Not every ugly reporting center should be torn down, and not every platform that still opens is worth stacking on. Look at three assets: whether definition documents still exist, whether extracts still reach documents, and whether anyone can change the schedule. If those three are alive, a reskin, drill-down, and tighter permissions usually cost less than a rebuild. If all three are empty, more patches only paint another layer on forked algorithms.
Finance and operations already accepted the KPI definitions; source APIs still live; the pain is hard to use, no phone, messy permissions, a few key charts that cannot drill. Close algorithm versions on the current subjects, add detail, cut unused reports. Risk is lower, and the weekly meeting does not go suddenly blind.
The “platform” is a scheduled Excel dump; one KPI has three SQL texts; the license locks people in; the vendor or implementer is gone; opening a chart never opens a document. Adding tables will cost more than rebuilding, and the more you patch, the less you dare turn the old one off.
A common trap: keep a ten-year-old cube so leadership keeps familiar menus. If no menu item opens an order, familiarity is only familiarity with disappointment. A rebuild can keep KPI names and the weekly agenda, move URLs and permissions, and leave the wrong algorithm behind. Put the few sheets sales and finance actually use on the migration list first. Browse-only reports can wait. Choose off-the-shelf BI vs custom after you see the assets; see Off-the-shelf BI vs a custom dashboard: what actually differs. Mid-size companies should not use a rebuild as an excuse to buy a big platform; see Do mid-size companies have to buy a data platform.
Extending is not changing colors. Write the drill path, the one algorithm, and extracts you can reconcile on failure. Otherwise you only swapped the decorative tool and the disease stays; see Buying BI as decoration is the same as doing nothing. Time and money depend on whether you stop at repair or rebuild; see What dashboards or a data platform cost, and how long they take. If numbers stay wrong and the platform cannot pull a document, trace first, then choose the cut; see The system is live and numbers are still wrong. How to trace it. A license expiry is not an automatic rebuild, but get export and algorithm rights into your hands. When the account dies, “how we actually calculate” dies with it.
During cutover, keep one official number. The old platform can be read-only for comparison. The new one ships the weekly pack. Two systems that both edit and both claim to be official hurt more than any technical debt. When DaXi assesses an old platform, we open the schedule, pull a KPI and SQL list, ask sales and finance which sheets they actually open, then recommend extend or rebuild—not a default teardown. Send live report URLs and the few sheets you dare not turn off to the Data service page. Overall sequence, see From brief to launch: how a data project actually runs.
Expiry is not an automatic rebuild, but export and algorithm rights must be yours. When the account stops, “how we calculate” stops with it. That is asset protection before you extend.
On a rebuild, migrate the three to five sheets the weekly meeting actually uses. Do not move two hundred old reports into a new tool as-is. If nobody looks after the move, you bought a more expensive empty shell.
Do not turn the old platform off before the new one is accepted. An overlap that lands on a semi-annual count costs more than the rebuild. Pick a cut date away from close and peak season.
Send the live report URLs and the few sheets you dare not turn off. We will recommend extend or rebuild, not a default teardown.