Single Source of Truth

Definition
A single source of truth is one authoritative place a fact lives, so every report and automation reads it there rather than a copy.

Why it matters

The moment a fact lives in two places, the two drift apart and nobody can say which is right. Forecasts built on a rep's private spreadsheet become guesses. Meetings turn into arguments about whose number is correct. The problem grows with automation. An AI agent or workflow reads whatever it is pointed at, so a stale copy leads to confident wrong action at speed.

How to apply it

  • List the facts the business depends on and name one owning system for each.
  • Point dashboards and automations at that system directly, never at an export.
  • Treat a second copy of a fact as a bug to fix, not a convenience to tolerate.
  • Give agents read access to the live source.
  • Every quarter, look for spreadsheets and trackers quietly duplicating a record.
  • Write down definitions as well as numbers. "Active customer" should mean one thing everywhere.

What it is

For any fact a business relies on, such as the value of a deal, the price of a plan or the status of an invoice, one system is named as the owner. Everything else reads from it or is rebuilt from it. The CRM owns deals. The accounting system owns revenue. Copies are allowed only when they refresh automatically from the owner, like a dashboard built on live data.

The opposite is a shadow copy: a spreadsheet exported last month, a figure typed into a slide, a private tracker kept because the official tool feels slow.

Common mistakes

  • Naming no owner. If two systems can both be edited, both will be, and they drift.
  • Exporting instead of connecting. A CSV is a copy from the moment it is saved.
  • Tolerating the "convenient" private tracker. Each one is a second source, and a future argument.
  • Fixing the numbers but not the definitions. If "active customer" means different things in different tools, a single source still gives conflicting answers.
  • Pointing automations at copies. An AI agent or workflow acts on whatever it reads, so a stale copy leads to confident wrong action.
  • Treating it as one-off. Without a regular check, copies grow back.
Worked example

Say a consultancy of eight people runs its deals in Pipedrive, yet two reps keep private spreadsheets of pipeline value, and the monthly forecast is built from one of those copies. It is 30 per cent too high because one sheet still counts a deal that was lost in March. The finance lead names Pipedrive as the owner of deal value and asks everyone else to stop keeping a copy.

The fix takes a day. The private sheets are deleted, and a Databox dashboard reads pipeline figures straight from the live data, not from an export. The definition of an active deal is written down once, next to the dashboard. The next forecast uses the figures everyone sees on screen, so the Monday meeting is spent on what to do rather than whose number is right.

Tools in the example

Some links are affiliate links: we may earn a commission at no cost to you. It never decides a ranking. How we work with partners

  1. Article

    Single customer view

    One record per customer, built on the same idea.

  2. Article

    Data warehouse

    Often where many sources are joined.

  3. Article

    Data modelling

    Fixes each definition once.

  4. Article

    Data Hygiene

    The upkeep that keeps the source clean.

Where it shows up

  • Managing your sales pipeline and revenue operations so deals don't fall through cracks. RevOps connects the two so forecasts hold up.
    27 chapters