting can't say which half.

Client-side tracking is blocked or consent-gated for a large share of visitors, AI answers send buyers who arrive already decided, and last-touch attribution quietly credits whichever channel happened to be last. We rebuild the measurement layer so the numbers are trustworthy, state plainly what still cannot be attributed, and then use it to test conversion changes rather than argue about them.

Book a strategy callSee pricing
SURFACESorganicai answersreferraldirectattribution+ surveypipelinedashed = arrives already decided, never attributed

every surface that contributes, including the ones nothing can attribute

Tracking that survives blockers and ITP
Server-sideTracking that survives blockers and ITP
Modelled, not ignored
Zero-clickModelled, not ignored
Built to the regime you operate under
Consent-safeBuilt to the regime you operate under
Conversion changes proven, not asserted
TestedConversion changes proven, not asserted

What we rebuild

Measurement first, because conversion work without trustworthy numbers is just redesign with extra confidence. Then experimentation, because opinion is the most expensive way to decide a layout.

Server-side tracking

Client-side tags are blocked, throttled or consent-gated for a large and growing share of visitors, and the loss is not random — it skews toward exactly the technical, privacy-aware buyers many B2B firms sell to. We move collection server-side so the data is complete, durable and yours.

  • Server-side container in your own infrastructure
  • First-party cookie and identity handling
  • Event schema versioned and documented in the repo

Attribution that admits its limits

We build multi-touch attribution where the data supports it, model the zero-click share where it does not, and add self-reported attribution to catch what neither sees. Then we report those separately, because merging them into one confident number is how reporting starts lying.

  • Multi-touch modelling across the paths you can observe
  • Self-reported attribution wired into forms and CRM
  • Explicit unattributed share, stated rather than buried

Consent & privacy compliance

Consent mode, regional rules and retention handled as engineering rather than a banner someone bought. Including the part most setups get wrong: what happens to data collected before consent, and whether your vendor terms actually permit what you are doing.

  • Consent mode and regional configuration
  • Retention and deletion that reach every store
  • Vendor and sub-processor review

Measurement plan & event design

One document defining every event, property and conversion — what it means, who owns it, and what decision it informs. Events that inform no decision get deleted, which is usually most of them.

  • Event and property dictionary, versioned
  • Conversion definitions agreed with sales, not just marketing
  • Deprecation plan for the events nobody uses

Conversion research

Before changing anything, we find where and why people leave: session replay, funnel analysis, form analytics, and asking customers directly. Most conversion problems turn out to be clarity or trust problems rather than layout ones.

  • Funnel and form drop-off analysis
  • Session replay and heatmap review
  • Customer interviews and on-site polling

Experimentation

Tests designed with the power calculation done first, so you know before starting whether your traffic can ever produce a result. Plenty of sites cannot, and we will tell you — sequential testing or a straight ship-and-measure is the honest answer at low volume.

  • Power analysis before a test is built
  • Server-side and split-URL testing without flicker
  • Documented result including the losers

Dashboards people actually open

One view per audience — an operator dashboard, a board pack, a channel deep-dive — each answering a specific question. Built on a semantic layer so a metric means the same thing everywhere it appears.

  • Role-specific views rather than one dashboard for all
  • Definitions shared with the warehouse metric layer
  • Automated board and investor reporting

Site performance & Core Web Vitals

Speed is a conversion lever and a ranking one. We work field data rather than lab scores, fix the largest-contentful-paint and interaction problems at the template level, and hold the gains with monitoring that fails a build when they regress.

  • Field-data diagnosis via CrUX and RUM
  • Template-level LCP, INP and CLS fixes
  • Budgets enforced in CI so gains stick

What lands

  • Server-side tracking, deployed

    Running in your infrastructure, in your repo, with the event schema documented and versioned.

  • Measurement plan

    Every event and conversion defined, owned and tied to a decision it informs.

  • Attribution model

    Multi-touch where observable, modelled where not, self-reported alongside — with the unattributed share stated.

  • Conversion research findings

    Where people leave and why, evidenced by replay, funnel data and customer interviews.

  • Experiment backlog and results

    Prioritised by expected value, with every completed test recorded including the failures.

  • Dashboards per audience

    Operator, executive and channel views built on shared definitions rather than three conflicting ones.

What we work with

Collection

  • GA4
  • Server-side GTM
  • Segment
  • Snowplow
  • PostHog

Warehouse

  • BigQuery
  • Snowflake
  • Postgres
  • dbt

Experimentation

  • GrowthBook
  • Optimizely
  • Statsig
  • Edge split testing

Qualitative

  • Microsoft Clarity
  • Hotjar
  • Customer interviews

Performance

  • CrUX
  • Lighthouse CI
  • Web Vitals RUM

Reporting

  • Looker Studio
  • Metabase
  • Cube
  • Automated board packs
Process

The first 90 days

  1. Weeks 1–2

    Audit what you have

    How much data you are losing to blocking and consent, which events are wrong, and which reported numbers currently disagree with each other.

