The objects
Whatever a CRM calls them, its building blocks are the same. A record type, often called an object, defines what kind of thing is stored; a record is one instance; a property or field is one piece of information on that record; and a relationship or association joins records to each other. Marketing's work in the CRM is almost entirely about getting the right values into a few fields on a few objects at the right moment.
| Object | What one record represents | Key relationships | What marketing uses it for |
|---|---|---|---|
| Contact | One person, usually keyed on an email address | Belongs to a company; associated with deals and activities | Source, consent, lifecycle stage, engagement; the unit most lists and workflows act on |
| Company (account) | One organisation | Has many contacts and deals | Account-level reporting, deduplication by domain, firmographic fit for scoring |
| Deal (opportunity) | One potential sale, with stage, amount and expected close date | Associated with contacts and a company | Revenue attributed back to a source; the end of the marketing reporting chain |
| Activity | One interaction: an email, call, meeting, note or logged task | Attached to a contact, company or deal | Evidence that follow-up happened and when; speed-to-lead reporting |
| Campaign | One marketing effort, with a name, channel, dates and cost | Associated with contacts and deals it touched | The join between spend and outcomes |
| Task | One piece of work with an owner and a due date | Attached to a record | The output of a hand-off workflow |
| List or segment | A saved set of records matching criteria | Built from any object | The audience for a send or a workflow |
Two distinctions trip up new users. First, a contact and a deal are different things: a person can have several deals over time, and a deal can involve several people, so revenue lives on the deal and the person lives on the contact. Reporting that puts revenue on the contact will double-count the moment a second buyer is added. Second, a company is not the same as the company field on a contact. A free-text company name is a string; a company record is an object that can hold its own fields and be associated with many contacts. Business-to-business reporting needs the object.
Some CRMs add service tickets, products, quotes and custom objects. They matter to marketing only where they change a contact's state, for example a support ticket that should pause a promotional sequence, and the right response is a field on the contact that the workflow can read rather than a workflow that reads the ticket directly. The tutorial on pausing marketing to customers with an open issue builds that field as a hold with a reason, an owner and a release rule.
Properties and required fields
A property has a name, a type and, usually, a set of permitted values. The types that matter are short text, long text, picklistA CRM field type that accepts one value from a defined list of permitted values, so that records can be reported on instead of typed in many spellings. (one value from a list), multi-select, number, date, true or false, and lookup (a reference to another record). The single most useful habit in CRM design is choosing a picklist wherever a free-text field would do, because free text cannot be reported on. A field called industry with values typed by hand produces dozens of spellings of the same thing; a picklist produces a report.
Required fields are enforced at two moments. At creation, the record cannot be saved without them: for a contact that is normally email, source and owner; for a deal it is name, stage, amount (or an explicit "unknown" value) and the associated contact. At a stage change, further fields become required: a deal cannot move to proposal without a decision-maker contact, or to closed-lost without a reason. The second kind is where most of the value lies, because it turns the pipeline into a record of why things happened, and it is designed alongside the stages in the pipeline guide.
Marketing should ask for a small set of properties on the contact and defend them: lifecycle stageA CRM field recording where a contact stands in the relationship, such as subscriber, lead, opportunity or customer. (the marketing view of where a person is), original source and most recent source (covered below), consent status and its date for each channel, and the engagement fields the automation platform writes, such as last email click or last page visit. Everything else is a candidate for deletion. CRMs accumulate fields the way cupboards accumulate cables, and each unused field is one more thing a report can be built on wrongly. The tutorial on auditing and documenting CRM fields finds which fields can go and retires them in stages so that nothing reading them breaks.
Naming follows the same discipline as the rest of the stack. A field name should say what it holds and, where it is written by an integration, which system writes it. The naming conventions guide sets out conventions for campaign parameters that carry straight into CRM source fields, and keeping the two aligned is what lets a campaign name in an ad account be found again on a won deal.
Source and attribution fields
Source fields answer the question every marketing report eventually asks: where did this person come from. Answering it well needs more than one field, because a person arrives more than once.
Original source records the channel and campaign of the first known touch, written once at creation and never overwritten. Most recent source records the latest touch, overwritten on each new visit or form submission. Between them they support first-touch and last-touch AttributionAssigning conversion credit to touchpoints. directly, and they are the minimum. A richer design stores every touch as its own activity record with a timestamp and campaign, which supports multi-touch models later; but a team that cannot keep two fields filled should not start with a dozen.
The fields are populated from the website. Campaign parameters (UTMURL tags identifying traffic source. source, medium, campaign, content and term) are read from the landing page URL, held in a cookie or local storage for the session, and written into hidden form fields at submission, so they arrive on the contact with the form data. Click identifiers from ad platforms are captured the same way and are the key to the integrations in the CRM integration guide. The landing page, the referrer and the form name are worth storing too; they cost nothing and they explain the parameters when the campaign naming turns out to be ambiguous.
Granularity should match the questions. Source at channel level (organic search, paid search, paid social, email, referral, direct, offline) answers budget questions; campaign level answers which message worked; content level answers which ad. Store the raw parameters and derive the channel grouping from them with a documented rule, rather than asking salespeople to pick a source from a list, because a salesperson's guess is the least reliable source value in the system and the one most likely to be "word of mouth".
Offline sources need their own values. Events, partner referrals, outbound calls and inbound phone calls all create contacts without a landing page, and they need a source picklist that sales fills at creation, with the event or partner named. Leaving them blank means every report shows a large unexplained segment that is, in practice, the sales team's own work going uncredited. For opportunities, the tutorial on writing the source rule for pipeline sets out which origins carry over from the account's first touch, which the creator picks with no default, and when the value locks.
