Data dictionary: the 10 definitions that cause conflict

The revenue operations data dictionary is the document that gives every term used in the board meeting a single name, a business definition, a calculation rule, a source, an owner, the matching CRM field and an update frequency. Without it, “active customer”, “qualified lead”, “revenue” and “churn” mean one thing to each team, and the board gets three numbers for the same question, all of them correct by the rule of whoever calculated them. A useful dictionary starts small, with the ten or fifteen definitions that cause the most conflict, goes through formal approval by the leaders, gets a version and an effective date and is published where the team looks up the data, with training for those who record and for those who read.

The problem has a measure. In a survey of 299 sales leaders conducted from May through June 2020, inconsistent definitions of metrics and KPIs came up as the second most cited reason for poor data literacy, mentioned by 47% of respondents, because they make it hard for different functions to share data and analytic output. The survey’s own recommendation to sales operations is to build a data dictionary with common data definitions (Gartner, 2020).

Why does each team have its own active customer count?

Each team counts customers by the question it needs to answer. In a hypothetical example, the CRM holds 455 accounts with a won contract and no recorded cancellation, and sales presents that number. CS knows that 43 of those accounts have already ended their contract or said they will not renew, without anyone updating the deal, and presents 412 customers with a contract in force. Finance counts only those who paid an invoice in the last 30 days and excludes 24 accounts in arrears, reaching 388. The numbers add up (455 minus 43 is 412, and 412 minus 24 is 388), and each team defends its own with good reason.

The conflict starts with the name. The three teams use “active customer” for different concepts, and the board dashboard mixes one team’s count with another team’s rate. The same happens with “revenue”, which can be the signed amount, the invoiced amount or what accounting recognizes, and with “churn”, which can count accounts or money and record the loss at the notice or at the end of the contract. In the assessments I run, the first question I ask the board is how many active customers the company has today, and two people in the same room tend to answer with different numbers.

What does a data dictionary solve, and what belongs in other documents?

The dictionary records the meaning of each term and field that feeds a decision, and it joins two layers that governance literature usually keeps apart: the business glossary, with the terms people say in the meeting, and the technical dictionary, with the fields and allowed values in the system. In a mid-sized commercial operation, keeping both in the same entry links the word the director uses to the field the salesperson fills in.

The RevOps metrics matrix holds the formulas for CAC, LTV, payback and NRR, and the sheet for each indicator in the sales dashboard says which screen shows the number and who explains the variation. Both use “new customer”, “churn” and “revenue” as ingredients, and the dictionary fixes what each ingredient means. The design of objects and associations lives in the CRM data model; the dictionary describes what each field in that model means for the business.

What does a data dictionary record?

Each term becomes an entry with the same attributes.

Attribute	What to record	Error it prevents
Official name	One single name, with the banned synonyms	Two teams using the same word for different things
Business definition	One or two sentences with what is in and what is out	A definition only the analyst understands
Calculation rule	Criterion or formula, unit, reference date and exceptions	Each report filtering its own way
Source	Master system for the data and systems that receive a copy	A number built in a side spreadsheet
Owner	The leader who answers for the definition and approves changes	A rule changed by whoever configures the report
CRM field	Object, internal name, type and allowed values	An approved term with no address in the system
Frequency	When the data updates and in which meeting it is read	Comparing numbers from different dates
Version and effective date	Version number, start date and what changed	A historical series broken without notice

The first three columns belong to the leaders. The others link the definition to the operation, and the revenue operations or CRM person fills them in with IT, using the map of master systems described in sales stack.

A completed entry template

The example uses the definition of “active customer” I usually suggest as a starting point. The field values are illustrative.

Attribute	Active customer
Official name	Active customer. Banned synonyms: base customer, active logo
Business definition	A company with at least one contract in force on the cutoff date, whether or not it pays on time
Calculation rule	Accounts with a contract started on or before the cutoff date and ending after it; a cancellation notice counts until the end date; arrears is a separate attribute
Source	ERP for contract dates and status, with a copy in the CRM
Owner	CS leader, with finance validating the dates
CRM field	Company, “customer status”: active, cancelling, former customer, never a customer
Frequency	Daily update through the integration; read in the monthly metrics meeting
Version and effective date	1.0, from the first day of the quarter

Which definitions cause the most conflict in the board meeting?

The list gathers the ten terms I most often see stall a results meeting in B2B operations. The suggested definition is there to open the conversation among the leaders, and each company adjusts it to its own contract model.

