An innovation agency's startup-sourcing method, rebuilt as a system

Client

Date

US corporate innovation agency

06/2025

The challenge

A US corporate innovation agency finds startups for large companies. Its method was good and it lived in people's heads: search the space, profile what comes back, judge which candidates are worth showing a client. Every engagement started that sequence again, and the depth of any search depended on how much time the person running it had that week.

The solution

FifthRow encoded the firm's own sourcing-and-evaluation sequence as a single system. A search space goes in; sourced candidates come out already profiled to a common standard and assessed against the criteria a corporate client would apply, so a shortlist arrives judged rather than merely found.

4 rebuilds

from a search step to a full evaluation system

112 sources

behind an average single sourcing pass

Method encodedSearch, profiling and evaluation as one sequence

Sourcing depth112 sources behind an average pass

Enterprise screeningCorporate-readiness criteria encoded into the sequence

Team running it2 people, both building and running

01 — THE SEARCH STEP

The method started where the firm's own work started: finding candidates in a defined space. The first version did that one job, so it could be judged before anything was built on top of it.

Core Activities

Encoded the firm's search approach for a defined space as a first working step
Ran it against real search briefs rather than a demonstration case
Set the sourcing depth well past what a manual search would reach
Compared what came back against the kind of list the firm produces by hand
Kept the step narrow, so the next rebuild had one thing to improve

The Results

A working search step producing candidate lists the firm recognised as its own kind of output, tested on live briefs rather than accepted on a demonstration.

1 step

encoded and tested before anything was built on it

02 — PROFILING TO A COMMON STANDARD

A list of names is not a shortlist. The second rebuild added profiling, so every candidate arrived described the same way and could be set against the others.

Core Activities

Added structured profiling so each candidate returned in a consistent shape
Held every profile to the same fields, so candidates were comparable
Built the profile depth out across successive passes rather than in one specification
Pulled the evidence behind each profile from published research and public material
Ran profiling on real candidate sets from the search step

The Results

Candidates arriving described identically rather than researched to whatever depth time allowed, so a comparison between two of them meant something.

1 shape

every candidate profiled comparably

03 — ENTERPRISE-GRADE SCREENING

The firm's clients are large corporates, and a startup that is interesting is not the same as a startup a corporate can actually work with. The third rebuild put that judgement into the method.

Core Activities

Added screening for whether a candidate is credible to a large corporate buyer
Encoded the readiness criteria the firm applies on its clients' behalf
Added scouting by funding stage, so maturity became a searchable parameter
Identified candidates' intellectual property positions as part of the assessment
Kept the screen separate from the search, so either could be changed alone

The Results

A shortlist filtered on the thing the firm's clients actually care about - whether a company is ready to be worked with - rather than on how interesting it looked in a search result.

1 screen

candidates judged as a corporate client would judge them

04 — ONE EVALUATION SYSTEM

The last rebuild collapsed the sequence. Search, profiling and screening stopped being three steps run in order and became one system the firm could point at a brief.

Core Activities

Assembled the separate steps into a single evaluation workflow
Ran the assembled system on live briefs to confirm the sequence held end to end
Kept the constituent steps runnable on their own where a brief needed only part of it
Left profile generation available as a standalone for candidates found elsewhere
Handed the assembled system over as the firm's own method

The Results

The firm's sourcing method existing as a system rather than as a practice, with the individual steps still available separately, and the whole sequence runnable against a new client brief.

1 system

running the firm's whole sourcing sequence

What it proved

Over three weeks the firm's own sourcing method was built four times, each pass adding the next thing the method needed. It started as a search step, tested on live briefs. Profiling was added so candidates came back in one comparable shape. Enterprise-readiness screening was added so candidates were judged the way a corporate client judges them, with funding stage as a searchable parameter and intellectual property positions assessed as part of the read. The last pass collapsed all of it into one evaluation system. An average sourcing pass rests on 112 sources.

What the firm was left holding was its own method in a form it could run rather than recall. The sequence it had been repeating from scratch on every engagement - search, profile, judge - existed as one system pointed at a brief, with the individual steps still usable alone. Two people did the whole thing, and they were the same two who built it, which is what put the reasoning behind the method somewhere a colleague could read it instead of in the heads of the people who had always done the work.

Results

Metric	Benchmark	Result	Improvement
Where the method lived	In the practice of the people who ran it	One system, built by the people who run it	A method that can be inspected and handed over
Candidate profiling	Each candidate researched to whatever depth time allowed	One profile shape applied to every candidate	Candidates genuinely comparable
Sourcing depth	Bounded by the hours available for a given search	112 sources behind an average sourcing pass	Depth set by the method, not by the week
Corporate readiness	Judged case by case, late, on experience	Encoded as a screen inside the sequence	Readiness assessed on every candidate
Starting a new brief	The full sequence rebuilt from the beginning	One system pointed at the brief	The sequence starts from a working method

BUILD ONE STEP, TEST IT, THEN BUILD THE NEXT

Four rebuilds in three weeks beat one specification written up front. Each pass had exactly one thing to improve, which is why the method got narrower rather than broader.

A LIST OF NAMES IS NOT A SHORTLIST

Sourcing produces candidates; profiling to a common shape produces a comparison. Without the second step, whoever reads the list does the profiling in their head, differently each time.

ENCODE THE JUDGEMENT, NOT JUST THE SEARCH

The valuable part of a sourcing practice is knowing which candidates a corporate can actually work with. That judgement is encodable, and leaving it out means automating the easy half.

MATURITY IS A PARAMETER, NOT AN IMPRESSION

Funding stage turns "is this too early" from a judgement call into a search criterion, which means it gets applied consistently instead of when someone remembers to ask.

KEEP THE STEPS RUNNABLE ALONE

Collapsing a sequence into one system is only an improvement if the parts still work separately. Briefs often need the profiling and not the search.

THE BUILDERS SHOULD BE THE RUNNERS

When the two people who use a method are the two who encode it, the reasoning ends up written down where a colleague can read it. Built for them by someone else, it stays a black box.

Recent case studies

See more proof

Finding the next high-growth US metro in five days

Client

$5B+ Global Design and Engineering Firm

Date

08/2026

5 days

naming the next growth market

Making consultants 3–5× more productive on the research behind their work

Client

$30B+ Global IT Services and Consulting Group

Date

08/2026

3–5×

productivity on consulting delivery

Matter More

Let's Talk