Deal Sourcing Pipeline: What to Log and When to Review It
A deal sourcing pipeline you can actually track needs three things: stages defined by an exit rule rather than a label, a source record detailed enough to separate one kind of referral from another, and a pass reason written on the day of the decision. Add a fixed review cadence and the pipeline starts to answer the question that matters: which sources produce companies your fund ends up backing.
This post gives a stage table, a source format, a reason list and a review calendar. It uses one of the few published data sets that follows real deals through the funnel, so you can check your own numbers against it.
What a real funnel looks like
Out of every 760 or so deals screened, roughly one gets funded. That chain comes from a University of St. Gallen project by Wolff, Hess, Grichnik and Wincent, which had data access to about 14,500 deals screened by early-stage European venture firms from 2017 to 2020, including 1,362 evaluated in more detail.
| Stage | Reported by the study | Share of all deals | Deals needed per investment (our arithmetic) |
|---|---|---|---|
| Screened | Few-minute check that rejects more than 90% | 100% | About 760 |
| First evaluation | 1,362 deals scored in detail | About 9% | About 71 |
| Pitch to the GPs | 6.2% of first evaluations | 0.6% | About 4.4 |
| Funded | 22.6% of pitches | About 0.1% | 1 |
The study rounds the result to one in 1,000. Multiplying its stage figures gives about one in 760. Either way, the working use is capacity. At these rates, a fund making eight new investments a year would need about 6,100 screens, 570 first evaluations and 35 pitches. That is an illustration, not a target. A fund that sources narrowly will screen far less and convert far more.
Treat these numbers as a sanity check, not a benchmark. The sample is European, early-stage and pre-2021. The paper does not say how many firms took part, and it gives no limitations section. It also reports two deal counts (14,549 and 14,459) and two rates for reaching the pitch (6.2% and 6.7%). Its repository record dates it 2021 and marks it as not refereed. Your own conversion rates are the only ones you should plan on.
Define each stage by its exit rule
A stage is only trackable if two people would put the same company in the same place. "Interesting" fails that test. "A partner has agreed to a first call" passes. These are the stages we would start from, adapted from the sequence in the study. The entry and exit rules are our recommendations.
| Stage | A company enters when | It leaves when | Must be logged to leave |
|---|---|---|---|
| Identified | Someone records a company against the mandate | It is screened, or moved to Watch | Source record, first-seen date |
| Screen | Someone has read enough to judge fit | It passes or is rejected | Decision, reason code, screener |
| First evaluation | A call or meeting with the founders is held | The team agrees to pursue or pass | Scorecard, interest record, next action |
| Diligence | The partners decide to pursue | The work is finished or a blocker ends it | Open questions and who owns them |
| Pitch and decision | The founders present to the GPs | The fund decides | Decision, reasons, independent votes |
| Watch (a status, not a stage) | The company fits but is too early or not raising | Its re-check date arrives or a trigger fires | Re-check date, trigger, owner |
Watch matters most for a fund that sources early. A company found through early signals may have no round to evaluate. If it sits in Screen, your stage-age figures fill with companies nobody can act on. If it is rejected, you lose it. A separate status keeps it visible without distorting the funnel.
Log the source in three fields, not one
Source is the field most pipelines flatten, and the data shows why that is costly. The St. Gallen authors split deals into quality deal flow, which the funds classed by source, and everything else. What happened to that group changes by stage:
| Point in the funnel | Quality deal flow | Comparison |
|---|---|---|
| Share of deals in | 23% of original deals | |
| Share of deals passing screening | 60% | About five times the pass rate of other deals (our arithmetic, assuming both shares describe the same pool) |
| Invited to pitch after first evaluation | 8.0% | 6.7% for the average evaluated deal |
| Reached the pitch, deals passed on by co-investors after their own process | 62% | 6.7% for the average evaluated deal |
| Funded after the pitch | No significant difference | Quality "equalizes irrespective of source" |
Two findings should change how you log. First, source buys attention at the top of the funnel and fades after that. The authors found that a simple referral from another VC is not enough. Second, the group that performed was narrow. According to the practitioners the authors interviewed, these deals were recommended by frequent co-investors and had already been through a similar screening and evaluation process with them. Some of these deals skipped screening and first evaluation and went straight to short-term diligence before the pitch.
