Private Equity Deal Origination Pipeline Stages and Metrics

Executive Summary

A private equity (PE) origination pipeline should describe three linked lifecycles: the company research universe, the contact or relationship record, and the transaction opportunity. A company enters the research universe when a source identifies a plausible entity. It becomes a screened target only after identity and required evidence have been reviewed; it becomes an outreach candidate after a named approver authorizes contact; and it becomes a live opportunity only after a defined commercial response or deal-team acceptance. This separation reflects the documented distinction between relationship records and deal stages ^[1], while published investment processes distinguish initial screening from later detailed review ^{[2] [3]}.

The stage dictionary below defines observable transitions from identification through deal-team handoff. It records unknown, excluded, parked, duplicate, stale, withdrawn, and reactivated as states or events, with reasons and dates. Generic sales guidance recommends measurable stage exit criteria ^[4], but those criteria need adaptation to a PE universe in which most identified companies are research records rather than sale opportunities. A change in mandate creates a new assessment against a versioned rule set; it should not overwrite the old decision. Docket's published research model retains sources, excerpts, and collection dates for researched answers (Source: docket.capital) and keeps corrections in record history (Source: docket.capital).

The denominator is the heart of the reporting design. Research coverage compares distinct in-scope identified companies with a documented estimated universe. Screen completion divides reviewed eligible targets by eligible targets in the same mandate cohort. Evidence completeness counts supported, acceptable answers over required fields; an unresolved answer stays visible. Outreach response uses attempted or delivered contacts according to a stated channel definition, and a deal outcome uses a separate opportunity cohort. In the fictional dashboard below, identified coverage moves from 50% to 41.7% even though the number of identified targets rises, because the estimated universe expands. This is arithmetic in a labeled hypothetical example, not an industry benchmark. Published counts illustrate why unit choice matters: the U.S. Census Bureau distinguishes small employer firms with fewer than 500 employees in 2023 from nonemployer establishments ^{[5] [6]}, and Companies House distinguishes its total from effective company register ^{[7] [8]}.

No public source here supplies a comparable, universal PE origination conversion benchmark. Published CRM activity and U.S. deal counts describe different populations; PitchBook's transaction counts do not measure the earlier target universe ^[9]. A useful operating dashboard therefore publishes its cohort date, mandate version, entity unit, event definition, and denominator with every rate. The goal is a comparable record of research and decisions, not a prediction of deal success.

Introduction and Background

A head of origination often inherits several lists bearing the same label: companies in a market map, people known to partners, messages sent to owners, banker-led sale processes, and opportunities undergoing diligence. Calling all of them the deal pipeline makes routine management questions hard to answer. A count of researched companies cannot be interpreted as deal flow; a count of emails cannot be interpreted as target coverage. The unit and event behind each count must be visible. Deal sourcing teams can use this separation to keep research work and owner contact activity in different reports.

Published sources show why a single borrowed stage list is insufficient. Salesforce's sales pipeline guide asks managers to define specific actions, milestones, and qualifications for advancement ^[10]. That is a useful design principle, yet a PE origination process also has research before any contact. Allvue's platform description separates accounts, contacts, and opportunities ^[11], and Affinity describes PE deal flow as nonlinear ^[12]. Neither vendor statement establishes a neutral industry taxonomy. Institutional examples likewise differ: one SEC-filed process describes an initial screen when opportunities are identified ^[2], while another describes review and logging of sourced materials in a proprietary database ^[13]. These are examples of organizational practice, not a reason to impose identical stages on every mandate.

This report proposes a research-first definition for a PE team configuring its customer relationship management (CRM) system, research database, and reporting. It answers when a company is identified, screened, approved for outreach, contacted, or accepted into a live deal process. It also specifies what should happen when evidence is missing, the company is duplicate, the mandate changes, or activity returns after a pause. The design assumes that the team can identify a canonical company record, record dated decisions, and keep the source behind material facts. Docket's existing research library separately covers market maps, screening criteria, entity identity, and evidence operations (Source: docket.capital); the focus here is the event and denominator crosswalk that connects them.

