Core Data Objects: The CRM Data Model and the Buying Group
Written by LeadScale
Every CRM arrives with the same furniture. Accounts, contacts, opportunities, a lead queue, a campaign hierarchy: the objects were in place before the first record was created, and every workflow, score, report and routing rule the team builds afterwards runs on them. Most operators never chose this model. They inherited it from platform design, and it quietly decides what the demand engine can see.
That matters because the model has a known weak point. B2B purchase decisions are made by groups of people, and the standard object set represents organisations and individuals well while representing the deciding group far less consistently. The consequences are operational: same-account leads suppressed as noise, routing that fires on one person’s engagement while ten people weigh the deal, forecasts hanging off opportunities with a single contact attached.
This article sits in the system foundations layer of the demand generation engine. It defines the CRM data model and what each core object is for, makes the case for the buying group as a data object, gives a dated account of what the platforms ship today, answers the contact-roles objection, sets out the patterns teams use to approximate a group object, and closes with the gate that decides what enters the model at all.
What the CRM Data Model Is
A CRM data model is the set of standard entities a CRM uses to represent a market: accounts for organisations, contacts for individuals, opportunities for potential revenue, leads as a pre-qualification holding area, and campaigns as the engagement layer connecting marketing activity to the rest. Every score, route and report in the system operates on these objects.
This model is a convention. It comes from CRM platform design, and it has been reproduced so widely, in documentation, in training, in every implementation partner’s default build, that it reads as the only possible way to structure customer data. The useful question for an operator is what the model under-represents. That is where the rest of this article goes.
What Each Object Is For
Here is the standard set, each object with its operational purpose.
The account is the organisation: the anchor for ownership, territory, hierarchy and firmographics. Almost everything else in the model hangs off it.
The contact is the individual, a person tied to an account. It carries job titles, roles and reachability, and it is usually where marketing permission lives.
The opportunity is the revenue event. It carries stages, amounts, close dates and an owner, and the forecast is built on it. Sales typically owns this object.
The lead is the temporary bucket: a prospect whose account affiliation has not yet been confirmed, held apart from the main model until someone or something qualifies it. Lead-to-account matching is the bridge out of that bucket, the process that works out which organisation a lead belongs to, and it has to run before lead routing can send the record anywhere sensible.
The campaign is the engagement layer: the record that connects marketing touches to the people and accounts they landed on, so that activity can later be joined to outcomes.
Those definitions cover organisations, individuals, revenue events and marketing activity. What they do not cover well is the group that makes the decision.
The Buying Group as a Data Object
A buying group, treated as a data object, is the set of contact records from one account grouped around a specific potential purchase, each carrying a role, with engagement and state tracked at group level rather than only on the individuals. It is the entity that decides a B2B purchase, and it is the newest and least consistently modelled of the core objects.
The argument for it has a specific origin. In 2017, SiriusDecisions (the B2B research firm since acquired by Forrester) rebuilt its widely adopted Demand Waterfall around what it called demand units: a buying group organised around a need that a solution can address. Its stated reason was that “demand units, not accounts, are usually the true targets” marketing and sales should have in mind. The prior, lead-centric waterfall dated to 2006; the 2017 revision moved the unit of demand from the individual inquiry to the group. Forrester sells advisory services around this framework, so read the positioning with that in mind, but the architectural point has held for nearly a decade.
The platforms have responded, up to a point. As of July 2026, the platform documentation reviewed for this article does not show a universal, first-class buying-group object in the core sales data model. The documentation reviewed was the Salesforce Winter ’20 release notes covering the opportunity contact role, the Salesforce B2B Commerce documentation for the BuyerGroup objects, and HubSpot’s association-labels documentation. What the platforms have been adding, for years now, are group constructs: Salesforce’s Winter ’20 release gave the opportunity contact role full Object Manager treatment, with custom fields, buttons and links, custom page layouts, validation rules and triggers, delivering the long-standing IdeaExchange request titled “Make Opportunity Contact Role a First Class Object”; HubSpot supports association labels, which let a contact be labelled on its relationship to a deal (a contact labelled “Buyer” on a deal, in HubSpot’s own example) and used in segments, workflows and reports, a relationship-labelling mechanism rather than a group object. Where group capability exists, deployed practice still tends to run through the opportunity object, contact-role conventions, custom objects and overlay tools rather than through a group used as the operating unit.
