A data integration checklist is a set of checks that shows whether the connections between your business systems are ready, running and owned. This one holds 20 checks in five areas, starting with data integration requirements. Each check says what to bring and when it counts as traced. When added up, the marks give a picture of your data integration readiness. AWS’s explanation of data integration describes data integration as the process of achieving consistent access and delivery for all types of data in the enterprise. The list sits under the Signisys use case on bringing business data together.
Last reviewed 2026-11-02 by Manish Sharma.
Data Integration Checklist: What the 20 Checks Test
The checks suit a project in planning or already running. You run them yourself, so the result is a self-check, not an audit or a certificate. In Signisys’ working method, the checks fall into five areas, and both the areas and the tests are our own, not a standard. For the order of work behind such a project, see the seven steps from business questions to a named owner for each feed.
| Area | What it asks | Checks |
|---|---|---|
| Requirements | Is it written down what the combined data must answer? | 1 to 4 |
| Sources | Can each system be reached, with its owner’s agreement? | 5 to 8 |
| Movement | Does each flow run, report failure and rerun cleanly? | 9 to 12 |
| Combined figures | Do the totals match a named source report? | 13 to 16 |
| Ownership | Does a named person answer for each connection? | 17 to 20 |
This list scores the work, but it does not teach it. It does not give the order of work, explain why two reports disagree or compare ETL with ELT. Cleaning values and setting data rules sit with the data quality use case. For background, see how data ends up in separate systems.
Copy the five area tables into a sheet. Then bring the records each check asks for to one session with the people who own the systems.
Data Integration Checklist Scoring Rules
In Signisys’ working method, each check gets one of three marks.
- Traced (2): the team meets the test, and a dated record shows it.
- Partly traced (1): the test holds for some flows only, or the record covers only part of the flows.
- Untraced (0): the test fails, or nobody has tried it.
A check with no dated record scores Untraced, regardless of what the team believes. Some checks may not apply, for example check 8 in a project with one source. In that case, mark them N/A, write the reason beside each and take 2 points off the maximum for each one. If no check is N/A, the maximum is 40.
Before you start, choose a sample, such as the two flows behind the board report, and write it down. Then score only what the sample shows, and leave the rest unmarked rather than assumed. No target score or pass band defines data integration readiness here.
For example, take check 13 on last quarter’s sales total. If the combined total equals the sales report and you save the comparison with its date, mark Traced. Where it matches for one region but not another, mark Partly traced. If nobody has run the comparison, mark Untraced.
Key Takeaway Count a check only when you can follow one figure back to its source system, with the record to show it.
Data Integration Checklist, Area 1: Requirements
The four checks below test whether the purpose exists in writing before any data moves. Data integration requirements, in this list, are the questions, fields, terms and match reports that the project agrees in advance.
| Check | What to bring | Counts as traced when |
|---|---|---|
| Check 1: Each report or use has a written business question. | One line per report, saying what the combined data must answer. | If someone outside IT reads the line, they can say what the report is for. |
| Check 2: The data each question needs is named. | A field list per question, with the system each field comes from. | Each question has its field list, and each field has a source. |
| Check 3: Shared terms carry one agreed meaning. | A terms sheet with one sentence and one owner per term. | Two owners, asked apart, give the same meaning for “customer”. |
| Check 4: A source report is named for matching. | The existing report whose total the combined data must equal. | Each question names that report and the period to match. |
Data Integration Checklist, Area 2: Sources You Can Reach
Each question in the data integration requirements of area 1 points to one or more systems, as well as their owners. Next, these checks test whether you can reach them, with each owner’s agreement.
| Check | What to bring | Counts as traced when |
|---|---|---|
| Check 5: Each source system has a named owner. | A list of systems and the person who approves access to each. | Each owner on the list confirms the entry in writing. |
| Check 6: Each system has an agreed route out. | The route for each system in writing: export, connection or database access. | The owner has agreed to that route by email or signature. |
| Check 7: The update timing of each source is known. | A note of when each system last updated the data you need. | The note gives a date and time from this month. |
| Check 8: A shared key links records across systems. | The ID that marks the same customer, product or supplier in each system. | A sample of records you pick matches on that ID in all the systems. |
Data Integration Checklist, Area 3: How Data Moves
These checks test each flow, that is, each route from one source to the shared store. In Signisys’ working method, check 9 counts as traced when a flow’s method and the oldest data its users accept are written side by side, and the two fit.