The proposed thresholds and fields are an operating template. Each investment mandate must decide its own hard exclusions, permissible uncertainty, approval authority, and meaning of engagement. The template deliberately avoids a universal conversion target: a sector map, a relationship campaign, and a banker-led process start with different populations and produce rates that cannot be compared without common definitions.

Three Lifecycles, Three Objects

The research lifecycle asks what companies exist in a defined market, which are relevant, and which facts support a screening decision. Its primary object is a canonical company or target, preferably keyed to an identified legal entity and linked to brands or operating units. A market map is a candidate generator and denominator exercise: Docket's market-mapping guide states that coverage is a ratio requiring an estimated universe and its basis (Source: docket.capital). Its deduplication process resolves each candidate to an entity and website before calculating the numerator (Source: docket.capital). A target may be deeply researched even when no owner has been contacted.

The relationship lifecycle asks whom the firm knows, what contact route is authorized, and what interaction occurred. Its objects are people, organizations, relationships, attempts, and replies. A PE team's existing contact can precede the current mandate, and a conversation may concern a different fund or business. Allvue says its system can track contacts, emails, conversations, meetings, and documents ^[14]; Affinity's own guide distinguishes a CRM's people, companies, and interactions from a deal pipeline's terms and approvals ^[15]. These are vendor descriptions, useful here only as evidence that the objects need different keys.

The transaction lifecycle begins when the deal team accepts a specific potential investment or sale process for assessment. It can then include commercial evaluation, indicative terms, diligence, committee review, and closing according to the firm's policy. An SEC-filed investment process describes a deal team forming after an initial screen when an opportunity merits detail ^[3]. Churchill's published secondaries process includes company and industry analysis, management meetings, and exploratory financial diligence at distinct steps ^{[16] [17]}. These sources concern particular strategies; their chronology should not be treated as a standard for every buyout mandate.

The objects connect through identifiers and retain separate statuses. One company can have several contacts, more than one mandate assessment, and multiple potential transactions over time. An owner may respond without presenting a sale opportunity; a deal may arrive via an intermediary before the firm has mapped its sector. Affinity's description of nonlinear PE deal flow supports allowing such paths ^[12]. A practical database needs links from the opportunity to the company, mandate version, originating channel, people, and source events. It must also permit a direct inbound opportunity while marking its research coverage as incomplete rather than pretending it passed every earlier research gate.

This object model prevents two common arithmetic errors. First, counting contact attempts as distinct companies inflates the market-map numerator. Second, using all identified companies as the denominator for a transaction win rate makes the rate depend mostly on how broad the researcher searched. The following stages define the research path and handoff while preserving the contact and transaction records as separate measures.

Stage Dictionary and Observable Transitions

A stage should answer four questions: what entered, what evidence was required, who decided, and what event ended it. The dictionary is a proposed template, not a representation of how a vendor or a particular fund must work. Salesforce recommends explicit exit criteria for sales stages ^[4]; Affinity documents required fields as a control for moving deals through its pipeline ^[18]. The adaptation here is to apply those controls first to a research record and then to an approved outreach or opportunity record.

Table 1 gives the stage, object, event, owner, minimum evidence, allowed next states, and denominator. A slash between next states means a choice; it does not mean that each state is an automatic progression.

The early stages should be auditable without pretending that every target will reach outreach. Docket describes its own triage step as checking company identity and fit before deeper work (Source: docket.capital); its screening guide explicitly calls for disqualifiers, scoring rules, and treatment of missing information (Source: docket.capital). An advisor's buy-side process similarly describes screening and ranking candidates before presenting a prioritized list ^[19]. Those sources support the ordering of tasks, while the precise gate in Table 1 remains a design choice for each mandate.

