TapCub is live — analytics, insights and live chat, free on one platform
Glossary

What are data silos? Causes, costs and how to connect them

A data silo is data one team or tool holds that others cannot reach, so one person gets counted as several. Causes, warning signs, a worked example and three steps to connect them.

Su He Reviewed by Ethan Brooks Also called data silos Last updated Nov 19, 2026 Glossary

Quick answer

A data silo is data held by one team, tool or system that the rest of the organization cannot easily reach. The usual result is that one person is counted as several, because no shared key links their records in web, app, support and billing systems.

Formula: Distinct people = people in A + people in B − people in both

What is a data silo?

A data silo is a collection of data that one team, tool or system holds and the rest of the organization cannot easily reach. IBM describes data silos as isolated pockets of data that prevent sharing between departments, software systems and business units. Web analytics holds one list of visitors, the app tool another, the helpdesk a third, billing a fourth. Each is accurate on its own terms. None of them agrees with the others.

The practical symptom is this: the same person is counted as several people. A customer who browses on a laptop, opens the app on a phone, chats with support and pays by card shows up as four records in four systems, with no shared key to say they are one. Every dashboard looks right, the totals never reconcile, and nobody can answer "what has this customer actually done?" The rest of this page shows how to spot that, measure it and start fixing it.

A data silo is not the same as a data warehouse that is hard to use, and it is not the same as the fix. Merging everything into one warehouse is one way to remove silos. The related ideas are defined separately: single customer view is the result you want for one person, and single source of truth is the discipline of agreeing which system wins for each fact.

What causes data silos?

Silos are rarely designed. They accumulate. IBM lists organizational structure, IT complexity, culture, limited resources, regulatory requirements and business growth. In a product-and-marketing setting the causes look like this:

CauseHow it creates a siloTypical sign
Each team buys its own toolMarketing, product, support and finance each pick the tool that fits their jobFour tools, four definitions of "user"
No shared identifierSystems store email, device ID, cookie ID or customer number, never the same oneJoining two exports needs fuzzy matching
Different platformsWeb, iOS, Android and mini programs sent to different toolsMobile and web "users" can never be combined
Mergers and growthAcquired companies bring their own databasesTwo customer tables for the same market
Regional or privacy rulesData kept in separate regions or systems on purposeLegitimate separation that still needs a plan to report across it
Spreadsheets as the glueExports passed around by emailA "final_v7" file that only one person understands

The fourth row matters because not every separation is a mistake. Keeping some data apart can be deliberate. The problem is separation nobody chose, with nobody owning the join.

How do data silos show up and what do they cost?

IBM gives a short list of warning signs. Translated into questions you can ask this week:

  • Do two reports give different numbers for the same metric? (conflicting figures)
  • Does answering one question mean exporting from three tools and merging by hand? (slow, manual gathering)
  • Do teams quote different customer counts in the same meeting? (inconsistent definitions)
  • Can support see what the customer did before writing in? (fragmented customer view)

The costs follow directly. Analysts spend time reconciling instead of analyzing. Decisions rest on partial or conflicting views. Marketing sends a "welcome" message to someone who has been a customer for two years. And when a person exercises their right to a copy of their data, which GDPR Article 15 provides in the EU, every system that holds it has to be searched, so a silo you forgot about becomes a compliance gap, not just an inconvenience. (This is not legal advice.)

Worked example: how many people do we actually have?

Sample data: a product has a website, an iOS and Android app, and logged-in accounts. Four months of data:

SystemWhat it countsCount
Web analyticsBrowser visitor IDs that logged in1,100
App analyticsDevice IDs that logged in900
Sum of the two"Identified customers" if you add the dashboards2,000

Now join them on the one thing both know after login, the user ID:

  • The 1,100 logged-in browser IDs belong to 700 distinct users (about 1.57 browsers per person: 1,100 ÷ 700).
  • The 900 logged-in devices belong to 600 distinct users (1.5 devices per person: 900 ÷ 600).
  • 400 of those users appear in both web and app.
  • Distinct identified users: 700 + 600 − 400 = 900.

How to recompute: the union of two sets is the first plus the second minus their overlap. The naive total of 2,000 overstates real people by 2,000 ÷ 900 = 2.22 times. Anything computed per "customer" on the naive number, such as revenue per customer or conversion per customer, is understated by the same factor. And 300 of the 700 web users (700 − 400) do not appear in the app data at all, while 200 of the 600 app users (600 − 400) do not appear in the web data. Those gaps are invisible until the data is joined.