A single "Referral" dropdown hides that difference. Use three fields instead:
- Channel: own sourcing, referral, co-investor, or cold inbound.
- Named source: the person or firm, so you can count outcomes per referrer.
- Prior work: whether the source has already evaluated the company, and how far it got.
Write the reason before you know the outcome
Log the pass reason on the day you decide, and do not edit it later. In Baruch Fischhoff's hindsight experiments, people who learned an outcome changed what they judged relevant and were largely unaware of the shift. Fischhoff argued this limits the ability to learn from the past. A reason reconstructed a year later, after a passed company has raised well, will suffer the same problem.
Use a short code list plus one sentence of free text. The first two codes below are the ones the study found most common at screening.
| Reason code | Covers | What a rising share tells you |
|---|---|---|
| No venture case | Market too small or crowded, business model does not scale | Sourcing is reaching the wrong kind of company |
| Outside mandate | Sector, stage, geography | Channel or search filters need tightening |
| Team | Founder fit or gaps | Worth reading with the scorecards |
| Traction or timing | Too early, re-check later | Candidates for Watch |
| Terms | Price or structure | Your view of the market may be out of date |
| Lost | The founders chose another investor | Speed or access problem |
Record who wanted the deal as well. In Kerr, Lerner and Schoar's study of angel groups, Tech Coast Angels kept a count of members expressing interest in each deal, across more than 2,500 ventures, and CommonAngels polled members after pitches. That record is what later let researchers examine the decisions. The Tech Coast Angels count was not a financial commitment, only a belief that the deal should be pursued further. A venture fund can keep the same kind of record: a count of partners who would take the next step, entered before group discussion. At the pitch stage, the St. Gallen funds had GPs score deals anonymously and independently, which protects the record from the loudest voice in the room.
Require fields at the gate, and keep the list short
Screening takes minutes, so logging at that stage has to take seconds. Ask for more only as a deal moves forward. The stage table above is built that way.
Incomplete records are common even in careful work. Of the 1,362 scorecard evaluations in the St. Gallen data, 427 (31%) offered a full picture of all evaluated criteria. The paper does not say whether that gap came from the funds or from the research, so read it as a warning rather than a measure. If a deal can move forward with blanks, your conversion rates by source will rest on the minority of deals that happen to be complete. Making a field required to change stage is the simplest fix.
A review calendar
Each review should have one question and one output. The cadence below is a working default, not research. Adjust it to your deal volume.
| When | Question | Look at | Output |
|---|---|---|---|
| Weekly | What is stuck? | Deals past your stage time limit, and deals with no next action or owner | Each one moved, assigned or closed |
| Monthly | What is due? | Watch list items at their re-check date or with a trigger | Re-open or re-date |
| Quarterly | Which sources are working? | Conversion by channel and named source, and the mix of reason codes | One change to where the team spends time |
| Annually | What did we miss? | Passed companies that later raised well, read against the reason logged at the time | Changes to the mandate, rubric or codes |
Two rules keep these numbers honest.
Measure by cohort. Group deals by the month they entered and follow them forward. A snapshot of this month's stage counts mixes old and new deals and flatters whichever stage is backed up.
Respect small numbers. A fund with 35 pitches a year has too few for a meaningful rate by source. Report counts over a rolling four quarters, and act on differences that persist.
Where software helps
Everything above can run in a spreadsheet or a CRM. Software earns its place at the top of the funnel, where finding companies early and keeping them visible takes the most manual effort.
Juno is our ready-to-use deal sourcing software. It helps investors discover emerging companies through early building activity, investigate founder and company context, keep companies on a radar and track outreach.
If your fund's sourcing logic or outputs need to be its own, we build a fund-specific discovery system around your mandate and review process instead. Either way, investment decisions stay with your team.