Before you can close a synthesis gap, you need to know exactly where it is. You can spot your data fragmentation problem in 20 minutes if you know what to look for — the signals are already there, scattered across tools that don't share a view. This audit maps exactly where yours are.

Why does a data stack audit matter?

A 2025 study by Precisely and Drexel University's LeBow College of Business — surveying more than 550 data and analytics professionals worldwide — found that 67% of organizations don't fully trust their data for decision-making, up from 55% the year before. Most growth-stage companies are aware something is off. The harder question is which specific decisions the fragmentation is distorting right now.

Growth-stage companies average 106 SaaS apps by the time they're three to six years old, and each one captures signals in its own domain without seeing what the others see. Knowledge workers lose an estimated 30% of their workday searching for data that exists somewhere in their stack — just not in one place, and not in a form that answers cross-tool questions automatically.

This is what the Optimizing Posture produces at scale. The Listening Posture had automatic synthesis — a small team seeing everything connected signals by default. The Optimizing Posture loses it, because specialized functions each see their own domain and nobody has the job of connecting them. The audit is how you see exactly what was lost.

What does the Four-Check Fragmentation Audit look like?

Four checks. Twenty minutes. Each one targets a distinct failure mode of fragmented data:

• Check 1 maps the breadth of the synthesis gap — how many routine decisions require manual data assembly across tools.

• Check 2 targets the LTV blind spot — whether your acquisition spend is being optimized against retention data.

• Check 3 measures churn signal lag — how early your data showed what your team didn't see until after the fact.

• Check 4 surfaces the decision loop — whether two functions made overlapping calls about the same customers without shared context.

You don't need to pull new data for any of these. You're testing whether your stack can answer questions it should already be able to answer.

Check 1 — How many of your last decisions required more than one tool? (5 minutes)

List three business decisions your team made in the last 90 days. For each one, count how many separate tools you opened before making the call.

• An acquisition strategy decision typically needs ad platform data plus product activation rates.

• A feature prioritization call typically needs usage analytics plus customer support data.

• An expansion bet typically needs CRM data plus behavioral product signals.

Each time you opened a second tool to complete a picture the first tool started, you patched a synthesis gap manually. That gap didn't go away — you just crossed it on foot instead of having a bridge.

Score: 0 of 3 decisions required multiple tools = synthesis gap is narrow. 1 of 3 = the gap exists but affects specific decisions only. 2 or 3 of 3 = the synthesis gap is costing you on routine decisions.

Check 2 — Can you answer the LTV blind spot question? (3 minutes)

Open your ad attribution tool. Identify your top three acquisition channels. Now open your product analytics tool. Try to answer this: which of those three channels produces your highest-retention customers at 90 days?

If you need to export and manually join data to answer it, your acquisition spend is being optimized without retention data in the loop. You concentrated budget on a channel because the CPAs looked strong, without seeing the 90-day retention story until a cohort had already compounded.

That's the Invisible Cohort failure — one of the Two Failures of Fragmented Data. The data existed. It just lived across two tools nobody was connecting.

Score: Can answer from a single integrated view = no active synthesis gap here. Requires a manual join = active synthesis gap at the most consequential point in your funnel.

Check 3 — How early did you see your last churn? (7 minutes)

Find your last three churned accounts. Look backwards through your stack: were there behavioral signals — login frequency dropping, feature engagement declining, support ticket volume rising — that appeared before the churn event?

Those signals existed almost certainly did. The question is when they surfaced, and whether anyone received them before the churn happened.

If the signals were in your product analytics tool but nobody flagged them proactively, your stack runs on reactive intelligence only. The Two Modes of Intelligence come down to this distinction: reactive shows you the churn number after it happens. Proactive surfaces the behavioral pattern while you still have time to act. If your last three churns were visible in the data before they became numbers, and nobody caught them, the synthesis layer isn't watching.

Score: Signals appeared and were flagged proactively = proactive intelligence working. Signals appeared but not flagged = reactive only. Signals weren't detectable = your stack isn't monitoring the right behaviors.

Check 4 — Has your team made the same decision twice? (5 minutes)

Look at decisions made by two different functions — growth and product, or product and customer success — about the same customer segment in the last six months.

Ask whether either team made a call without knowing what the other team had already decided. Did growth concentrate budget on a segment that product had already identified as low-retention? Did product deprioritize a feature that customer success had flagged as a top churn driver?

Even one instance is the same-decisions-twice pattern. Two teams, two accurate views of the same time period, two calls made without shared context. The synthesis layer makes that loop impossible — it's the shared view that neither function had on its own.

Score: No overlapping decisions without shared context = synthesis working. One or more overlapping decisions = active synthesis gap confirmed.

What does your score mean?

Each check targets a different fracture point. Check 1 shows breadth — how many routine decisions cross the synthesis gap. Check 2 targets the Invisible Cohort failure mode. Check 3 targets the reactive-to-proactive divide. Check 4 targets the decision loop.

Score one check actively and you have a specific, bounded gap. Connecting two identified data streams addresses it directly.

Score two or more and the synthesis gap is structural. That's the point where Layer 3 — the synthesis layer — becomes unavoidable. Adding another tool adds to Layer 1 or Layer 2. Neither closes the synthesis gap. The Four Signs Your Company Needs an Intelligence Layer are the diagnostic. This audit is how you see which signs are live in your stack right now.

The next post makes the case for why more tools don't solve this — and what building Layer 3 actually requires.

— Steven

FAQ

How do I audit my data stack?

• Run the Four-Check Fragmentation Audit. Count how many recent decisions required opening more than one tool — any decision requiring a manual join is evidence of a synthesis gap. Test whether you can answer the LTV blind spot question from a single view: which acquisition channel produces your highest-retention customers at 90 days? Look backwards through recent churned accounts and ask when behavioral signals appeared relative to the churn event, and whether anyone flagged them proactively. Check whether any two functions made overlapping decisions about the same customer segment without shared context. Two or more active scores mean the gap is structural in your stack.

How do I know if my company has a data fragmentation problem?

• The clearest signal is the LTV blind spot: if you can't tell which acquisition channel produces your highest-retention customers without manually joining your ad attribution and product data, your stack isn't connected at the most consequential point in your funnel. A second clear signal is insight arriving after decisions close — the churn number moves, the team investigates, and finds behavioral signals that were in the data two weeks earlier. If two or more checks from the Four-Check Fragmentation Audit score actively, the synthesis gap is structural in your stack.

What signals am I missing from my growth stack?

• The most commonly missed signals are cross-tool correlations — patterns that require comparing data from two sources your stack never connects simultaneously. The three most common blind spots: acquisition channel to 90-day retention linkage, which requires joining ad attribution and product data; early churn behavioral signals, which require connecting product usage and support data; and feature value versus retention correlation, which requires connecting product usage and CRM expansion data. Those signals exist in Layer 1. What's missing is the Layer 3 synthesis function that watches them simultaneously and surfaces what they mean together.