One honest limit: this only fixes the people who log in. Visitors who never identify themselves stay anonymous, and an anonymous browser on a laptop cannot be tied to the same person's phone. Joining reduces the problem; it does not erase it.

How do you connect the silos?

Three steps, in this order. Buying a platform before step 1 just creates a larger silo.

  1. Choose the join key and write it down. Usually this is a stable internal user ID issued when someone signs up or logs in, not an email address that can change. Decide who issues it and that every system will carry it. Treat raw personal data with care; see PII before putting emails or phone numbers in analytics.
  2. Make every system carry the key. For each tool, record the event or field where the ID is attached, usually at login. Document it in a tracking plan so a new tool does not start a new silo. A user ID that is set at login is the minimum; identity resolution is the process of merging records that belong to the same person once you have it.
  3. Bring the data together and agree what counts. Either send events from all platforms into one analytics project (so the join happens at collection), or load them into a warehouse or a customer data platform. Then agree one written definition for "active user", "customer" and "conversion" and put an owner next to each.

Check the result with the same test as the example: does the union of the systems equal the sum minus the overlap? If your distinct-user count is still close to the sum, the join is not working.

What are the common mistakes?

  • Treating the tool as the fix. A new platform with no shared ID and no owner is one more silo.
  • Joining on email only. People change addresses, share them, or use different ones in different systems. Prefer an internal ID, with email as a secondary key.
  • Ignoring anonymous traffic. Joined, logged-in users are a subset. Report logged-in and anonymous separately instead of pretending everyone is identified.
  • Merging records without a way to undo it. A wrong merge combines two real people. Keep a log of merges and make them reversible.
  • Fixing the data but not the definitions. If two teams still define "active" differently, the joined data still disagrees.
  • Removing deliberate separation. Regional or privacy-driven separation may be required. Plan a compliant way to report across it rather than deleting the wall.

Data silos in TapCub

TapCub's answer to the web-versus-app split is to keep those platforms in one project from the start, and to give logged-in people one profile. It does not pull data out of your CRM, billing or helpdesk systems on its own.

  1. One code, one project. The website script, the iOS, Android, Flutter, React Native and HarmonyOS SDKs and the mini program SDKs all send events into the same project, using the same event and user model. Web and app data are therefore not separate silos to join later.
  2. Attach the login ID. After a user logs in, call tapcub.login('u_123', {...}). The person then appears in All-platform analytics › Users › People with one timeline of behavior, and earlier anonymous events on that device are filled in. This needs identified mode, which requires the Pro plan or above and, in the console, a contact with support to switch on; anonymous visitors never get a profile.
  3. Check what was merged. All-platform analytics › Users › Identity map lists every bind, split and rejected ID, and shows how many duplicate counts the merge removed. That is the live version of the 2.22× check in the example.
  4. Plan the events. Record which events each platform sends in All-platform analytics › Data › Tracking plan, so a new platform follows the same names.
  5. Bring in the rest. The server-side HTTP interface can send events from backend systems, and the Data import page takes CSV events and user properties. TapCub is not a CRM or a customer data platform, and it does not sync with one automatically.

Sources

Was this explanation clear?

We use your answers to fix and expand this page.

Su He

Data Analyst

Explores traffic sources, conversion funnels, and retention, using clear metric definitions and concrete examples to explain what the numbers mean.

FAQ

Questions people ask

Short answers to the follow-up questions that come up most.

Still have a question?

Chat with the team behind TapCub — we usually reply within a few hours on business days.

What is a data silo in simple terms?

It is a pile of data that only one team or tool can see. Web analytics, an app tool and a helpdesk each hold their own records of the same customers, and nothing links them, so every system gives a different answer.

What causes data silos?

Teams buying their own tools, no shared identifier across systems, separate platforms for web and app, mergers that bring in other databases, regional or privacy rules, and spreadsheets used as the link between systems.

How do you break down data silos?

Choose a stable join key such as an internal user ID, make every system carry it, then bring the data together in one project, warehouse or customer data platform and agree written definitions with owners. Tools alone do not fix it.

Are all data silos bad?

No. Some separation is deliberate, for example keeping data in separate regions for privacy rules. The problem is separation nobody chose and nobody owns, so plan a compliant way to report across the walls you keep.

See this number on your own site

Add one script tag for web and apps, and the first reports fill in within minutes.

See your data clearly. Find your growth.

Every click, backed by data. Install one line of code and see your first numbers in a minute.

No credit card · Free plans for analytics and chat · Cookieless analytics