Skip to content
Sales Label Consulting Book a call
Article

Sales Data Model: How to Design One Your Team Can Trust

A sales data model defines the objects, fields and relationships your CRM runs on. Learn what to include, how to design it, and how it feeds reporting.

Sales Label Consulting17 Nov 2026 · 15 min read

A sales data model is the agreed blueprint of what your CRM tracks (accounts, contacts, leads, opportunities, products, activities), which fields each record needs, and how the records connect. Get it right and your forecast, routing and reporting read from one version of the truth. Get it wrong and every dashboard becomes an argument. Below is what the model contains, how to design it around your sales process, and where it breaks as you scale.

Real talk: most Heads of Sales inherit a data model rather than design one. It grew one custom field at a time, and nobody can say why “Stage 3” exists. This guide is for the leader who wants to fix that deliberately.

Saleslabelconsulting
Build a Sales Model Your Team Trusts
Sales Label Consulting supports sales teams with sales audits, sales enablement, and demand generation for more predictable results.

Visit

TL;DR

  • Sales data is the raw information; the data model is the structure that holds it. The model defines objects, fields and relationships. The data fills them.
  • Start from your revenue model and buyer journey, not from the CRM’s defaults. List the “nouns” of your business first, then decide what is standard and what is custom.
  • Keep the core small: account, contact, lead, opportunity, product, activity. Add entities only when a defined workflow demands them.
  • Your CRM model and your analytics model are different shapes. Operational systems favor normalized tables; reporting favors facts and dimensions.
  • Model problems show up as pipeline problems: duplicate accounts, misrouted leads, stages nobody trusts, forecasts that move every week.

Table of Contents

What a sales data model is (and what it is not)

Salesforce describes sales data as the raw information collected about sales activities and customers. It splits into external data (prospect demographics, interests, behavior and engagement as they move through the funnel) and internal data (deal attributes such as product type and pricing, plus rep performance metrics). Sales analytics is the next layer: analyzing that data for patterns and predictions.

The data model sits underneath all of it. It decides where each piece of information lives and how it connects to everything else. In Salesforce’s Trailhead training, a data model determines how the information you track relates to your users and your business, and it defines the relationships between different sets of data. Two building blocks do the work:

  • Objects are the home for a type of data, like a form. Account, Contact and Opportunity are objects.
  • Fields are the specific facts recorded on that form: deal amount, close date, industry.

A third word you will hear is record: one filled-in form, such as one specific company or one specific deal. And the glue is relationships: the rules that say a contact belongs to an account, or that an opportunity carries products.

A simple way to hold the difference: sales data answers “what happened?” The data model answers “where does it go, what does it connect to, and who owns the truth?”

The core entities every B2B tech sales model needs

You do not need fifty objects. You need the six or seven that your sales motion actually touches, defined cleanly.

Entity What it represents Typical relationship
Lead An unqualified person or company showing interest Converts into an account, contact and opportunity
Account A company you pursue or serve Parent of contacts and opportunities
Contact A person at an account Many contacts to one account
Opportunity A potential deal Belongs to an account; links to contacts and products
Product What you sell Attached to opportunities through line items
Activity A call, email, meeting or task Linked to a contact, account or opportunity
User / Team The reps and managers who own records Owns accounts, leads and opportunities

This is not a vendor quirk. Pega’s sales automation documentation lists a similar standard set of business objects (account, activity, contact, lead, opportunity, quote and others), each with its own attributes that interact as records move through the sales life cycle. Salesforce’s Sales Cloud data model gallery goes further and includes objects for contracts, orders, price book entries, forecast items, territories and campaigns.

The lesson is the same in every system: the core is small, and the extensions (contracts, orders, partners, territories) appear when a real workflow needs them.

How the relationships actually behave

The relationship type matters more than most leaders expect, because it decides what happens when someone deletes or merges a record. In Salesforce’s own reference diagram, Account to Contact and Account to Opportunity are lookups, while Opportunity to Opportunity Line Item is master-detail with cascade delete. In plain words: a lookup is a loose link, and a master-detail link means the child record cannot live without its parent. Delete the opportunity and its line items go with it. The same reference shows a lead converting into an account plus a contact.

CRM relationships and lead conversion

This is why a messy account structure is expensive. If duplicate accounts exist, contacts and opportunities hang off the wrong parents, and no report can tell you the real size of your pipeline in a named account. If that sounds familiar, our guide to lead-to-account matching covers how misrouted pipeline starts.

