Plateau Candy

Starting point

Plateau Candy is a branding and positioning agency for B2B deep tech. The promise to its own clients: translate brilliant technology so enterprise buyers grasp the benefit in thirty seconds — not three meetings.

That promise did not apply to its own new business. Pipeline ran on network, referral and the founders’ personal contacts. The classic agency contradiction: sell positioning to the outside world, have no predictable acquisition on the inside.

• New business depended on referrals and the founders’ network, not a process
• No target profile sharper than “B2B tech startups in DACH”
• In-house assumption: decision-makers hold the title “Geschäftsführer” (managing director) and sit in companies with 20+ employees
• No data to test that assumption — and no capacity to test it manually

Solution

CegTec did not build a campaign; it built an engine that corrects its own target profile. The approach in three steps:

1. Define signals, not industries. Instead of an industry list, a weighted set of signals was put in place — observable states that correlate with need: the offering is explanation-heavy. The website reads like a technical paper, not a brand. A technical founder leads communication. The technology is validated, market perception is missing. Each signal carries its own weight, each is checkable against the company.

2. Screen against the signals, not against a list. 8,358 companies were screened against the signal set, 4,635 decision-makers identified, 4,009 of them qualified as a fit (86.5%). Outreach ran via LinkedIn and email, personalised to the respective signal — not to the company name.

3. Every reply feeds back into the target profile. Replies were classified and aggregated: which signal, which title, which company size, which channel. From these aggregates came three rolled-out plays and two refined follow-up generations. The target profile ended up being a result of the campaign, not its precondition.

Results

Over seven weeks of active sending: 3,646 decision-makers contacted, 310 replies (8.5%), of which 143 positive (3.9%). 32 decision-makers expressed concrete meeting intent.

The real yield was the corrections to the target profile:

The assumed decision-maker was the weakest. “Geschäftsführer” (managing director) got the largest volume with 1,082 contacts — and delivered 3.05% positive replies. “Founder” reached 6.02% on 615 contacts, “VP” titles hit 7.14% on smaller volume. The audience the agency had believed in replied at half the rate of the one the data found.

Small clearly beat large. Companies with 1–10 employees replied positively at 5.17%, 11–50 at 3.02%, 51–200 at 2.18%. At 200+ employees: zero positive replies from 58 contacts. The mid-market segment assumed to be the well-funded part of the market was the weakest zone.

One channel carried the result. 46 of 47 positive reply signals ran via LinkedIn, exactly one via email. An insight that reallocated the budget for everything that followed.

A wrong hypothesis cost eleven days instead of a quarter. One play tested the signal “currently hiring in marketing” on 135 decision-makers. Result: 55 explicit rejections and zero leads that reached the prospect stage. The play was shut down after eleven days. That is exactly the point of a signal-based approach — not that every signal carries, but that a wrong bet becomes visible in days, not quarters.

The learning curve was measurable. Over the runtime, the positive reply rate rose from 2.48% in the first month to 4.72% in the second — while volume simultaneously grew fivefold (525 → 2,437 contacts). Not through better copy, but because the engine knew whom it no longer needed to contact.

By the end of the runtime, a target profile existed that was built on 3,646 real first contacts instead of an assumption: founders and technical leadership at B2B deep-tech companies with under fifty employees, approached via LinkedIn, on the signal “validated technology without market perception”. Two refined follow-up plays were already built on it.