A company becomes identified when the source event exists, not when an analyst believes it could be a fit. It becomes screened only when the criteria were applied and a verdict was recorded. It becomes approved for outreach only when authority and message scope were recorded; it becomes contacted when an attempt was actually logged, not when a draft was prepared. Shortlist review and approval of outreach messaging should be separate recorded decisions. A company becomes a live opportunity only after the receiving deal team accepts a defined transaction case. An SEC-filed process illustrates that a team may form after preliminary screening, while another investment process puts committee discussion later, before comprehensive diligence ^{[3] [20]}. Handoff policy must therefore be explicit and local.

Exceptional States and Backflow

Pipeline reports often hide difficult records in a single lost category. That destroys useful information. Unknown means required facts remain unresolved, including the source attempts and date of the last check. Excluded means a stated hard condition was met under a particular mandate version. Parked means the target remains plausible but no action is scheduled until a trigger or review date. Duplicate means the candidate is an alias or repeat of a canonical record. These are research states, not failed deals. Keep unanswered questions visible and link duplicate records while preserving their history. Affinity also recommends explicit merge rules for CRM hygiene ^[21].

Stale is a freshness flag applied to a fact or assessment, not a new funnel stage. Its clock should depend on the field. A dated ownership source may require different refresh treatment from a stable industry classification. Withdrawn records an approved outreach or live opportunity that the team no longer pursues, with the reason and decision date. Reactivated is an event that returns a parked or withdrawn item to a new review path. It should point to the previous state and create a new assessment or contact plan rather than silently resetting the first entry date.

A mandate change is the clearest test of the data model. (Hypothetical Example) Suppose a fund changes its minimum scale or geographic boundary. The old screen remains an event under mandate version A; the new decision is a separate event under version B. If the new mandate expands the universe, coverage's denominator changes as well. A team can report a consistent same-mandate cohort, or show a bridge that attributes changes to newly in-scope companies, newly discovered candidates, identity corrections, and review decisions. It should not compare raw stage percentages across mandate versions as if only analyst performance changed.

Backflow is also ordinary. A reviewer can return a target to screening for new evidence, an outreach candidate can be parked before any attempt, and a deal team can return an opportunity for more preliminary work. PE deal flow can be nonlinear; GRI Equity's published process describes opportunities being sourced through direct team contact ^[22]. Neither establishes a universal route. Preserve the backflow event, its actor, reason, and timestamp. A current-stage snapshot is useful for the work queue, but the event history is the evidence for aging, conversion, and cohort reconstruction.

Operationally, keep two separate reason codes: disposition reason for why an item stopped or paused and evidence status for whether the underlying facts are supported. An unknown ownership field does not necessarily mean the target should be excluded; a clear exclusion does not mean the analyst failed to complete research. That distinction keeps the unknown queue visible and avoids treating incomplete work as a negative investment judgment.

Event Ledger, Roles, and Record Ownership

One mutable status field cannot reconstruct what a team knew when it made a decision. Store a timestamped event ledger and derive the current stage from the latest valid event for each company, mandate, and opportunity. Salesforce documents opportunity field-history reporting with old and new values ^[23], and Affinity describes timelines of historical meetings and emails ^[24]. Those features illustrate parts of history capture; the schema below is a recommended analytical contract, not a claim that either product supplies every field or an immutable decision log.

A compact event schema should include the following fields. The event's natural key is stable even when a display name changes:

• event_id and entity_id: unique event and canonical company identifiers; retain alias and merge links.
• mandate_id and mandate_version: the exact rule set under which the event was evaluated.
• object_type and object_id: company, contact attempt, relationship, or transaction opportunity.
• event_type, from_state, to_state: explicit action and prior state, including backflow and reactivation.
• occurred_at and recorded_at: when the action occurred and when it was entered; preserve late entry.
• actor_id and authority: researcher, reviewer, origination owner, or deal-team approver.
• source_id and evidence_snapshot_id: retained source and the answers visible at decision time.
• reason_code and free-text rationale: controlled categories plus context; record uncertainty separately.
• cohort_id and supersedes_event_id: frozen reporting group and reversible correction link.