Term	Where the conflict starts	Suggested definition	Owner
Qualified lead (MQL)	Marketing counts whoever downloaded content; sales counts whoever has a project	A lead that meets the fit and intent criteria approved in the SLA, with the qualification date recorded by the CRM	Marketing, with sales
Opportunity	The rep creates it at first contact; the manager counts it after the meeting	A deal created after a meeting with a stated pain and an identified decision maker	Sales
Pipeline	Whether overdue deals and renewals count	Value of open new customer and expansion deals with a future close date in the period; renewals in a separate pipeline	Sales, with RevOps
Won deal	Verbal yes, approved proposal or signed contract	A contract signed by both parties, with the close date equal to the signature date	Sales
New customer	Reactivation and a new unit of a group already served	A company with no contract in force in the previous 12 months that signs its first contract; reactivation and a new unit of an existing group get their own category	Sales, with finance
Active customer	Contract in force, payment up to date or product usage	A company with a contract in force on the cutoff date; payment and usage are separate attributes	CS, with finance
Revenue	Signed, invoiced or recognized amount	Three terms: contracted revenue (at signature), invoiced revenue (on the invoice, from the ERP) and recognized revenue (by the accounting rule)	Finance
Monthly recurring revenue	Whether implementation, one-off services and temporary discounts count	Monthly value of recurring contracts in force, net of discounts, without implementation and one-off services	Finance, with CS
Churn	Counts customers or money; notice or end of contract	Customer churn: accounts that stopped being active in the period, on the end date. Revenue churn: recurring revenue lost to cancellation; downgrade has its own term	CS
Expansion	Index-based price adjustment, upsell and a new contract from an active customer	Growth in an active customer’s recurring revenue through an additional sale; index-based adjustment has its own term	CS and sales, with finance

Active customer, revenue and churn need a board decision

Leaders settle seven of these terms in one working session. The other three usually need arbitration, because they touch targets and compensation. The definition of active customer decides whether an account in arrears stays in the book that pays renewal commission, and the definition of revenue decides whether the sales target is measured on the signed or the invoiced amount, a gap that can exceed a quarter when implementation comes before the first invoice.

Churn holds the longest dispute. Recorded at the notice, it gives CS time to win the customer back, but it swings with notices that are later withdrawn. Recorded at the end of the contract, it matches finance and arrives late for management. The way out I usually propose splits it into two terms: official churn on the end date, which goes into the revenue metrics, and the “cancelling” status recorded in the CRM from the notice onward, which feeds the CS risk dashboard.

How do you write a definition two people read the same way?

The test I use is to hand the definition and five real CRM records to two people from different teams and ask them to classify each record. If they disagree on any case, the definition still leaves room for interpretation. The version that passes says what is in and what is out, names the reference date and settles the edge case, such as the customer who cancelled one product and kept another.

Avoid the circular definition (“an active customer is one that is active”) and the term that depends on judgment, such as “lead with real interest”. Write the rule the way the CRM can apply it: instead of “engaged customer”, use “customer with at least one user login in the last 30 days”. When the operation needs two versions of a concept, give them two names, because “contracted revenue” and “invoiced revenue” sit well on the same dashboard and two numbers both called “revenue” confuse the board.

How do you approve and maintain the data dictionary?

The dictionary gains authority through formal approval and a single path for changes. Definitions go through the monthly committee described in data governance, with a fixed block for the dictionary. The marketing, sales and CS leaders take part as entry owners, along with finance, the data steward, usually from revenue operations, and the board sponsor, who arbitrates when two teams cannot agree.

Each entry has a version. Version 1.0 takes effect on an announced date, and each change creates a new version with what changed, who approved it and when it takes effect. Older numbers stay calculated under the old rule, and the dashboard marks the break in the series so nobody compares March churn, measured at the notice, with April churn, measured at the end date. Within governance, the dictionary is the piece the commercial team feels first.

Change step	Who does it	What it delivers
Request	Any leader or RevOps	Term, proposed change and reason, in writing
Impact analysis	RevOps, with IT	Reports, targets, commissions and integrations affected
Decision	Committee, with the board arbitrating	Approval or rejection recorded in the minutes, at the monthly meeting
Configuration	CRM administration and IT	Fields, rules and reports adjusted and tested
Communication	Entry owner, with RevOps	Version published, notice to the teams and break marked on the dashboard

Who owns, who executes and who approves?

The split that works separates whoever answers for the meaning from whoever handles the configuration.

Role	Responsibility	Example decision
Board (sponsor)	Approves version 1.0, arbitrates conflicts and rejects numbers outside the dictionary	Sales target on contracted or invoiced revenue
Entry owner	Answers for the definition, proposes changes and explains the number	CS defines when an account becomes “cancelling”
Finance	Validates revenue, cost and contract dates	Implementation stays out of recurring revenue
Data steward (revenue operations)	Keeps the document, brings requests to the committee and analyzes impact	Shows that changing “new customer” alters commissions
CRM administration and IT	Configures fields, rules, integrations and reports	Creates the “cancelling” value and the rule that fills it
Marketing, sales and CS teams	Record according to the definition and flag cases it does not cover	A rep flags a purchase made by another company in the group

The role most often missing is the owner. When the definition of churn belongs to everyone, whoever configures the report decides the rule without noticing, and the board finds out months later, when the series jumps.

Where should you publish the data dictionary?

The official version lives at a single address, such as the internal wiki or the document portal, with version history and read access for the whole operation.