Duplicates and hygiene
Duplicates arrive from predictable places: the same person using a work and a personal email, a list import that did not match on email, a salesperson creating a contact by hand rather than searching first, a form that creates a new record on every submission, and company names typed with and without the legal suffix. Each duplicate splits a person's history between records, so engagement looks lower than it is, workflows may fire for both records, and the one sales is working may not be the one marketing is nurturing. When subscribers start receiving the same email twice, the diagnosis for subscribers receiving duplicate emails separates duplicate records from the other causes, such as two systems sending or a flow letting people in twice. When the merge queue refills every week with the same kinds of pair, the diagnosis for duplicate contacts in the CRM groups the pairs by the entry point that created the second record.
Prevention is cheaper than merging. Match on email for contacts and on website domain for companies at every entry point: forms, imports, integrations and manual creation. Where a CRM supports matching rules, set them to update an existing record rather than create a new one, and make the exception (create anyway) require a deliberate choice. Domains alone do not settle every case, and the guide to matching contacts to company records covers personal addresses, shared domains and groups of companies.
Merging needs rules written down before the first merge, because the merge decides which value survives when the two records disagree. The usual rules are: the older record's original source wins; the newer record's most recent source wins; the most recent consent value wins; activities and deals are combined; and the owner is the one with the open deal, or else the older owner. A merge is irreversible in most systems, so the queue is reviewed by a named person rather than cleared automatically.
Hygiene beyond duplicates is a routine. A weekly pass over the merge queue. A monthly report of records missing required fields, grouped by the form or import that created them, so the cause is fixed rather than the symptom. A quarterly review of contacts with no activity for a long period and of companies with no contacts. And a standing rule on bounced email addresses and unsubscribes, which are flags to respect rather than records to delete. The data quality guide describes the equivalent routine for event data; the two reviews are best scheduled together.
Permissions and ownership
Every record needs an owner: the person responsible for it, who receives its tasks and notifications and who is accountable for its state. Contacts and companies are usually owned by the salesperson or account manager; deals by the salesperson working them; and new leads by whoever the hand-off workflow assigns. Records with no owner are the ones nobody follows up, and a report of unowned records is one of the fastest ways to find leaks.
Assignment rules decide who gets what. Round-robin spreads new leads evenly; territory rules assign by region or segment; named-account rules send anything from a known company to its account manager. The rule is a workflow, and it lives with the hand-off design in the pipeline guide.
Permissions decide who may change what. Marketing normally needs to create and edit contacts and companies, write source and consent fields, build lists and workflows, and read deals; it rarely needs to edit deal stages or amounts, and giving it that power invites accidental changes to the pipeline. Sales needs the reverse. Integrations should authenticate as their own named users with the minimum scope they need, so that an audit log shows which system changed a field, and a departed contractor's credentials can be revoked without breaking a connection. Export rights deserve the most care, because a contact database is both a commercial asset and, in most jurisdictions, personal data with legal duties attached.
Reporting basics from the CRM
Marketing should be able to pull a short list of reports from the CRM without help, and the list is a good test of whether the fields above are working. Leads created by original source by month. Lifecycle stage conversion by source for a cohort of leads created in a period. Deals won, and revenue won, by original source. Average time from creation to first sales activity, by owner. Records missing required fields, by creating form. And, once spend is joined from the ad platforms, cost per acquisition (CACCost to acquire one new customer.) calculated on won deals rather than on form fills.
The join to spend is the step most teams never make, and it is where the CRM earns its place in marketing. Campaign names in the ad platform and campaign values in the source fields have to match, or a mapping table has to translate between them; the naming conventions guide linked above is the way to avoid needing the table. Where the business already has a reporting layer, the definition of each KPIA metric tied directly to an objective. belongs in it rather than in individual CRM reports, so that a renamed field or a changed stage does not silently change a number; the metrics layer guide explains how.
A training provider finds its missing source
A professional training provider runs a monthly report of contacts created by original source and finds that 2,340 of the 6,100 contacts created in the last quarter have no source at all. Grouping by creating form shows that 1,900 of them came from a single brochure-download form added by the web agency without the hidden parameter fields the other forms carry, and the rest from a conference list imported without a source value. The form is fixed in an afternoon; the import is backfilled with the source "event: spring conference". The next quarter's report shows 3 unexplained contacts out of 5,800, and the paid search campaign that had looked like it produced 40 leads a month turns out to have produced 95, which changes the budget conversation entirely.
The lesson in the example is general: a reporting gap in the CRM is almost always a capture gap at an entry point, and the report that groups missing values by their creating form or import is the one that finds it.
Common mistakes
- Free-text fields where a picklist would do. Industry, source and reason fields typed by hand cannot be reported on and will be typed differently by every user.
- Overwriting original source. A single source field updated on every visit makes every customer look like they came from the last email they clicked.
- Revenue on the contact. Amounts belong on the deal; a second contact on the deal otherwise doubles the revenue.
- Letting salespeople choose the source from a list. The guess becomes the data; capture parameters from the page instead and reserve manual values for genuinely offline sources.
- Merging without rules. Which record wins has to be decided before the first merge, not during it.
- Unowned records. A lead with no owner is a lead nobody is calling; report on them weekly.
- Shared integration credentials. One departed user, one revoked password, every connection down; give each integration its own user.
- Definitions in reports rather than a metrics layer. A renamed stage changes a KPIA metric tied directly to an objective. silently; define it once and reference it.
The CRM for marketers hub places this guide beside its siblings, and the pipeline guide picks up where this one stops, at the stages a well-formed record moves through.