The approval matrix should follow the work. Research operations owns candidate capture and entity resolution. An analyst completes screening evidence. A reviewer accepts or returns the screen. The origination owner authorizes contact and owns the attempt log. The deal team accepts a handoff and owns subsequent transaction stages. ILPA's due diligence questionnaire asks managers how sourcing is staffed, conducted, and documented ^[25]; an SEC-filed process describes investment committee discussion at a later diligence gate ^[20]. These are prompts for local authority design, not evidence that an investment committee must approve every outreach message.

Table 2 distinguishes the systems and records that may hold these events. It includes Docket as a direct provider of research records, using only its published product description, alongside the other record layers rather than implying that one system must replace another.

The systems may be integrated, but an integration is only as reliable as its key and event mapping. Allvue claims lifecycle tracking from pipeline through funding ^[27], and Altvia says touchpoints and documents can link to a deal record ^[28]. Those are capabilities described by providers, not proof that a particular firm's research definition has been preserved. The implementation test is whether a reviewer can reconstruct the target, mandate version, source, decision, approval, and transaction link after a merge, reassignment, or changed criterion.

Denominators and a Cohort Dashboard

A useful funnel chart needs a denominator dictionary before any percentage is calculated. Define the object unit, eligibility filter, event window, first or latest occurrence rule, and treatment of duplicates and reactivated records. A count of companies, people, messages, and opportunities must never share a denominator merely because they appear in adjacent CRM columns. Salesforce's duplicate documentation applies specifically to accounts, contacts, and leads ^[29]; this is a reminder to specify where company and opportunity deduplication occur rather than assuming one generic rule covers all objects.

For a mandate cohort, the principal research measures are:

• Coverage: distinct identified in-scope companies divided by the estimated in-scope company universe, with source, geography, date, and uncertainty range published. A register count is a bound or input, not automatically an investable-company denominator (Source: docket.capital).
• Identity resolution: candidates with an accepted canonical match divided by candidates requiring resolution. Show duplicate and unresolved counts separately.
• Screen completion: reviewed eligible targets divided by eligible targets assigned to the same mandate cohort. Freeze the cohort membership date and define whether excluded records are eligible for deeper review.
• Evidence completeness: required answered fields with acceptable support divided by all required fields for eligible records. Report unsupported answers, conflicts, and not-found outcomes as separate counts.
• Review yield: approved outreach candidates divided by reviewed targets. This is a decision mix under the mandate, not a universal quality or performance score.
• Outreach response: qualifying replies divided by valid attempted contacts, or by delivered contacts when delivery can be established. Declare whether the unit is person, company, or attempt and whether repeat attempts count.
• Stage aging: median and 90th-percentile days from entry to exit among completed stage episodes, alongside open-item age; a parked interval can be reported separately.
• Deal outcomes: process progress and closed transactions divided by a defined accepted-opportunity cohort, never by identified research records.

Table 3 is a fictional example of a dashboard. Every value is invented solely to demonstrate the calculations; no row is a PE industry benchmark. The cohort windows and mandate versions are shown because without them the percentages cannot be interpreted.

The coverage row shows a denominator bridge: the estimate rises by 400 companies while the identified numerator rises by 100. Reporting only the count of names would hide a decline in estimated coverage. The evidence row distinguishes a supported answer from a filled cell. The response row is intentionally not labeled a deal conversion rate; a reply is a relationship event, and a handoff is a separately accepted opportunity. The example's p50 and p90 values describe completed stage episodes only, so a second view is needed to reveal unresolved open cases.

A team should publish the raw numerator and denominator beside each percentage and retain the query or snapshot that produced them. This permits a later analyst to rebuild the same cohort after company merges, late-entered events, and source corrections. If the universe estimate is a range, publish a coverage interval using the high and low denominator rather than an unjustifiably precise point estimate. Docket's mapping guidance explicitly treats statistical counts and candidate names as different components of the map (Source: docket.capital).

