During a systems migration, the historical data that opponents rely on is left behind, so their expertise rests on memory nobody can check and their arguments lose their evidence.
An example
Cutover was fixed for the first day of the fiscal year as a meat-processing company moved its six plants onto a new production and quality system. A plant manager with twenty years in had spent the previous year arguing against closing his plant's smoked-products line, using fourteen years of yield and spoilage data that showed the line was profitable every winter quarter. In August, a change request cut the historical data load to "three years, for budget reasons." The cutover date stayed. Older records would remain in the retired system, which was due to be switched off in December. At the January portfolio review he cited the winter pattern, and the operations vice president asked where the data was. He answered that he remembered it clearly. The vice president noted that the system showed three years, two of them unusual. The line was scheduled to close in May.
Why it works
Without its records, the opponent's expertise loses its evidence: a long history becomes "trust me, I remember," which an audience used to dashboards cannot evaluate, so the argument looks worse. Arguing from history becomes costly, because dissenters must rebuild the data themselves from retired systems and paper. Expertise anchored in records the experts did not hold themselves is most exposed. Project governance that accepts cuts to historical data without escalation, no regulator or auditor requiring continuous records, and a fast cutover complete the picture, since a long parallel run gives dissenters time to extract what they need. The organization loses that memory for good, which costs it when an old pattern recurs.
The thinking behind it
George A. Akerlof The Market for "Lemons": Quality Uncertainty and the Market Mechanism (1970) shows that when buyers cannot verify quality, they discount everything to the average. Once the records are gone, a veteran's knowledge becomes a claim nobody can verify, and the audience cannot tell sound memory from nostalgia, so it discounts all of it. Extracting the records before cutover keeps the claim checkable, which is the heart of the defense.
Carl I. Hovland Communication and Persuasion: Psychological Studies of Opinion Change (1953) reports that the same message is judged by the perceived expertise and trustworthiness of the person delivering it. A plant manager's expertise was credible because records stood behind it. Without them, the same argument from the same person persuades far less, so a migration's scope decision is also a decision about whose credibility survives it.
Warning signs
Historical data loads get cut back in a change request while the cutover date stays fixed. The cut falls hardest on the records one group cites in its objections. The retired system has a switch-off date soon after cutover. Budget is the stated reason, and nobody escalates it. After the migration, arguments based on history start being called anecdotal. The parallel-run period is short.
How to defend yourself
Raise the scope change when it happens, in the migration's own governance forum, while it can still be reversed.
- Extract copies of the records you depend on before cutover, into files you control, with their source noted.
- Press for a longer parallel run, and for the retired system to stay readable until the decisions that rely on its data are made.
- Find out who proposed the reduced scope and who signed it off, and ask for the reasoning in writing.
- Where records must be kept continuously for regulatory, contractual, or audit reasons, bring in the people responsible for that.
How a leader shuts it down
Treat any cut to the historical data being migrated as a decision with business consequences, reviewed beyond the project team, and ask whose arguments depend on the data being dropped. Keep retired systems readable for a set period. Require that records be preserved somewhere before any switch-off, because the institutional memory lost in a migration does not come back.