For most retailers, switching people counting systems raises an immediate concern: what happens to the data we already have?

You may have five years of hourly counts, seasonal patterns your teams know well, budgets based on last year’s figures and board reports that depend on year-on-year comparisons. Starting again can sound like losing all of that context.

But importing historical counts directly into a new system creates a different problem. It combines measurements produced by two instruments with different error patterns and presents them as one continuous record.

There is a better approach. Keep the records separate, measure the relationship between the two systems and restate the historical series for valid comparisons.

What retailers need to know

• Do not merge data from the previous system into the new verified record.

• Start the new record only after the system has been commissioned and tested at each location.

• Run the old and new systems in parallel to measure the difference between them.

• Use that measured relationship to restate the old series for historical comparison.

• Document the limits and uncertainty of every restated figure.

• Plan how a mixed estate will report during a phased rollout.

Why importing old data creates a misleading record

Every people counting system has its own pattern of error. That pattern is not evenly spread across the day or week.

A break-beam counter may miss people entering together. A thermal sensor may struggle to separate individuals in a crowd or as ambient temperature rises. These conditions often occur during the busiest periods, so any undercounting or overcounting can be concentrated exactly where commercial decisions matter most.

If data from the old system is imported into the new one, the combined history contains both error patterns. After the records have been merged, it becomes difficult to determine whether a change reflects visitor behaviour, trading conditions or the change in measurement technology.

The problem becomes more serious when the data is used for comparisons across locations or operators. A benchmark is only useful when participants are measured on a comparable basis. If part of one retailer’s history carries the margin of a different instrument, the resulting position is difficult to defend.

The practical rule is simple: data produced by the previous system should not enter the new verified record. It should not become an entry, baseline, reference average or contribution to a market comparison.

When the new record should begin

The new record should begin on the date the system has been commissioned and verified at that location.

During commissioning, cameras are positioned, counting lines are configured and performance is checked. The system may produce data during this period, but those figures should be excluded from reporting because the measurement was still being tuned.

This has three practical consequences.

First, metrics and baselines should not reach back before the verified start date. If a reference window would normally include an earlier period, it should be shortened and clearly disclosed.

Second, a comparison that requires more history than the new record contains should not be presented as if it were complete. Four months of verified data cannot produce a true year-on-year comparison on its own.

Third, a location should join any wider market comparison only from its verified start date. This prevents unverified measurements from affecting other participants’ positions.

Keep the old history, but do not merge it

The previous series does not need to disappear. It needs to remain separate from the new record.

The useful distinction is between merging and restating:

A restated series can support the year-on-year comparisons the business already uses. It should always be identified as historical data carrying a correction, not presented as part of the new verified record.

It should also remain outside market benchmarks. Restatement helps a retailer compare its current performance with its own past. It does not make measurements from two different systems identical for cross-operator comparison.

Run both systems in parallel

The most reliable way to preserve historical comparison is to operate the old counters and the new system together for an agreed period.

Both systems then measure the same locations on the same days. This makes their relationship visible:

• whether the new system reads consistently higher or lower;

• how large the difference is;

• when the difference increases or decreases;

• which traffic conditions affect it.

This measured relationship becomes the bridge between the old series and the new record.

Avoid relying on one correction factor

Dividing one total by another and applying a single factor across the full history is attractive because it is simple. It is also likely to hide important variation.

An older counter’s error may be concentrated during peak hours, when groups enter together or when environmental conditions affect the sensor. A single factor can appear reasonable at weekly level while still under-correcting the busiest periods.

The correction should therefore be calculated by relevant hour bands and traffic levels. The documentation should also state the level at which the restated data can be used reliably. A correction may support daily or weekly comparison without being precise enough for individual hours.

Every restated figure should carry two sources of uncertainty:

• the declared accuracy of the previous system;

• the uncertainty in the correction calculated during the parallel run.

If a reported difference is smaller than those margins combined, it should not be treated as a meaningful commercial finding.

This exercise is best described as a bridge or reconciliation. It is not a benchmark. A benchmark compares performance with comparable locations or operators, while a bridge connects two measurement systems used by the same retailer.

Questions to ask the vendor

Before committing to a switch, ask:

• Will the new system run alongside our existing counters?

• How long should the parallel period last?

• Which locations and trading conditions will it cover?

• At what level will the correction be calculated?

• How will uncertainty be reported?

• At what level can the restated history be used reliably?

• Will we receive the method and results in a written document?

• How will verified start dates be handled across a phased rollout?

A vendor should be able to explain both the technical method and how the business will report during the transition.

Plan the reporting decision internally

The transition can fail even when the measurement work is sound.

Before the parallel period ends, decide which series is authoritative for each reporting line. Document the decision and communicate it across operations, finance, analytics and leadership.

Without a clear reporting policy, different teams may quote different figures for the same location and week. That creates a governance problem that can continue long after the technical switch is complete.

How to sequence the rollout across an estate

For a multi-site retailer, the order of rollout matters more than the speed.

Begin with locations that have clean historical data and relatively stable trading conditions. These sites are more likely to produce a reliable bridge between the systems.

Avoid beginning with locations undergoing a refit, tenant change, major campaign or other unusual trading condition. In those cases, the parallel period may capture disruption rather than the relationship between the instruments.

The first locations will usually take longer because they establish the bridge and reporting process. Verification still needs to happen at every location, but the transition method can then be applied more efficiently across the estate.

A mixed estate is normal during a phased rollout. It becomes manageable when the reporting rules, verified start dates and responsibilities are agreed in advance.

What the first year looks like

Months 1 to 3

The new system has a short verified record, while the parallel run establishes the correction. Comparisons with previous periods use the restated historical series, with the relevant margins and limits disclosed.

New measures that did not exist in the previous system can be used immediately. Depending on the system, these may include where visitors go, where they stop and how different visitor groups engage with the store.

Months 4 to 9

The new record becomes increasingly useful for internal comparison. Seasonal context may still come from the restated series, provided teams understand how it was produced and the level at which it is valid.

From month 12

The new system has a full year of verified history and can support year-on-year analysis without relying on the bridge for the equivalent period.

The temptation to avoid is forcing a full year-on-year result from a few months of new data. The calculation may produce a number, but it will not represent the comparison that decision-makers think they are reading.

The vendor answer that matters

A vendor offering to import all previous history may appear to provide the easiest transition. Unless the differences between the two instruments are measured and disclosed, that convenience comes at the cost of a trustworthy record.

A stronger vendor will explain why the old and new records must remain separate, then show how a parallel run and documented restatement can preserve the comparisons your business needs.

Ask about historical data early in the buying process. The answer reveals how seriously a vendor treats measurement integrity, reporting continuity and the realities of switching systems across a retail estate.