Reconciling two sources that disagree
A brand's shipment data and a measured channel's scan data will not agree, and treating the gap as an error is the wrong starting point. They measure different events at different points in the chain, over different universes, on different calendars.
The productive question is not which one is right. It is what the difference is made of.
What are the structural gaps?
Timing, universe, and definition. Shipments leave a warehouse before they sell; scan data records a sale. Shipments cover every customer; a measured channel covers the stores in it. And a case shipped is not a unit sold until somebody normalises pack configuration.
Each of those is a legitimate difference and each can be quantified. A reconciliation that accounts for all three usually closes most of the gap.
How does inventory in the chain distort it?
By moving the timing. Stock sitting with a distributor or in a store's back room was shipped and has not sold, so a period of trade loading shows shipments far above scan sales, and the following period reverses.
That pattern is diagnostic. A gap that oscillates around zero over a longer window is a pipeline effect; a gap that persists in one direction is something else.
What does a persistent gap suggest?
Usually universe. Shipments include channels the scan data does not measure, and if a brand sells through several channels the measured one is a subset by construction. The gap is then not a discrepancy but a channel mix.
Other persistent causes: unmeasured stores within the channel, product sold under a code that is not mapped, and returns handled differently in the two systems.
What is the reconciliation method?
Work down a ladder. Start with total shipments, remove other channels, convert cases to units, shift for pipeline timing, and adjust for the measured universe's coverage. Each step has a number and the residual at the bottom is the part actually unexplained.
The value is in the ladder rather than in the residual. A brand that can state the size of each step understands its own chain; one that cannot has a single unexplained number.
How should disagreement be presented internally?
As a decomposition rather than as a dispute. "Scan sales are below shipments by an amount consisting of pipeline, unmeasured channels and coverage" is a sentence that ends an argument. "The data is wrong" continues it.
The decomposition also makes clear which number answers which question. Shipments answer what the business sold; scan data answers what shoppers bought in a defined channel.
What about comparing two scan providers?
Different problem, same discipline. Two providers differ on universe, weighting, definitions and week convention, and a reconciliation walks through the same list. Rarely will they agree exactly, and consistent direction is usually the realistic standard.
Where they disagree on direction, that is worth pursuing, because one of them is describing something the other is not seeing.
What should be agreed before the comparison?
Which source is the internal standard for which purpose. A business that uses shipments for supply and scan data for market performance has no conflict; one that uses whichever is more favourable has a governance problem rather than a data problem.
What is the honest limit?
That a residual will remain. Perfect reconciliation is not achievable across systems built for different purposes, and a small unexplained remainder is normal. Chasing it to zero costs more than it returns.
Frequently asked questions
Which number should go in a board pack?
Both, labelled, with the gap explained once. Presenting one without the other invites the question at the worst moment.
How often should a reconciliation be run?
Quarterly for most brands, and after any structural change — a new channel, a pack change, a distributor switch. Those are the events that move the ladder.
What if the gap suddenly changes?
Investigate immediately, because a step change usually has a specific cause: a listing, a delisting, a mapping error or a trade event. A sudden gap is more informative than a persistent one.
Does this apply to a small brand?
Yes, and it is easier, because there are fewer channels and fewer codes. A small brand can usually reconcile in an afternoon.
Can distributors supply the middle of the ladder?
Sometimes, and it is worth asking. Distributor withdrawal data fills the pipeline gap directly and is the single most useful additional input.
Is one source more accurate?
Neither is more accurate in general; they are accurate about different things. Shipments are near-exact about what left; scan data is near-exact about what sold where it was measured.
How do I handle returns?
Consistently, and by deciding which side they are recorded on. Returns handled in one system and not the other are a recurring source of gap.
Does this affect forecasting?
Considerably. Forecasting shipments from scan trends without modelling the pipeline is a known way to produce a plan that is systematically early or late.
What is the first step for a brand that has never done this?
One period, one brand, one channel, down the ladder. The first attempt usually finds something structural that has been distorting every comparison for years. General retail definitions published by the Census Bureau are a useful calibration for what a careful definition looks like.
### Who should own the reconciliation?
One person, with both datasets and the authority to ask questions of supply and of the provider. Splitting it between a commercial team reading scan data and a supply team reading shipments guarantees that the ladder is never built.
The output belongs in a standing document rather than in an email, because the structural components change slowly and rebuilding them each time is most of the work.
Working with retail data? See NRS Insights data solutions or contact NRS Insights.