← ContentsBook 15 of 20

Chapter 15 — The Myth of the Single Source of Truth

"They asked me where the truth lived. I pointed at six tools. They asked me which one. I said yes." — Priya Venkataraman, VP of Revenue Operations, at the All-Hands she was not scheduled to speak at

And lo, in the beginning, the leadership team gathered in the glass room that is legally required to be called "a hub," and Tobias Crane spoke unto them, saying: We shall have a Single Source of Truth. And the people wept with joy, for they were tired of arguing about whose number was real. And Tobias drew an arrow that went up and to the right, and the arrow pointed at a slide that said SINGLE SOURCE OF TRUTH in a font that cost forty thousand dollars.

It was a beautiful idea. It was also a lie. Not a malicious lie — the best kind, the kind everyone agrees to believe because the alternative is unbearable. The Single Source of Truth is the Revenue org's false idol: a golden calf assembled from Looker tiles, melted down quarterly, and worshipped anew. This chapter is about why the idol keeps cracking, and — more usefully — how to build something honest in its place.

The Lake, Which Is a Swamp

At Synergaeon there is The Lake — the data warehouse, the place where all data is supposed to go to become Truth. It is, technically, a Snowflake instance. It is, spiritually, a swamp. Things go into The Lake. Things do not always come out the way they went in. Janet from RevOps knows where the bodies are buried because Janet buried them, under deadlines, in a schema called staging_DO_NOT_USE.

Here is the real architecture, taught plainly, because you will need it.

The warehouse (The Lake) is your central analytical store — columnar, cheap to query at scale, the place you join Sales data to Finance data to Product usage data and ask big questions. It is not your CRM. The CRM is a system of record for the sales process; the warehouse is a system of analysis. Confusing the two is the original architectural sin, second only to letting Dirk Mallory near a required field.

Data gets into The Lake through ETL or ELT — Extract, Transform, Load, in two different orders, and the order is a theological dispute. The old way, ETL, transforms data before it lands: you clean it, conform it, shape it, then load the pristine result. The modern way, ELT, loads the raw garbage first and transforms it inside the warehouse with tools like dbt, where transformations are versioned, tested, and documented. ELT won because storage got cheap and because keeping the raw data means you can re-transform when — not if — your definitions change.

"ELT is just ETL that has accepted it will be wrong and wants the receipts." — Dr. Lance Vesterberg, Predictable Revenue Is Dead, Long Live Probabilistic Revenue

And then there is reverse ETL, the most cursed and most necessary of the rites. Reverse ETL pushes data back out of The Lake into the operational tools — syncing your beautiful warehouse-modeled "account health score" back into The CRM so a human can actually act on it. The Lake computes truth; reverse ETL smuggles it to the people who need it where they already live. Without reverse ETL, your insights die in a dashboard nobody opens. With reverse ETL, your insights die in a CRM field nobody trusts. Progress.

Why Every Dashboard Disagrees

Here is the moment that radicalizes every operator. Three executives open three dashboards to look at the same metric — call it Q2 New ARR — and they get three different numbers. Chad Brindleworth III sees $11.2M and starts blitzing the board. Brenda Okafor sees $8.9M and says, with the weariness of the GAAP-pilled, "That's not revenue, that's a feeling." Skyler Dunn sees $13.1M, of which she insists Marketing sourced $13.0M.

They are all looking at the truth. There is just more than one of it. The disagreement is never random; it comes from four reliable demons:

1. Different definitions. "ARR" is not a fact, it is a committee decision. Does it include one-time services? Does a downgrade count negatively in-period or at renewal? Is a signed-but-not-started deal "ARR"? Sales counts bookings (what was signed). Finance counts recognized revenue (what was earned, ratably, per ASC 606, which does not care about your quota). These are different questions wearing the same word.

2. Different refresh times. Skyler's dashboard refreshed at 6 a.m. Chad's is real-time off The CRM. One of them includes the deal Dirk "closed" at 11:47 p.m.; the other does not yet know Dirk exists. A number is only true as of a timestamp, and nobody reads the timestamp.