Design the model from your sales process, not from the tool

The most common mistake is opening the CRM and starting with fields. Start with how you sell.

HubSpot’s guidance on planning a data model is a good template, whatever CRM you run. It advises understanding your revenue model and ideal customer journey first, identifying the “nouns” of your business, listing every type of record you need before deciding standard versus custom, and then mapping how those entities relate (a contract signed as part of a deal, for example).

Here is the version we would run with a Head of Sales and a CEO in a room.

Step 1: Write down how a deal really happens

Walk through your last ten closed-won deals and your last ten losses. Who showed up? Who signed? Was there a partner, a reseller, a pilot, a procurement team? Every actor and artifact that matters to the outcome is a candidate entity or field.

For an EU-focused B2B tech company, this often reveals things the default CRM ignores: multiple legal entities under one customer group, a separate buyer and user, a paid pilot before the contract, or country-level subsidiaries that buy independently.

Step 2: List the nouns, then pick standard or custom

For each noun ask one question: does an existing standard object do this job? Use standard wherever it does. Custom objects carry a permanent cost in reporting, integrations and training, so reserve them for concepts your process cannot work without.

Step 3: Define the relationships in writing

For every pair of entities, answer: one-to-many or many-to-many? Who is the parent? What happens on delete or merge? Write it as one page. If you cannot draw it on one page, it is too complicated to govern.

A typical set of answers: many contacts belong to one account. One opportunity belongs to one account but involves several contacts. One opportunity holds several products. Many activities point to one contact or opportunity.

Step 4: Decide which field is authoritative

This is where models quietly fail. If “industry” lives on the lead, the account and in your enrichment tool, which one wins? Name a single source of truth for each important field and say which system and which role may change it.

For every field, record four things:

  1. Definition: what it means, in one sentence.
  2. Owner: the role accountable for it.
  3. Required or optional: and at which stage it becomes required.
  4. Source: typed by a rep, synced from another system, or calculated.

Step 5: Build stages and statuses into the model

Opportunity stages are part of the data model, not decoration. Each stage needs entry and exit criteria that map to fields, or reps will move deals forward on feel. EntityModel’s example CRM schema treats the opportunity stage as a reference table, with the pipeline stages in order and a default win probability, which is the right instinct: stages are controlled values, not free text. Our opportunity hygiene rules show how to turn those criteria into a pipeline you can forecast.

Step 6: Test it against the questions you must answer

Before you build anything, list the ten questions the board, the CEO and you ask every month. Pipeline by segment. Conversion by source. Win rate by rep tenure. Sales cycle by deal size. For each, check that the model holds every field needed, in a place where it will be filled in consistently. Our piece on RevOps metrics that predict revenue lists the formula, source, cadence and owner for each one, which is exactly the check this step needs.

Fields that earn their place

Every field you add is a tax on your reps’ time. A field earns its place only if someone makes a decision from it. Use this filter:

  • Decision test: which decision changes based on this field? If none, drop it.
  • Capture test: who fills it in, when, and is it realistic? A field that needs ten minutes of rep research will be empty.
  • Automation test: can it be calculated or synced instead of typed?
  • Consistency test: is it a picklist with controlled values, or free text that produces forty spellings of the same industry?

The result is a model with three kinds of fields: required fields that gate a stage, optional fields reps may add, and calculated fields (ages, totals, scores) the system maintains. Inconsistent values are the main reason sales leaders distrust their own reports, and cleaning them up is a discipline of its own. We cover it in our CRM data normalization guide.

Your CRM model is not your reporting model

Here is a distinction that saves many leaders from a painful rebuild. The model your CRM uses to run day-to-day sales is built for a different job than the model you use to analyze performance.

According to the Kimball-style excerpt published by Wiley, operational systems such as a sales system that captures orders are designed in third normal form (a layout that stores each fact once) to support fast inserts, updates and deletes. Analytic systems do the opposite: they run read-only queries across large volumes of history, and their dimensional models organize data as facts (measurements) and dimensions (context) in a star schema.

What does that mean in practice?

  • Operational (CRM) model: optimized to create a deal, change its stage, reassign an owner. Many linked tables, each fact stored once.
  • Analytical (warehouse or BI) model: optimized to answer questions. Coginiti’s guide to dimensional modeling explains that a fact table holds one row per measurement event, while dimension tables carry the who, what, where and when. Its example: total sales amount by product category for stores in one state during a holiday season. Sales amount is the fact; everything else is context.

