Orchestration / Decision frameworks and tooling

Every week brings tweaks.

Nothing compounds.

Testing logic, optimisation roadmaps, budget guardrails, cross-channel governance. The rules that keep performance consistent when conditions shift. Usually missing entirely. We write them.

scroll
How we think

Most performance marketing isn't being optimized. It's being tweaked.

Someone notices CPA drifting and pulls a lever. The agency shifts budget when something looks off. A new audience gets tested because it came up in a meeting. Nothing is planned. Nothing builds. Tests cancel each other out because no one's tracking what's been tried.

We replace improvisation with a written playbook.

What we build

The testing, rules and guardrails your optimisation runs on.

01 · Optimization Roadmaps

Optimisation roadmap, tests that compound.

We replace ad-hoc weekly tweaks with a disciplined optimisation roadmap. We decide what variables get tested, when they get tested, and how long a test must run before making a structural pivot. Creative, audience, and landing page variations are evaluated against a statistical baseline, not a hunch. Tests build on each other rather than cancelling each other out.

test roadmapQ3 · 8 weeks
cumulativebidding+8%creative+14%landing page+19%audiencequeuedweek 1week 8each test starts from the last one's result
02 · Bidding Infrastructure & Automation

Bidding guardrails, parameters you set.

Unconstrained ad networks overpay for clicks and optimize for empty signals. In B2B, low conversion volumes often starve algorithms of the data density they need to learn. We engineer cost ceilings, map proxy signals that predict pipeline, and deploy custom scripts inside the accounts where native automation falls short. The machines run at maximum velocity, strictly inside parameters set by your commercial margins.

bid governancetarget CPA · 12 weeks
18090cost ceilingvolume floor · data densityunconstrained · CHF 215governed · CHF 172week 1week 12the ceiling is your margin, not the platform's suggestion.
03 · Cross-Platform Reallocation

Cross-channel arbitration, when platforms disagree.

Google's algorithm wants you to spend more on Google. Meta's wants you to spend more on Meta. Both will claim credit for the same conversion. Left alone, each channel scales toward its own claim. We sit above the platforms with one unified view and move budget toward where the incremental value actually is. That requires data the platforms don't show you, which is why our media plans start from external data sources and your first-party data, not from platform exports.

claimed vs incrementalsame 30 days
Google+8%Meta-24%LinkedIn+16%incrementalclaimedbudget move620 claimed between them. 410 actually recorded.
04 · Attribution Logic

Attribution rules: what counts, what it's worth, who gets credit.

Platforms count differently and ship with defaults that are too generous. We define the rules upstream: which events count as conversions, what each is worth, the credit window per channel, and how post-click and post-view stay separated in the reporting. The rules get passed back to the platforms so they optimize for the business pipeline, not cheap digital interactions. The reporting describes the business, not the channels.

credit rulesone journey
conversionMeta · viewpost-view · excludedLinkedIn · click7d window · expiredGoogle · clickcreditedday 0day 30platform defaults would have counted all three.
The playbook

Tests compound.
When the system decides what runs.

01Backlog
Every idea scored on impact, effort, confidence. The score decides what runs next, not the loudest voice in the room.
02Live
Only what passed the score. Three tests, not thirty. Each one resourced enough to read clean.
03Concluded
Result logged, uplift named, decision to scale or kill. Both are answers worth keeping.
04Scaled
Winners move into the always-on layer. The next test builds on top of what's already earning.
Testing pipeline · last 90 daysQ3
Backlog
Live
Concluded
Scaled
Backlog
tROAS vs tCPA bidding8.4
UGC creative angle7.6
LP hero rewrite7.1
Negative kw expansion6.4
Video hook (first 3s)5.8
RSA headline set5.2
Brand keyword pause3.9
Live
tROAS bidding8.4
day 11 / 18spend €34k
UGC creative7.6
day 4 / 14imp 1.2M
LP hero rewrite7.1
day 8 / 10n = 14,820
Concluded
tROAS bidding+22%
UGC angle+14%
RSA headlines+6%
Brand kw pausekilled
Carousel reorderkilled
Scaled
tROAS bidding+22%
Always-on · 64 days
UGC angle+18%
Always-on · 142 days
Compounded lift
+47% vs. baseline
scoringImpact×Confidence÷Effort=scorethe score decides, not the loudest voice
What changes on your side

Once the rules are written, the tweaks become a plan.

A documented playbook that survives rotations, agency changes, and holidays.
A testing roadmap where each test builds on the last, not against it.
Budget pacing that follows market windows, not the calendar.
Algorithms running at full speed, inside boundaries set by your margins.
Clarity on when to hold the strategy and when to intervene.
Start here

Tooling engagements depend on context.

Start with a conversation. We scope from there. If measurement needs to come first, we'll say so.

Reply within one working day · Switzerland or Hong Kong