Then take the definition to the point of use. The most widely used CRMs accept a description on each field, shown when the rep hovers over it, and it should repeat the business definition with a link to the official version. Each report on the board dashboard carries the term and the version used in its footer. If the company runs BI or AI agents on top of the CRM, the dictionary goes in as context, because an agent that reads “active customer” without the rule calculates whatever version the data allows and hands leadership a fourth number.

Common mistakes when building a data dictionary

Starting with every field in the CRM

A mid-sized company’s CRM has hundreds of fields, and the project that tries to document all of them ends before reaching the ones that matter. Start with the terms in the board meeting and in the commission rule, and document the fields that feed them.

Letting IT write the definitions alone

IT knows the structure of the system and does not decide what counts as a new customer. When the technical team writes the dictionary alone, the definition describes what the database already does and the disagreement between teams stays undecided. IT fills in source and field, and the business definition comes from the leaders.

Approving a definition with no CRM field

If the churn definition uses the contract end date and that date exists only in the ERP, the approved entry becomes text in a folder. The integration goes into the plan before the effective date, and the dashboard shows the term as incomplete until then.

Changing a definition without a version

The company adjusts the qualified lead criteria in July, and MQL to opportunity conversion rises twenty points with nothing changing in the funnel. Without a version and a marked break, the board celebrates or questions a variation that came from the rule.

Training: who needs to learn what

The dictionary changes behavior when each role knows what the definition asks of it. In the same survey cited at the start, sales reps came out with the lowest data proficiency in the organization, 43%, below marketing and customer service (49% each) and finance (53%) (Gartner, 2020).

Leaders learn to read and defend the other teams’ definitions, because the marketing director who understands why CS counts a cancelling customer as active stops challenging the number in the meeting. Reps and SDRs learn opportunity, won deal and new customer, with the effect of each field on commission, and CS learns to record cancellation notices and expansion the way the entry asks. Whoever administers the CRM learns the change flow, so no rule changes on an informal request. In the teams I work with, resistance to the dictionary usually comes from the leader who loses the most favorable version of their own number, which is why leaders go through training before their teams.

30-day roadmap for leadership

1. Days 1 to 3: run the three-spreadsheet test. Ask each team for active customers and last month’s revenue, with the rule used, and take the gaps to the board.
2. Days 4 and 5: name the sponsor and the steward. The board appoints who arbitrates conflicts, and the data steward, from revenue operations, takes over the document and the committee agenda.
3. Days 6 to 10: choose the terms for version 1.0. Start from the ten definitions, add the terms in the commission rule and stop at fifteen entries.
4. Days 11 to 17: write the entries with the owners. Each leader drafts the definitions for their area in the template, and finance reviews everything involving revenue and dates.
5. Days 18 to 20: run the two-reader test. Apply each definition to five real records and rewrite the ones that caused disagreement.
6. Days 21 to 23: approve in committee. Leave active customer, revenue and churn to the board and record the decisions in the minutes.
7. Days 24 to 27: connect to the CRM and the dashboard. Fill in source and field, write the description of each CRM field and mark on the dashboard the terms with incomplete data.
8. Days 28 to 30: publish and teach. Publish version 1.0, show each team the terms that touch its records and schedule the first quarterly review.

The dictionary completes the SLA between marketing, sales and customer success: the SLA sets the commitment between the teams, and the dictionary makes sure they all count that commitment the same way. If your leaders reach the board with different numbers for the same question, that is the work of the in-company Revenue Operations training. I run it with the marketing, sales and CS leaders, finance and the operations person in the same room, on your operation’s data, and the group leaves with the SLA between marketing, sales and CS and the RevOps metrics matrix built on the same vocabulary.

Frequently asked questions

What is a data dictionary?

It is the document that records, for each term and field that feeds a decision, the official name, business definition, calculation rule, source, owner, system field and update frequency. In revenue operations, it ensures that marketing, sales, CS and finance calculate active customers, revenue and churn the same way.

What is the difference between a data dictionary and a business glossary?

The business glossary defines the terms people use in conversation, and the data dictionary describes the system fields, with type and allowed values. In a mid-sized commercial operation, it pays to keep both layers in the same entry, linking the word in the meeting to the field in the CRM.

Who should own the data dictionary?

The revenue operations or CRM person maintains the document and runs the changes, and each entry is owned by the leader of the team that answers for the definition. The board sponsors it, approves the first version and arbitrates conflicts that touch targets and compensation.

How many definitions should the first version have?

Between ten and fifteen, chosen from the terms that appear in the board meeting and in the commission rule. A first version with hundreds of entries takes months and arrives out of date, and the remaining terms come in at the quarterly reviews.

How often should the dictionary be reviewed?

One-off changes follow the request, impact analysis and decision flow at the monthly meeting. A full review each quarter checks whether reports calculate what the entries say and adds new terms from products or contracts.

How do you prepare the team to use the data dictionary?

Train by role, showing each person the terms that touch their records and the effect of each field on the board’s number and on commission. Leaders go through the training first, because they need to read and defend the other teams’ definitions, and the description on each CRM field takes the definition to the moment of recording.