| Check | What to bring | Counts as traced when |
|---|---|---|
| Check 9: Each flow has a chosen method. | The method per flow (scheduled batch, near real time or direct connection) and the data age its users accept. | Both are written, and the method keeps data within that age. |
| Check 10: A failed run reaches a named person. | The record of the last failed run and who heard about it. | The person in that record replied on the same day. |
| Check 11: A rerun gives the same result. | Totals before and after one deliberate rerun. | The two totals agree, with no duplicate records. |
| Check 12: Reports read from one shared store. | A list of reports and the store each one reads. | Each report reads that store, and its edit log shows no hand changes. |
Reshaping data before or after loading is a separate question that this list does not cover.
Data Integration Checklist, Area 4: Do the Combined Figures Match?
These checks compare the combined data with the reports people relied on before the project began. They test matching only, rather than cleaning. For why business numbers disagree, see the explainer on agreeing who owns each fact.
| Check | What to bring | Counts as traced when |
|---|---|---|
| Check 13: Totals for one period agree with the source report. | The combined total and the source report total for the same period. | The two totals agree, and the comparison is saved with its date. |
| Check 14: Each difference has a written explanation. | A log of differences found and the cause given for each. | Each open difference has a cause and a person against it. |
| Check 15: Picked records trace from source to report. | Records you picked, followed from the source screen to the report line. | Each picked record appears in the report with the same values. |
| Check 16: The combined data shows its refresh time. | A screenshot of the report showing when the data last loaded. | The report states the date and time when the data last loaded. |
Wrong, missing or repeated values in the sources, however, are a different matter. For cleaning and rules, see the use case on improving data quality.
Data Integration Checklist, Area 5: Who Owns Each Connection
Finally, these checks test whether someone answers for each data feed after launch.
| Check | What to bring | Counts as traced when |
|---|---|---|
| Check 17: Each connection has a named owner. | A list of connections with an owner and a deputy for each. | Each connection has both, and both have confirmed it. |
| Check 18: A stopped feed reaches a person. | The last alert about a stopped feed and the reply to it. | Someone replied to that alert, and the reply carries a date. |
| Check 19: Source field changes get a review first. | The last change to a source field and its review note. | The review took place before the change reached the feed. |
| Check 20: The terms sheet has a set review date. | The terms sheet with its last review date. | The sheet shows a review within the period the owners agreed. |
Key Takeaway Fix each Untraced mark before you improve a Partly traced one.
Reading Your Data Integration Readiness Score
Add the marks for each area to see data integration readiness, then write the number of Untraced marks beside each one. Then read the areas in order. Areas 1 and 2 come first, because the later checks depend on their answers. A flow in area 3 needs a source from area 2, and a match in area 4 needs a report named in area 1.
Overall, the total can show progress in data integration readiness so far. By contrast, the Untraced count per area shows where the project has gaps. A sheet with several Untraced marks in area 1 needs those data integration requirements settled before anyone scores area 4 again.
Save the scored sheet with its date and its sample. The next run then shows which marks moved, and why.
What a Data Integration Checklist Will Not Tell You
This list does not replace an audit, and it is neither a certificate nor legal advice. It scores the readiness of your data integration records rather than the quality of any tool. The scope of this data integration checklist is moving and combining data. By contrast, linking applications and business processes is a different job. For that, see connecting applications and processes. This article claims no data rule for India or the GCC.
A second view helps when the owner of a source is out of reach, or when the area 1 terms stay disputed.
Signisys provides services for connecting and analysing business data through its business intelligence and analytics solution and its data and AI services. Read this paragraph as a description of a Signisys service, not neutral advice.
Frequently Asked Questions
What does a data integration checklist check?
Can the score be high while the figures are still wrong?
Which area depends on which?
Can we copy these tables into our own sheet?
How is this checklist different from a guide to the steps?
Who signs off an Untraced mark?
Security Insights, Delivered.
Practical, vendor-neutral analysis of threats, tools, and architecture decisions.