Measurement Planning
Part of Ecommerce analytics operations
Planning an analytics migration without breaking historical comparisons
Protect ecommerce trends during an analytics migration with saved definitions, overlap checks and clearly labelled breaks.
Plan an analytics migration around the decisions that use a time series. Preserve old definitions and extracts, map each measure to its replacement, compare both views over overlapping dates where possible, and mark breaks that cannot be explained. Keeping an old label on a new report does not make the measures equivalent.
Identify the comparisons to protect
List the recurring comparisons that matter, such as orders, product sales, checkout progress and campaign views. For each, record the old source, counting unit, eligible population, date field, time zone, currency, filters, refresh cut-off and decision owner. Preserve a dated extract and its definition before changing collection or reporting.
Check available history before planning a bridge. Review the property's retention settings and any existing exports; those settings and exports determine what can actually be compared.
Bridge measures one at a time
| Question | Evidence to keep | Decision |
|---|---|---|
| Does the replacement count the same thing? | Old and new calculations, units and included records | Join the series only if the definitions align. |
| Can a difference be explained? | Overlapping extracts and selected source records | Record a mapping or show a break. |
| Is a long trend still usable? | Change date, coverage and unresolved gap | Label comparable segments separately. |
For purchase reporting, compare the order milestone, item amount, currency and refund treatment. Google's ecommerce guidance lists purchase and refund as measured actions, describes item details, and says to set currency at the event level when value is sent.
Use overlap without overstating it
If both systems can run together, compare the same dates, channels and eligible orders. Keep old and new totals side by side and classify differences by definition, coverage, processing state and unmatched records. Record extraction times and repeat provisional comparisons when appropriate.
A similar aggregate can hide offsetting errors. Where identifiers and access permit, inspect representative purchases, cancellations, refunds and orders near the cut-over. If record-level data is unavailable, state that the bridge was assessed only at aggregate level.
Release with a visible boundary
Record the cut-over date, approved new definition and any historical period transformed under a documented rule. Do not silently fill missing old data or relabel old measures as though they followed new collection rules. Keep the old extract, mapping decisions, exceptions and approver with the new report.
After launch, monitor missing fields, unmatched orders and late revisions. The decision owner should know which periods remain comparable and which questions the gap prevents the series from answering.