The GoodData University course on data model examples makes the same point from the other side: a source model that handles creating a user or changing an opportunity status perfectly is the wrong shape for analytics, which needs a transition to a star schema.

For a Head of Sales, the takeaway is practical. Do not contort your CRM to make one particular report easy. Keep the CRM model clean for the people entering data. When you need heavy reporting (cohort conversion, multi-year win rates, attribution), build that on a separate analytical layer fed from the CRM. The only thing connecting them is clean keys: consistent account IDs, opportunity IDs and stage definitions.

Where sales data models break as you grow

The model that works at 5 reps rarely works at 30. Watch for these signals:

  1. Two teams, two definitions. Marketing’s “qualified lead” and sales’ “qualified lead” are different fields. Our sales and marketing SLA and MQL to SQL benchmarks pieces show how to align them.
  2. Account sprawl. The same customer exists as three accounts, one per country or per rep. Territory and coverage rules then assign work on bad data. See account segmentation models and our sales coverage model guide.
  3. Custom field creep. Every new manager adds a field. Nobody removes one. Reports multiply and contradict each other.
  4. Stages without exit criteria. Deals sit in a stage for reasons nobody recorded.
  5. Tool sprawl. Enrichment, engagement and billing tools each write to the same fields. A sales tech stack audit reveals which system should own what.

Real talk: structure beats heroics. A rep who manually fixes data every Friday is a symptom of a model that was never designed, not a hero.

A one-page model review you can run this month

If you already have a model, do not rebuild it. Review it.

  • Count your objects and fields. How many are used in a report or a workflow in the last quarter? Retire the rest.
  • Check the relationships. Pick 20 accounts at random. Are contacts and opportunities attached to the right parent?
  • Check required fields by stage. What percentage of opportunities at each stage are missing a required field?
  • Check definitions. Ask three people what a “qualified opportunity” is. Count the different answers.
  • Check ownership. For your five most important fields, can you name the owner and the authoritative source?

If the answers are uncomfortable, that is useful. It is the same logic behind a RevOps audit that gets fixed: find the specific break, assign an owner, and set a date.

Our perspective

We rarely meet a sales team with a data problem alone. We meet a sales team with a system problem that shows up as a data problem.

A data model is not an IT artifact. It is the written form of how you sell: who your ICP is, which roles touch a deal, what moves a deal forward, and who is accountable for each number. When the written form and the real process disagree, reps invent workarounds, and the CRM fills with noise.

That is why we work end to end instead of fixing isolated fields. The model, the sales roles, the pipeline stages and the performance governance have to agree with each other. Fix one in isolation and the others pull it back out of shape.

Our approach comes from hands-on practitioner experience in tech, not theory. We start from your real deals, design the revenue system around them, and make the data model the quiet backbone that lets you forecast without drama.

Antony

Where we can help

If your data model has grown by accident, the fastest route is to look at the whole revenue system, not just the CRM. Our Revenue System Diagnostics and Revenue System Design services are built for B2B tech companies in the EU that want their ICP, sales roles, pipeline and performance governance to run on one consistent structure. If you are a Head of Sales or CEO who wants a model your team trusts, start there.

Saleslabelconsulting

For a wider view of how the model fits into the operating rhythm of the team, read our guide to the sales operating model.

FAQ

What is the difference between sales data and a sales data model?

Sales data is the raw information you collect: prospect behavior, deal attributes, rep activity. The sales data model is the structure that decides where that information is stored, what it links to and which fields are required. The data changes daily; the model should change only on purpose.

Which objects should a B2B sales data model include?

Start with lead, account, contact, opportunity, product and activity, plus users and teams for ownership. Standard CRM models, such as Salesforce’s Sales Cloud model, add contracts, orders, quotes and territories. Add these only when a defined workflow in your company needs them.

Should I use standard or custom objects?

Use standard objects wherever they fit your process, because reporting, integrations and training already assume them. Create a custom object only for a concept your sales process cannot work without, such as a paid pilot or a partner referral. HubSpot’s planning guidance suggests listing every record type you need before deciding which are standard and which custom.

Is a CRM data model the same as a data warehouse model?

No. A CRM model is built for fast, accurate day-to-day entry and updates. A warehouse model is built for analysis, with facts and dimensions arranged in a star schema. Keep them separate and connect them with consistent IDs.

How often should we revisit the sales data model?

Review it whenever your sales motion changes: a new segment, a new product line, a new region, or a jump in headcount. A light quarterly check of unused fields, duplicate accounts and stage criteria is usually enough to keep it healthy.

Sources