The practice literature points the same way. A July 2024 piece on Salesforce Ben, produced as branded content with LeanData (a routing vendor, so a party with an interest in the pattern), describes the buying-group motion as opportunity-centric: the opportunity object serves as the container holding multiple contacts from a group, and adopting the motion “often involves rearchitecting existing systems, like Salesforce”.
The strongest objection to all of this is that the CRM already has a group construct: opportunity contact roles. Forrester itself made that case. When Salesforce made the opportunity contact role customisable in Winter ’20, Forrester published a walkthrough showing its Demand Unit Waterfall could be set up with standard objects, and called the contact role “the most natural object to house buying group members”. The objection is mainstream and deserves a straight answer on the mechanics. A contact role attaches to an opportunity, so it can only hold a group once an opportunity exists; the group that Forrester’s own method says you should detect from intent signals forms before any opportunity record does. Contact roles also carry no group-level state: there is no native home for the group’s stage, completeness or engagement rollup, only fields on individual role records. And role coverage in practice is often thin. Many opportunities carry a single primary contact regardless of how many people engaged, something the same Forrester walkthrough describes and most operators will recognise from their own pipeline.
How big is the group these objects need to hold? It depends on who measured it, when, and how. The verified figures, by source and date:
| Source | Figure | Data vintage | Method and population |
|---|---|---|---|
| Challenger (formerly CEB) | “Just under a dozen” (11.2 on average), up from 5.4 across successive studies | 2019 study; retrospective published January 2024 | Survey of 667 B2B buyers on recent complex purchases, 81% above $50,000 |
| Forrester | 13 people on average; 89% of purchases involve two or more departments | 2024 | State of Business Buying, 2024 (press release, December 2024) |
| Wynter | Median of 5 decision-makers | 2024 | Vendor primary; 100 B2B SaaS marketing executives at companies with 200+ employees |
| Gartner | 6-10 stakeholders | Current guidance | B2B buying journey research on typical complex purchases |
The spread carries the design lesson: group size varies by deal, method and segment, so a buying-group object must hold variable membership rather than a fixed set of role slots.
Forrester has converted the idea into an operational method. Its six-step buying-group guidance, built on the B2B Revenue Waterfall, says to detect active buying groups from first, second and third-party intent signals, identify the members, and treat second and third leads from the same organisation as “reinforcing signals of buying intentions” rather than duplicates. The same guidance recommends moving revenue development work off the lead object and onto an object such as the opportunity, which can hold and measure touches against several individuals at once. Forrester’s e-book on the shift is titled “It Was Never About Leads”, which states the analyst position about as directly as it can be stated.
What would first-class status mean in practice? An object that groups contact records from one account around a potential purchase before any opportunity exists; a role on each membership; group-level state and an engagement rollup, so the system can report how complete and how active the group is. Something has to instantiate it: an intent-signal threshold, the first qualified lead from a new organisation, or manual creation by an operator who can see a deal forming. Membership then needs maintenance, because people join the decision late, change roles and leave; a group object with static membership decays into another list.
Until the platforms settle this, teams close the gap with three recognisable patterns. The first is a custom junction object: a purpose-built object linking several contact records to a potential purchase before any opportunity exists. The second is the contact-role proxy: disciplined use of opportunity contact roles, with custom fields for role and engagement, along the route Forrester described. The third is an overlay group: an ABM platform or routing tool maintains the group construct and syncs it down into CRM fields. These are patterns rather than vendor recommendations; they differ in how much group-level state they can hold and who maintains it.
A scope note: this argument is B2B throughout, because buying groups and opportunity objects are B2B constructs and the evidence above (Forrester, Challenger, Gartner, Wynter) is B2B research. B2C systems meet the same problem from the other direction: a household of individuals who share purchasing power, which some B2C data models represent as a household or decision-unit object. The design question is the same; the direction of approach differs.
