Stalled Data Migration: Backing Up Data, Rescheduling Deadlines

Stalled Data Migration: Backing Up Data, Rescheduling Deadlines

A data migration is considered to have reached an impasse when the original schedule has already been postponed twice and no one on the project can say for certain how many records will actually be transferred without errors. At this stage, another sprint using the same method won't help; instead, the team needs to step back and address the root cause.

For senior management and IT leadership, the most important thing at this stage is ensuring that the data is transferred to the new system accurately and within a reasonable timeframe, without the next test run revealing the same errors again.

Short answer

A stalled data migration can be gotten back on track in four steps: a data quality audit that identifies the actual gaps, a systematic duplicate cleanup with clear responsibilities, four to six test migrations instead of a single dress rehearsal, and a reconciliation with fixed “go” and “no-go” criteria before the cutover. Skipping this sequence and setting a new deadline right away usually just leads to the next delay.

How to Tell If a Data Migration Has Stalled

Two signs almost always appear together in such cases. The go-live date has already been postponed without the reason being documented in the project plan, and test runs repeatedly fail at the same points, usually when merging customer or supplier data. Often, when asked, no one on the project team can provide a reliable figure for how many data records are currently being imported without errors.

These signs are not a minor technical issue. According to a According to a Trovarit analysis, more than 60 percent of implementation problems in ERP projects stem from data migration. A more detailed survey is forthcoming 36 percent of project managers who report specific problems with data migration, the most common individual problem in the entire project in both analyses.

more than 60 percent

According to Trovarit, so many of the implementation problems in ERP projects stem from issues with data migration.

Trovarit Study 2024, cited from erp-software.org, 2026

The Data Quality Audit as the First Step

Before the next test run, a data quality audit checks three things in the source system: the completeness of required fields, the plausibility of values, and the number of duplicates. The result is a report listing error categories in order of priority, which the project team addresses in collaboration with the business units.

One Rule of thumb from SAP migration practice It is said that if 5 percent of issues are found during data analysis, the actual migration will require about 25 percent more effort. The audit is therefore worthwhile even if it puts additional strain on the schedule in the short term.

Resolve duplicates and clarify responsibilities

Customers with slightly different names, suppliers with multiple entries per location, and material master records with inconsistent designations rarely stand out in day-to-day operations because employees usually know which data record is the correct one. However, when migrating to the new system, this tacit knowledge no longer applies, as a target system requires a single, unique version of each data record.

Those who leave the data reconciliation process entirely up to the tool will generally end up with a result that is technically correct but substantively incorrect. The decision as to which of two customer records is the master record belongs with the business unit, not solely with the IT department. Designating a specific data owner for each object type (such as customers, suppliers, and items) streamlines these decision-making processes in day-to-day project work.

Test migrations: four to six runs before go-live

A single test run shortly before the cutover weekend is not enough to build confidence in the new data set. Four to six complete test migrations, each using the same verification procedures as on the go-live date, will show whether the error rate actually decreases from run to run or whether the same problems recur.

Every run requires a thorough evaluation: how many data records were imported, how many had to be manually corrected, and how long the run took from a technical standpoint. Without this logging, it’s impossible to determine in the end whether the last test run was actually better than the first, or if it just feels that way.

Sample Analysis of Four Test Migrations
Run-through Typical Focus
1 Technical feasibility, approximate error rate per data object
2 Removed duplicates, initial reconciliation of master data
3 Movement data and open items, time measurement of the run
4 to 6 Repeat until the target quota is reached; dress rehearsal under production conditions

Reconciliation and the Go/No-Go Decision

After each test run, the reconciliation compares totals and quantities between the old and new systems, including open items, inventory levels, and order values. If the numbers differ, this is not a rounding error but an indication of an incorrect mapping.

The results of the most recent test migration yield a clear “go/no-go” list with fixed thresholds, such as a maximum allowable number of open data errors or a minimum percentage of automatically validated data records. These thresholds must be established before the final test run, not just the evening before the planned cutover, when the pressure is at its highest and every number suddenly seems negotiable.

Where this approach reaches its limits
A thorough audit and multiple test runs reduce the risk, but they do not eliminate it entirely. With very large or historically accumulated data sets, there will always be a residual percentage of cases that can only be resolved manually, and these cases often do not surface until the final test run. There is no reliable rule of thumb for determining the exact point at which further test runs are no longer economically viable. This depends on the specific case and the potential for damage caused by erroneous data in the respective process.

