In a survey of more than 100 SAP customers and experts by data migration specialist Xmateria, reported by SAPinsider in July 2026, 57% said data migration regularly delays their SAP projects. When it does, the typical slip is 10 to 25% of the planned timeline, and nearly a third of the delayed projects overran by up to 50%. Bloor Research puts it more bluntly: more than 60% of data migration projects overrun on time, budget or both.
Across 30+ years of SAP and ERP deals, the story behind those numbers has been remarkably consistent. Data is rarely late because the tools are poor. It is late because too few full rehearsals were planned, the ones that were planned had no exit criteria, and nobody priced the time between runs to fix what each one found.
What a data run is, and what it isn't
A data run, also called a mock load, test load or trial conversion, is a complete pass through the migration: extract from the legacy systems, cleanse and transform, load into the new system, then reconcile back to the source. Loading a few hundred records for a demo is not a run. A run ends with a reconciliation the business signs, or it hasn't finished.
The short answer: three plus one
For a typical greenfield implementation with a single go-live, we plan for three mock loads and a full dress rehearsal, after an early profiling run. With fewer than that, the moment the business first reconciles real volumes on production-like data tends to arrive during cutover weekend. SAP's own guidance, published by SAP PRESS, recommends at least two productive load tests, data cleansing completed by the last of them, and the whole migration and cutover tested at least once before go-live.
Each run has a different job. Click through the ladder to see what each one should prove, and the exit targets we typically agree with clients.
Six steps from legacy data to go-live
Select a run to see its purpose, volume and exit criteria.
Why every run needs a gap after it
A run is only as useful as the fix window that follows it. Each pass exposes mapping errors, missing values, duplicate records and transformation rules nobody wrote down. If the next run starts before those are fixed and re-tested, it simply finds the same defects again. In our experience, a full cycle of load, reconcile, fix and re-test takes three to five weeks on a mid-sized programme, and that time has to sit in the plan, not in somebody's weekend.
The shape you are looking for is simple: load success climbing towards 100% while open data defects fall towards zero. When the curves don't move between two runs, the problem is usually cleansing ownership, not technology.
Quality should improve with every run
Load success rate and open data defects across a typical programme. Hover or tap a run for values.
What changes the number
Three plus one is a starting point, not a rule. More legacy sources, poorer data, regulated processes, a tight cutover window or several go-live waves each add risk that only another rehearsal retires. A brownfield system conversion works differently again: the "runs" are technical conversions of a copy of production, and two full mock conversions before the dress rehearsal is the usual minimum. Try your own programme below.
How many runs does your programme need?
Answer seven questions. The plan updates as you go.
The cheapest line to cut from a bid
When suppliers compete on price, data runs are one of the easiest things to trim. Each run is a few weeks of a small team, so dropping one takes a visible amount off the total. Putting the time back once go-live slips is never small, because by then the whole programme team, client and supplier, is waiting on the data.
One more run, or a later go-live?
Enter rough figures for your programme. Slip scenarios use the range reported in the Xmateria research.
Two bids, same headline price
This is where data runs become a supplier selection issue rather than a delivery detail. Two proposals can arrive within a few percent of each other and describe very different programmes. The difference is rarely in the executive summary. It is in the migration approach, the RACI and the assumptions.
What the price includes is the real comparison
Bid A
- ×Two mock loads, volume not stated
- ×"Client responsible for data cleansing"
- ×No exit criteria per run
- ×Repeat runs treated as change requests
- ×Dress rehearsal "if time permits"
Bid B
- ✓Profiling run, three mocks, full dress rehearsal
- ✓Shared cleansing plan with named owners
- ✓Load and reconciliation targets per run
- ✓One repeat run priced in as contingency
- ✓Rehearsal timed against your cutover window
Ten questions to ask every supplier
Put these to every bidder in clarifications, and score the answers. Vague answers are an answer too.
Can your preferred supplier answer all ten?
Tick each question your preferred supplier has answered clearly in writing.
A note for system integrators
If you are on the bidding side, make the runs visible. Name each one, give it exit criteria, show the fix windows and price a contingency run. Evaluators comparing two similar prices will trust the proposal that shows how data actually gets to go-live, and it protects your margin as much as the client's timeline.
Download the playbook
Our free Data Runs Playbook puts all of this on paper: the run ladder with purpose, volume and exit criteria for each step, the sizing rules behind the planner, a cost-of-delay worksheet, a data RACI for client and supplier, a run-readiness checklist and the ten supplier questions as a scoring sheet. Every statistic is sourced.
The Data Runs Playbook: Planning Mock Loads That Protect Your Go-Live
10 pages. No sign-up, no email required.
Choosing an ERP supplier? Our buy-side tender support builds the data track into your requirement, so every bidder prices the same runs and you compare like with like. Book a free discovery call or email sales@dealcraftadvisory.ie.
Sources
- SAPinsider, reporting Xmateria's research "The True Cost of Data Procrastination", July 2026. More than 100 SAP customers and experts surveyed over six months.
- Bloor Research, Data Migration.
- Gartner, Data Quality, citing Gartner research from 2020.
- SAP PRESS, What are the data migration phases for an SAP S/4HANA project?
- SAP Community, Planning the data migration track of an SAP project.
Run targets, cycle durations, the planner's sizing rules and the illustrative chart reflect our practitioner experience, not a published standard. Agree your own criteria in the contract.