SPINS panel coverage: projection & suppression

Why SPINS panel coverage matters

The SPINS report says 28% ACV in Natural Channel over the last 12 weeks. Three weeks later, same report, it says 31%. You did not add a single door. You call your SPINS rep, and the answer is that a regional natural co-op just came back into the panel after a months-long gap. Your "growth" was never growth. It was a sample-mix change wearing a growth costume.

That is the sort of thing that quietly ruins an analyst's morning. SPINS data is not wrong, exactly. It is a projection, an estimate built by scaling a sample of stores up to a whole channel, and projections come with rules. Most analysts do not learn those rules from a manual. They learn them by getting burned. This page is the burn list.

Who this is for: the brand-side analyst at a natural, specialty, or wellness CPG brand pulling SPINS every week and presenting to commercial leadership. If your work touches retailer-level reads, post-promo lift, or distribution, the projection and suppression mechanics below should change how you read every number the portal hands you.

The three diagnostic questions for any anomalous cell

Before the methodology, three questions. They cut through most SPINS confusion fast, and the worked example later shows each one earning its place. Worth keeping somewhere you can find them.

1. Did ACV move with sales? If it didn't, you are looking at panel or projection noise, not real demand. A genuine sales crash drags ACV down with it, because stores stop ordering and distribution thins. A suppression or backfill event leaves ACV sitting right where it was.

2. Is the week before or after unusually high? Backfill happens constantly, and it tends to even out across a 4-week window. See a zero followed by a spike? Add the two weeks together. If the sum is close to a normal two-week total for that retailer, that is a backfill, not a real spike.

3. Direct-scan retailer, or distributor-flow estimate? Distributor- flow numbers swing harder, partly from shipping latency and partly because the projection coefficients are larger. A 30% week-over-week move in independent natural through KeHE or UNFI deserves a lot more suspicion than the same move at Sprouts, which scans direct.

The SPINS panel coverage methodology: what's happening under the hood

SPINS pulls the natural and specialty channel from two streams:

1. Direct retailer scan data from natural and specialty chains that license their POS to SPINS: Sprouts, Natural Grocers, and a long list of regional naturals and specialty chains. (Notably, Whole Foods Market does not report scanner data to SPINS; for Whole Foods coverage, brands rely on NielsenIQ — Whole Foods' own analytics provider — or a Circana estimate.)
2. Distributor-flow data from KeHE and UNFI, which captures what was shipped to independent natural retailers (the long tail of independent co-ops, single-store naturals, and specialty stores that don't license their POS individually).

Those two streams get reconciled, deduplicated, and then projected up to stand in for the whole channel. Projection just means scaling: if the sample stores in market X account for N% of estimated channel sales, you multiply the sample number by 1/N to get the total estimate. The catch is that the coefficient doing that scaling is not constant.

It moves with geography, for one. Coverage density is uneven, so the West Coast and Northeast, where natural retail clusters thickly, produce smaller projection multipliers than regions where natural stores are sparse. It moves with the channel slice too. Natural channel projects nothing like Conventional MULO, and the MULO+ extension (Specialty/Natural Enhanced plus Conventional MULO, with Circana data powering the conventional half) stacks two different projection models into one published number. And it moves with time. The coefficients update as the panel itself shifts, so a retailer joining or leaving mid-quarter can revise the coefficient retroactively.

For a total-channel trend line, none of this matters much. Where it bites is week-over-week movement on one specific cell, a single retailer crossed with a geography and a period, because the "movement" you are reading might just be the panel rearranging itself.

SPINS data suppression: when SPINS hides the cell entirely

Projection is half the story. The other half is suppression. When a cell's sample is too thin to publish, either because it would breach a retailer's contract or because the number just isn't statistically trustworthy, SPINS hides it. What you see instead is a blank, a zero, or an "insufficient data" marker sitting where you wanted a real figure.