What Delays Actually Cost

Every additional delay costs twice as much because the project team remains tied up and the funds originally budgeted for the data migration are usually insufficient. In budget analyses for ERP projects, the actual cost of data migration is often significantly higher than the planned amount, in an analysis, at about 12 percent of the total project cost instead of the budgeted 5 percent. The article provides more information on the causes of budget overruns in IT projects in general IT Project Budget Overrun: Analyzing the Causes Before Assigning Blame.

Those who recognize this discrepancy early enough can realistically reschedule deadlines instead of postponing them a third time by a few weeks and once again putting off the same uncertainty. A realistic new deadline based on actual test results will erode management’s confidence less than yet another last-minute postponement.

Have a stalled migration project reviewed by an external party
If no one internally can say how many records are being imported without errors, an outside perspective can help more quickly than the next internal crisis workshop. We’ll assess the status of your project and identify realistic next steps.

Get in touch

Frequently Asked Questions

How many test migrations are realistically necessary before go-live?

In practice, four to six full test migrations have proven effective, each using the same test routine as on the actual go-live date. The first runs primarily uncover structural issues and residual duplicates, while the later runs check transaction data, open items, and the actual runtime under realistic conditions. Fewer runs are rarely sufficient; significantly more runs usually indicate that the root cause analysis from the preliminary audit was incomplete.

Who should make the go/no-go decision before the cutover?

The decision should be made by a steering committee comprising representatives from IT, the affected departments, and senior management, not solely by the project team, which is under operational time pressure. The thresholds for data quality and error rates should be established and documented before the final test run, so that the decision made on the cutover weekend is based on numbers rather than the mood in the room. A clear “no” costs less than a premature “yes.”.

What would be the cost of postponing the migration deadline again?

In addition to the visible costs associated with contracted external consultants and the project team, the costs of the legacy system, which continues to run in parallel, accumulate; these are often underestimated because they get lost in the ongoing budget. Budget analyses of ERP projects show that the actual cost share of data migration is often far higher than originally planned. Anyone who postpones a deadline without addressing the root cause risks having to postpone it a second time, with the same costs and less confidence in the project.

torck takes on project rescue assignments in which migration to the new system has brought the entire schedule to a standstill, often as part of a temporary interim engagement that conducts a root-cause analysis without having been part of the original project team. This distance is helpful because an audit is more honest when no one has to defend their own previous estimates. The article describes typical crisis patterns in ERP projects as a whole ERP Projects in Crisis: Typical Patterns and Solutions. For more information on the acquisition process, see the page Interim CTO.

Schedule a meeting

Questions about this post?

Just a couple of sentences about your situation will suffice. The person responding builds these kinds of systems himself.

We'll respond within one business day.torck · Code mit Drehmoment
Florian Blischke
Managing Director at torck · Over 20 years of software development experience
Florian Blischke is the managing director of torck, based in Maxhütte-Haidhof near Regensburg. He has been working in software development for over 20 years. His focus is custom software, data and AI for industry and retail, ranging from the integration of physical processes and IoT to cloud architecture and data- and AI-driven systems. torck works with its own nearshore team in Rabat. At torck, he oversees projects including the Jouvoli energy platform and the KVM Fleet fleet management product. torck develops software at its locations in Maxhütte-Haidhof, Vienna and Rabat and focuses on software that actually works in operation.

Are you facing the same question?

Since 2017, we have been building software for industry and retail from Maxhütte-Haidhof near Regensburg, with teams in Vienna and Rabat. An initial consultation takes 30 minutes and costs nothing. Afterwards you will know whether the project is worth pursuing, even if the answer is no.

More Articles

Oracle Java License: Pay for a subscription or switch to OpenJDK

Oracle Java License: Pay for a subscription or switch to OpenJDK

On October 8, 2026, Oracle introduced fixed annual prices for the Java SE Universal Subscription for organizations with fewer than 3,000 employees; as of October 20, 2026, new updates for Oracle JDK 21 are covered by the OTN license. When does the subscription pay off, and when should you switch to an OpenJDK build?

Read more »