Implementation Sequence and Controls

Start with the mandate, not the software menu. Write a one-paragraph target definition, the hard exclusions, the fields required for a screen, and the evidence accepted for each field. Docket's screening guidance separates hard conditions from scored preferences and preserves missing information as a decision input (Source: docket.capital). Then draw the entity boundaries: company, brand, site, owner, person, and potential transaction. A key that merges several businesses into one record will distort both research coverage and contact controls.

Next, define each transition as an event test:

• Identify: retain the source and discovery date.
• Resolve: accept a canonical entity match or record an unresolved state.
• Screen: retain rule version, evidence, and verdict.
• Review: record reviewer decision and any returned questions.
• Approve: record the authorized actor, target, and permitted channel.
• Contact: log the actual attempt and its delivery status.
• Engage and hand off: apply written response and receiving-owner tests.

A validation rule can make a field mandatory at a stage, as Salesforce's opportunity-stage example shows ^[30], but the team must decide which field matters and who may override it. Keep exception reasons controlled, visible, and dated.

Then configure reporting as separate views. The research view shows universe estimates, identity resolution, screen backlog, evidence completeness, and review aging. The relationship view shows approved targets, valid attempts, replies, opt-outs, and next action. The transaction view shows accepted opportunities, deal-team ownership, diligence steps, and outcomes. Intapp describes stage tracking and transaction checkpoints in its DealCloud materials ^[26]; 4Degrees says its pipelines can reflect sourcing, diligence, and closing process ^[31]. These statements establish available configuration concepts, not a reason to copy a vendor's default columns.

Finally, run a small historical reconciliation before publishing trend lines. Pick several known targets that were merged, parked, reactivated, rejected, contacted, and handed over. Reconstruct what each dashboard would have shown at the old cut-off date. If a current status cannot reproduce the old result, add or repair the underlying event. Establish a monthly control queue for:

• Unresolved identities requiring a canonical-match decision.
• Unknown evidence with an overdue research follow-up.
• Missing reason codes on exclusions and parked records.
• Stale assessments whose source window has expired.
• Unlinked opportunities lacking a canonical company or source.
• Reactivated records missing a new mandate assessment.

A weekly report can support this cadence; GRI Equity's published process explicitly logs received opportunities in a weekly deal-flow report ^[32]. The cadence itself remains a firm choice.

Data Analysis and Evidence

External data explain why a PE origination dashboard must disclose its unit and population, but they do not supply a common research-stage conversion rate. PitchBook's U.S. PE breakdown counted 4,429 deals in the first half of 2025 ^[9]. It also reported 12,552 PE-backed companies in its tracked inventory through the second quarter of 2025 ^[9]. These are transaction and portfolio measures, respectively. Neither is the number of acquisition targets a particular fund identified, found eligible, or chose to approach. The report's 75.9% add-on share of U.S. buyout volume in that quarter further shows that deal mix can change what a source calls an opportunity ^[9]. A buy-and-build team might count add-on candidates differently from platform acquisitions, so its mandate version and transaction type belong in the cohort key.

Affinity's 2026 benchmark report draws on activity from more than 300 PE firms using its platform over January 2024 through the second quarter of 2026 ^[33]. It reports a typical user sending and receiving 1,087 emails per month and adding 36 new contacts per month ^{[34] [35]}. The report also describes introductions in its own tracked relationship base, with 33.9% for the typical firm's tracked relationships ^[36]. These are platform-customer activity measures, not a representative census of PE funds or a conversion from researched targets to transactions. Its counting rule makes even a seemingly plain introduction metric specific: one email involving three people can count as two recipient-pair introductions ^[37]. An origination team should document similarly precise counting rules for attempts, deliveries, replies, and company-level engagement.

For a market-map denominator, government business statistics present a different unit problem. The U.S. Census Bureau counted 30,427,808 nonemployer establishments in 2023, up from 29,811,495 in 2022 ^[6], and said nonemployers represented 78.4% of all establishments in that year ^[38]. It separately reported 5.58 million small employer firms with fewer than 500 employees in 2023 ^[5]. Establishments are locations or business units; firms are ownership groupings. Neither national total is a ready-made universe for a narrowly defined acquisition mandate. A denominator must apply industry, size, geography, ownership, activity, and entity-boundary rules to the same unit as the numerator.