  2. Weeks 3–5

    Rebuild collection

    Server-side tracking deployed, event schema defined and versioned, consent handling corrected. Old and new run in parallel so the change is measurable.

  3. Weeks 6–8

    Attribution & dashboards

    Attribution model built, self-reported capture added to forms, and the first dashboards replacing whatever people currently screenshot into decks.

  4. Weeks 9–12

    Research and first tests

    Conversion research on the highest-value funnel, then the first experiments — chosen for expected value rather than ease.

  5. Ongoing

    Test, hold, extend

    A running experiment programme, performance budgets enforced in CI, and measurement maintained as the site and the privacy landscape change.

Measurement engagements

The audit stands alone and frequently ends with us telling you the tracking is fine and the problem is elsewhere. That is a useful answer too.

Measurement Audit

$5,500

fixed fee · 2 weeks

How much of your data is wrong or missing, which numbers disagree, and what it would take to trust them.

  • Data-loss quantification against server logs
  • Event and conversion accuracy review
  • Consent and privacy compliance check
  • Attribution model assessment
  • Prioritised remediation plan

Anyone whose dashboards disagree with their CRM.

Get started
Most chosen

Measurement Rebuild

from$22,000

project · 8–10 weeks

Server-side collection, attribution and dashboards built and deployed into your stack.

  • Everything in the Audit
  • Server-side tracking in your infrastructure
  • Event schema and measurement plan
  • Attribution model with self-reported capture
  • Dashboards per audience
  • Handover and runbooks

Teams making budget decisions on numbers they privately distrust.

Get started

Conversion Retainer

from$6,500

per month · 6-month minimum

A running research and experimentation programme against your highest-value funnels.

  • Ongoing conversion research
  • Prioritised experiment backlog
  • Test build, run and analysis
  • Core Web Vitals monitoring and fixes
  • Monthly readout including failed tests

Sites with enough traffic to test properly and revenue that justifies it.

Get started

Prices are starting points, not quotes. Scope drives the number, and you get a fixed one before you commit.

Analytics & Conversion

Analytics & Conversion questions, answered

How much data are we actually losing?
The audit measures it rather than guessing, by comparing client-side events against server logs. The gap varies enormously by audience — a consumer retail site loses far less than a developer-tools company whose visitors run blockers as a matter of course. The important part is that the loss is not random, so it biases every channel comparison you make from it.
Can you attribute AI search traffic?
Partially, and we are careful not to overstate it. Referral traffic from ChatGPT, Perplexity and AI Overviews is identifiable where it exists. But much of the effect is zero-click — the buyer arrives via a direct visit already convinced — and no analytics package can see that. We model the gap, add self-reported attribution to catch what modelling misses, and report the unattributed share explicitly rather than pretending it is zero.
Isn't server-side tracking a way to dodge consent rules?
No, and anyone selling it that way should worry you. Server-side collection changes where data is processed, not whether you need a lawful basis for it. We build consent enforcement into the server-side layer, which is generally more reliable than the client-side equivalent. If your goal is to track people who declined, we are the wrong agency.
Our traffic is low. Can we run A/B tests at all?
Often not, and we would rather say so before taking your money. We run the power calculation first: at typical B2B volumes a classic A/B test on a 2% conversion rate needs months to reach significance, by which point the market has moved. For low-traffic sites the honest approach is qualitative research plus shipping the change and monitoring, with sequential testing where it genuinely applies.
Does this overlap with your Business Systems service?
They meet at the warehouse. Business Systems builds the reconciled data layer and the metric definitions for the whole company; this service handles marketing collection, attribution and conversion specifically. If your CRM and billing already disagree, that is a foundations problem and no amount of tracking work fixes it — the audit will tell you which situation you are in.
Will you work with our existing GA4 setup?
Usually yes. Most GA4 implementations are salvageable and a rebuild is rarely justified on its own. What we typically add alongside is server-side collection and a warehouse copy of the raw events, because GA4's sampling and retention limits are what stop teams answering the questions that actually matter.

Bring the number your team quietly doesn't believe.

Every company has one — a channel that looks too good, a conversion rate nobody trusts. We'll trace where it comes from on the call and tell you whether it's a tracking problem or a real one.

30-minute strategy call

With an engineer, not a closer

Book a strategy call

Typical reply time: under 4 business hours.

hello@searchsynth.ai