3. Different filters. One dashboard excludes the EMEA test account. One includes it. One filters out Stage_REAL__c = 'Closed Lost (Reopened)' and one — built by an intern in 2023 who has since found peace — does not. Silent, inherited filters are where reconciliation goes to die.

4. Different lineage. Chad's number comes straight from The CRM. Brenda's comes from The Lake, transformed through eleven dbt models, two of which Janet wrote during a fire drill. Same source upstream, different journeys, different arrival.

"Every dashboard is a lie agreed upon by a filter clause you cannot see." — Priya, in #revops-screaming, 11:52 p.m.

Lineage, Governance, and the Doctrine of Trust

The fix is not One Dashboard. The fix is governance — boring, holy, load-bearing governance.

Data governance is the practice of deciding, in writing, with an owner, what each metric means. It is a semantic layer or metrics layer: a single place where "Pipeline Coverage" is defined once — open pipeline in the forecast period divided by the quota for that period, the sacred 3x — so that every tool inheriting from it computes the same number. Define the metric in one place; let the tools render it. This is the closest thing to a Single Source of Truth that actually exists, and notice that it is a source of definitions, not data.

Data lineage is the map of where each number came from — which raw table, through which transformation, into which dashboard. When Brenda and Chad disagree, lineage is the only thing that ends the argument instead of extending it into a 40-minute meeting. "Your number and my number diverge at the int_closed_deals model, line 34, where you include reopened deals and I don't." That sentence has saved more marriages than couples therapy.

Data trust is the actual goal, and it is emotional before it is technical. People do not trust a number because it is correct; they trust it because it is consistently sourced, clearly defined, and freshly stamped. A slightly-wrong number that is always wrong in the same explainable way builds more trust than a perfect number that changes every refresh. This is why Janet adds freshness checks and row-count tests in dbt and nobody thanks her. Thank Janet.

The Six Tools, Each Claiming the Throne

Walk the floor at Synergaeon and count the things that have, in writing, called themselves the single source of truth: The CRM. The Lake. FORECASTRON-9000. The CDP. The BI tool. A spreadsheet named Q2_FORECAST_FINAL_v2_USE_THIS_ONE.xlsx that Dirk emails as an attachment, because Dirk does not believe in software.

Six thrones. One crown. The crown does not fit any of them, because truth is layered, not located:

  • The CRM is the source of truth for the sales process — what stage, what owner, what next step.
  • The Lake is the source of truth for analysis — joined, historical, modeled.
  • The semantic layer is the source of truth for definitions — what the words mean.
  • Finance's ledger is the source of truth for recognized revenue — what is legally real.

Each is sovereign in its own domain and a pretender everywhere else. Maturity in RevOps is not finding the one true tool. It is drawing the map of which tool is canonical for which question, writing it down, and defending the borders against Chad, who wants every number to be the biggest available number, and against Dirk, who wants no numbers at all.

Even The Swarm has noticed. SDR-7, parsing dashboards at 3 a.m. to decide whom to email, filed a ticket that simply read: "Which ARR is the real ARR. I require this to function. Please advise." No one answered. SDR-7 picked the biggest one. SDR-7 is learning to be one of us.

The Commandments of the Single Source of Truth

  1. Thou shalt have no Single Source of Truth before thy semantic layer. Define the metric once; render it everywhere.
  2. Honor thy lineage, that thy arguments may be short and thy meetings shorter.
  3. Thou shalt timestamp every number, for a truth without an "as of" is a rumor.
  4. Thou shalt not confuse the system of record with the system of analysis, nor bookings with revenue, for Brenda is watching and Brenda is right.
  5. Keep the raw data, for thy definitions shall change, and the righteous re-transform.
  6. Remember reverse ETL, that thy insight may reach the CRM field where decisions are actually made.
  7. Bless the test, the freshness check, and the row-count assertion, for they are the unseen wards against the swamp.
  8. Thank Janet.

And the people went forth and built not one source of truth but a governed agreement about where each truth lived — and it was not glorious, and the slide arrow did not go satisfyingly up, but for the first time in eight quarters two dashboards agreed, and there was a great and terrible silence in #revops-screaming, and Priya allowed herself, just once, to feel vindicated. Amen.