The UK Companies House register gives a second illustration. Its total register contained 5,479,045 companies at the end of financial year 2026 ^[7], while its separately reported effective register was 4,930,634 ^[8]. It recorded 815,277 incorporations during that year ^[39]. A registration stock therefore changes over time and depends on the inclusion rule. It also does not by itself establish whether an entity actively trades or meets a fund's target definition. A research team using register data should freeze the publication date and state whether its denominator is all registered entities, effective entities, or a narrower verified operator population.

These sources allow a clear conclusion: published deal totals, CRM activity metrics, and legal-entity registers each observe a different stage of the world. None supplies a valid universal identified-to-closed rate for PE origination. A local rate can be computed only after the firm fixes its entry events, entity unit, cohort dates, channel, and mandate version. The fictional dashboard above demonstrates the arithmetic; it does not infer predictive power or expected deal volume from an external statistic.

Implications and Future Directions

The immediate implication is that a pipeline configuration project is partly a data-definition project. The stage names matter less than their entry and exit events. A team should be able to open any reported number and see the target IDs, mandate version, event dates, and reason codes behind it. If it cannot, a persuasive chart may simply reflect mutable statuses or inconsistent use of the word opportunity. Salesforce's history reporting and stage validation documentation show that common CRM building blocks exist ^{[23] [30]}; the firm still has to design the research-specific record and denominator rules.

A second implication is that research throughput and outreach capacity can be managed independently. The research team can discover and review companies while outreach remains subject to relationship checks and approval. A buy-side advisor describes a sequence from target ranking to confidential introductions ^{[19] [40]}; another origination provider describes outreach to test whether identified companies are actionable ^[41]. Those examples support keeping a handoff between a research-qualified company and a contact or opportunity, without assuming that every good target should be approached immediately.

Future reports should expose versioned coverage bridges. When a thesis changes, show how many companies were newly in scope, newly discovered, reclassified, merged, or reassessed. When an outreach campaign changes channel or delivery definition, start a new response cohort. When a firm changes its handoff test, keep the old opportunity cohort intact. This does not make rates comparable by magic; it lets analysts explain why they are not comparable and construct a like-for-like subset when one exists.

The evidence standard should also move with the decision. An initial market-map candidate may need only source and identity clues; a reviewed target needs supported answers and uncertainty; an outreach approval needs authority and contact controls; a live deal needs a transaction thesis and receiving owner. Evidence review should separately test supporting material and unresolved findings. Intapp says investment committee packet requirements can be standardized ^[42]. Those are different controls at different points, and a single completeness score would mask that distinction.

Conclusion

A private equity deal origination pipeline is most useful when it distinguishes research coverage, relationship activity, and transaction process. The practical stage sequence is identified, identity resolved, minimum data collected, screened, reviewed, approved for outreach, contacted, engaged, and accepted by a deal team. Each stage needs an observable entry event, a documented exit decision, an owner, required evidence, allowed next states, and a denominator. Unknown, excluded, parked, duplicate, stale, withdrawn, and reactivated records remain visible with reasons and timestamps.

The reporting rule is simple: never publish a rate without its numerator, denominator, unit, cohort date, and mandate version. Coverage asks how much of the defined company universe the firm has identified. Screening completion asks how much of the eligible target cohort received a reviewed decision. Evidence completeness asks whether required answers have acceptable support. Contact response asks what happened to valid attempts or deliveries. Deal outcomes belong to an accepted-opportunity cohort. These measures describe different work and belong in separate reports.

A team can implement the model in more than one platform, but it needs stable identifiers and an event ledger that survives corrections, merges, backflow, and mandate changes. The resulting dashboard is an account of what the firm found, decided, and did at a known point in time. It gives origination leaders a sound basis for allocating research and follow-up work while leaving investment judgment with the people authorized to make it.