The Operational Cost of Individual-Level Records
Where does an individual-level model leak? Start with the mechanism, because it is regularly misnamed. Deduplication tends to get the blame, and true dedupe, merging two records that describe the same person, is doing its job. What destroys buying-group evidence sits next to it: lead-to-account matching failures, and the suppression and routing rules that treat a second lead from the same organisation as redundant. “Account already has an open opportunity: disqualify.” “Existing contact at this account: recycle.” Rules like these were written to protect sales time, and they discard the very signal Forrester’s method treats as qualification evidence.
Scoring and routing carry the same assumption. A score fires on one person’s engagement, and a route sends one record to one owner, while the state of the deal lives at group level. A highly engaged individual can be a lone enthusiast with no committee behind them; a quietly engaged individual can be one member of a twelve-person group in late evaluation. Individual-level records cannot tell those situations apart, so effort follows the wrong signal in both directions.
Forecasting inherits the problem next. An opportunity carrying a single contact role reads the same as one carrying a mapped, active group, even though they represent different levels of risk. Measurement feels it last, because joining the group’s touches to a revenue outcome requires identity resolution across every object involved, the plumbing covered in measurement and identity.
Underneath sits the substrate problem: whichever objects you model, the records inside them decay. Landbase, a data vendor aggregating third-party figures (2026, with no independent methodology of its own), puts B2B contact-data decay between 22.5% and 70.3% annually depending on industry and methodology. RevenueBase, a contact-data vendor tracking changes across its own records, measured 3.6% of business emails going stale in November 2024 alone, against a traditional monthly rate of 1.5 to 2.0%. Both are vendor figures, and they point the same way. The full treatment of decay and hygiene lives in data truth and CRM hygiene; the point here is that object design and record truth are separate problems, and fixing one does not fix the other.
Emerging Objects: What Has Not Settled
The acceptance layer of the lifecycle is moving too, and it has not settled. The Sales Accepted Lead (SAL) was built for the individual record: sales reviews one lead and accepts or rejects it. Some teams are now adopting a Sales Accepted Account (SAA), an emerging object where acceptance happens at account or group level instead. The case for it is coherent: if a group decides the deal, accepting an individual record accepts the wrong unit. The case against is equally practical: another lifecycle object adds definition and governance load, the market has not converged on what an SAA means, and renaming the acceptance point does not by itself improve qualification. Both cases are live, and this article reports the debate rather than adjudicating it.
Behind the acceptance question sits a wider tension: whether qualification is a status stamped on a record or a state read from a group as it forms. That tension has its own treatments elsewhere in this cluster: lifecycle definitions and thresholds in minimum viable lead, and the state-versus-status argument in dynamically qualified lead. One governance note applies to any new object, SAA included: it needs a written definition owned and agreed between the functions that use it, and the process for reaching that agreement is covered in sales and marketing alignment.
The Model Is Only as Good as What Enters It
One part of the data-model conversation rarely makes it into the object diagrams: admission criteria are part of the object contract. Defining an object properly means defining what a record must prove before it is allowed to instantiate or join it. An opportunity is only worth forecasting if it carried real evidence when it was created; a buying-group membership is only worth routing on if the contact behind it is a real, reachable person with a lawful basis for contact. Object design and admission control are the same decision, made in two places.
That is why the capture layer belongs in this article. A model, however well designed, holds whatever the front door lets in. Validation at the point of capture, checking each record for compliance, truth and value before it writes to the CRM, decides whether the accounts, contacts and group memberships downstream are true when they arrive. The chain is short and worth stating plainly: validation happens at capture, so what enters the CRM is what passed the gate, and the buying group on the account is therefore assembled only from contacts that passed. A contact that fails the gate never writes to the CRM, so it is never in the group, and it never carries a role on the opportunity. The full argument for validating at source rather than cleaning after ingest is made in the data truth article.
