Getze Consulting · AI in Sales Test report 04 of 5

From exhibitor directoryto lead radar.

This report is the odd one out: it tests the same method on a different data problem — sales data rather than engineering knowledge. 3,707 exhibitors become 35 solid leads in a single day, each with a hook and a source. You come from engineering? Then test report 01 is the more direct way in.

Conditions of this test
  • Public sources, no customer data
  • Every assessment with a source
  • Reclassifications checked by a human

The same method in the quotation process is on Sheet 5 · Technical Distribution. Where your data sits is your decision — three deployment models are set out on Sheet 1.

Test object

IFAT Munich 2026, the entire exhibitor directory.

The world's leading trade fair for water and environmental technology. We take the complete directory as the data basis and build a lead radar for a clearly defined ideal customer.

No
Ideal customer trait
Definition
Why
Checkable from
01
Raw data set
3,707 exhibitors
The official directory, worldwide, unfiltered.
Excel, API
02
Geographic focus
DACH region
The home market. Anything else is not workable for us.
country field, postcode
03
Size class
mid-market
No corporates, no pure resellers. Nobody there decides at short notice.
brand list, size
04
Business model
manufacturer
Own production and engineering. Only there does the problem we solve exist.
profile texts
The human defines the ideal customer and checks the borderline cases. The pipeline does the volume work. The same method in a quotation: Sheet 5 →
The measurement

Reduction by a factor of 105.

1 day instead of two to three weeks of full-time research, including the quality check. Every stage reduces the volume and enriches what is left.

Seven stages, each with a checkable criterion. No stage throws anything away without it being traceable why — the intermediate sets are kept and can be reopened at any time.

STAGE REMAINING SET Raw data 3,707 DACH filter 1,812 ICP cluster 736 Manufacturer + mid-market 184 1st research wave 57 A-leads 35 High priority 24 Factor 105 EVERY STAGE EVIDENCED 45 of the 57 have open engineering vacancies · 40 with relevant news from 2025/2026 No stage throws anything away without it being traceable why. The intermediate sets are kept. REMAINING SET Raw data 3,707 DACH filter 1,812 ICP cluster 736 Manufacturer + mid-market 184 1st research wave 57 A-leads 35 High priority 24 Factor 105 EVERY STAGE EVIDENCED 45 of the 57 with open engineering vacancies No stage throws anything away without it being traceable why.
Seven stages, each with a checkable criterion The measurement · Test report 04

Classic cold-call lists reach 5 to 10 per cent relevance. The difference is not the volume but that every deletion has a reason and can be undone.

Test procedure

Six stages, human and machine in turn.

The human defines the ideal customer and checks the borderline cases, the machine does the volume research. At two stages a human explicitly decides alongside.

No
Stage
Who
What happens
Based on
01
Data acquisition
machine
Direct access via structured sources: Excel export and official API.
public
02
Geographic filter
machine
DACH reduction via country field and postcode heuristics.
master data
03
Sector cluster
machine
Product groups 001 to 013 as a filter taxonomy.
fair taxonomy
04
Manufacturer or reseller
machine + human
Heuristic classification from profile texts, borderline cases reviewed.
profile texts
05
Corporate filter
machine + human
Mid-market focus via brand list, size and corporate signals.
company register
06
Trigger research
machine
One to two targeted web searches per company: vacancies, investments, acquisitions.
press, job boards
The engine does not replace a sales person. It takes the volume work off them. All stages with prices →
Test result

This is what an A-lead looks like.

Anonymised example from the real lead radar. Every trigger has a source — otherwise it would just be another claim.

Pump manufacturer A
Tier A Priority High

Southern Germany · Hall B1 · pumps and valves cluster

TRIGGERS, EACH WITH A SOURCE
€8m invested in a new production site, groundbreaking 04/2025. Source: press release
+18 % revenue growth year on year, named a growth champion. Source: trade magazine
3D-print prototyping in design, new BIM catalogue with 220 items. Source: company website
Open position: sales engineer, currently advertised. Source: job board
DERIVED FROM THAT

A new plant at 18 % growth plus product development with 3D printing means design and engineering are running at full stretch in parallel. An ideal moment for external engineering capacity as a peak buffer.

The derivation is a proposal, not a finding — it stands or falls with the four triggers above, and those are evidenced.

The view breaks off here — the lead radar carries on
  • All 35 A-leads at the same depth, sorted by priority
  • The 149 excluded with the criterion they fell out on
  • Intermediate sets for every stage, reopenable at any time
  • Location and hall, for planning the route around the fair
  • Contacts from public sources, no purchased data
  • Timestamps per trigger: how current the hook is
  • Reclassifications a human overrode, with the reason
  • What the pipeline could not do and where it was unsure

Everything from public sources, nothing purchased. Every assessment traces back to the page it came from — and every deletion is reversible.

One lead dossier with four evidenced triggers Test result · Test report 04
Assessment

What the pipeline can do. And what it deliberately does not.

This side of the line

What the pipeline does

  • data acquisition and structuring at scale
  • multi-stage filtering along formal criteria
  • trigger research from public sources
  • prioritisation and conversation hooks as a draft
Beyond the line

What sales does

  • reclassification and refining the ideal customer
  • contact qualification and personal research
  • the sales conversation, needs analysis, handling objections
  • negotiation, closing, relationship
The engine does not replace a sales person. It makes them sharper, because it takes the volume work off them — and because they can say of every lead why this one in particular.
Full report

All of it as a PDF.

With the pipeline architecture, the complete lead example and the impact analysis. You get the PDF straight after submitting.

Further test reports

Four more process chains, same method.

All five at a glance →

Next step

We build this engine on your target group.

This sales engine is an internal tool at Getze and an offering at the same time. On request we build it for your sales team on your own target group data.

Arrange a 30-minute call →
Test object
Lead qualification · IFAT Munich 2026, exhibitor directory
Data basis
3,707 exhibitors, public sources, no customer data
Method
Six-stage pipeline · one day · borderline cases reviewed by humans
Tested by
Andrej Getze · Machinery and pump engineering
Report
Test report 04 of 5 · Revision 28.08.2026
Approval
Stays with you
Test report 04 of 5