Teams running account-based marketing usually feel the object gap first, because ABM assumes the group is the unit from day one. For everyone else the practical next step is an audit with three parts. Check the object definitions: does every object in your instance, custom ones included, have a written purpose and an owner? Check buying-group coverage: for your last twenty opportunities, how many contacts were attached, in what roles, and would the group have been visible before the opportunity existed? Check the matching layer: trace recent same-account leads through your suppression and routing rules and see whether they surfaced as reinforcing signals or vanished as duplicates. Where the audit traces problems back to what enters at capture, a validation layer such as the LeadScale Engine checks each record on arrival, before it writes to the CRM. Start with the audit; it tells you whether the weakness is the model, the matching or the intake, and each of those is a design decision you can then make deliberately.
Frequently Asked Questions
A CRM data model is the set of standard entities a CRM uses to represent a market: accounts for organisations, contacts for individuals, opportunities for potential revenue, leads as a pre-qualification holding area, and campaigns as the engagement layer connecting marketing activity to the rest. Every score, route, report and workflow in the system operates on these objects, which is why the model shapes what a demand engine can see and do. It is a convention inherited from CRM platform design rather than the only possible architecture.
Five objects do most of the work. The account represents the organisation and anchors ownership and territory. The contact represents the individual and carries roles and reachability. The opportunity represents the potential revenue event and carries stages, amounts and close dates. The lead is a temporary bucket for a prospect whose account affiliation is not yet confirmed, with lead-to-account matching as the bridge out of it. The campaign is the engagement layer that connects marketing touches to the other objects so activity can be joined to outcomes.
A buying group is the set of people within an organisation who collectively make a purchase decision. Its size varies by source and method: Challenger’s research found groups growing to just under a dozen by its 2019 study; Forrester’s State of Business Buying, 2024 reports 13 people on average, with 89% of purchases involving two or more departments; Wynter’s 2024 study of 100 B2B SaaS marketing executives found a median of five; Gartner’s guidance is 6 to 10 stakeholders. Treated as a data object, the buying group links several contact records from one account around a specific potential purchase, with a role on each membership.
A demand unit is SiriusDecisions’ 2017 term for a buying group organised around a need that a solution can address. SiriusDecisions (since acquired by Forrester) rebuilt its Demand Waterfall around demand units that year, arguing that “demand units, not accounts, are usually the true targets” for marketing and sales. The concept matters for data modelling because it defines demand at group level: one account can contain several demand units, each representing a distinct buying decision with its own group.
As of July 2026, the Salesforce documentation reviewed for this article does not show a universal, first-class buying-group object in the core sales data model. Salesforce’s Winter ’20 release made the opportunity contact role a customisable object with full Object Manager treatment, delivering the long-standing IdeaExchange request titled “Make Opportunity Contact Role a First Class Object”, and Forrester has described that object as the most natural home for buying-group members. The BuyerGroup objects in Salesforce B2B Commerce are storefront entitlement constructs (price books, promotions, store access), a different concept with a similar name. In deployed practice, buying groups on Salesforce are typically operated through the opportunity object, contact-role conventions, custom objects or overlay tools.
The lead object is a temporary bucket for a prospect whose account affiliation has not been confirmed, and problems start when it becomes a permanent operating unit. Because leads sit apart from accounts, a second lead from an organisation you already know can be suppressed or recycled as a duplicate, when Forrester’s buying-group guidance says it should be treated as a reinforcing signal of buying intent. Parts of the industry now argue for retiring the lead object in favour of opportunity-centric models; the conservative position keeps it as a short-lived container with disciplined lead-to-account matching out of it.
A Sales Accepted Account is an emerging acceptance object some teams are adopting, in which sales acceptance happens at account or group level rather than on an individual lead record (the Sales Accepted Lead, or SAL). The case for it is that deals are decided by buying groups, so accepting one individual record accepts the wrong unit. The case against is that it adds another lifecycle object to define and govern, and the market has not settled what it means. It is currently an open design choice rather than a standard.
Directly, because an object model holds whatever the capture layer lets in, so the truth of the records decides the value of the model. B2B contact data decays between 22.5% and 70.3% annually depending on industry and methodology (Landbase, 2026, a vendor aggregation of third-party figures), and RevenueBase, a contact-data vendor tracking its own records, measured 3.6% of business emails going stale in November 2024 alone. Validating each record for compliance, truth and value at the point of capture, before it writes to the CRM, keeps the objects worth operating on. The full treatment is in the data truth article in this cluster.