The thresholds are contract-specific and nowhere in one tidy public reference. In practice, here is where it shows up:

• Banner-level Kroger detail is suppressed in the standard feed unless you've licensed banner-level breakouts (Ralphs, King Soopers, Fred Meyer, Harris Teeter, Smith's, Fry's, and others). Without that add-on, "Kroger total" is your only handle, and you can't separate King Soopers strength from Harris Teeter weakness.
• Independent natural in lower-density geographies often shows zero for multiple weeks running, then a backfilled number when distributor data catches up.
• Small-share categories in MULO can show suppression on retailer × week cells even when the rolled-up category is fine.

Suppression is not zero sales. It is SPINS declining to publish a number. Reading a suppressed cell as a zero is the single most common mistake analysts make in this space, and it quietly drags distribution, share, and post-promo lift calculations downward every time.

Worked example: a brand "losing" a banner

Made up, but it could be any wellness brand. Say it sells through roughly 40% of Sprouts stores nationally. Distribution is steady, velocity is steady. Then the weekly SPINS pull lands like this:

What really happened is dull. SPINS suppressed the W3 cell over a panel data delivery issue, and when the data finally arrived it backfilled those W3 dollars into the W5 number. Read it cold and you see a catastrophic week chased by a 100%-plus rebound. Read it right and every single week was normal. The five-week total is exactly what it should be.

The ACV column gives it away. If sales had genuinely cratered in W3, ACV would have dropped with them, because no stores selling means no distribution. Instead ACV sat at 38–40% the whole time. That column is the load-bearing diagnostic.

And notice all three diagnostic questions firing here. ACV held at 38–40%, so question 1 says no real distribution event. W5 came in at exactly double a normal week, the textbook backfill signature, which is question 2. And Sprouts is direct-scan, so by question 3 this kind of anomaly is an exception worth noticing, not the baseline you would shrug off in distributor-flow data.

How to validate a suspicious number before presenting it

Before a SPINS number goes into a deck headed for a buyer or a CEO, run it through three checks.

First, does it square with velocity? If the brand is in 200 Sprouts stores and the dollar figure works out to $4 per store per week, that is implausibly low for a SKU that normally moves. Do the arithmetic: dollars divided by the ACV-implied store count gives you implied velocity. If that lands more than 20–30% outside the historical range, the cell is suspect.

Second, is there a matching signal somewhere else? If the brand tracks inventory through a 3PL or distributor, a real KeHE or UNFI shipment anomaly should show up in both the SPINS read and the distributor's own sell-through report. SPINS spikes and the distributor stays flat? The spike is probably a panel artifact, not the business.

Third, call your SPINS rep. This is not an admission of weakness, it is the job. A rep can pull the data-delivery log for your account and tell you straight whether the odd week had a known panel event behind it: a retailer gap, a backfill, a coefficient change. Ten minutes on the phone beats a category director questioning your methodology in front of the room.

SPINS projection methodology: coefficient changes and retroactive adjustments

Here is a source of SPINS volatility almost nobody documents: the retroactive panel change. When a retailer joins or leaves the panel, or a major distributor renegotiates its data-sharing deal, SPINS will sometimes go back and revise historical data to account for the new panel composition.

So the 52-week report you pulled in January can show slightly different numbers than the same 52-week report pulled in April, if a panel event happened in between. The data is not wrong. It is, if anything, more accurate. But if you have been tracking brand performance against a fixed baseline, the baseline just moved under you.

The fix is unglamorous and effective. Lock down the historical baseline report at the start of each reporting cycle, say the fiscal year start, and stash it somewhere separate. Use the locked copy for year-over-year comparisons and the current version for current-period reads. When the two refuse to reconcile, ask your rep whether a panel composition change is the culprit. It usually is.

Doing this in Scout

The standard SPINS portal hands you the raw numbers and stops there. It does nothing to separate suppression from projection from real movement. A cell that should set off alarms looks exactly like a cell that should not.

Scout dashboards sit on top of your SPINS extracts and give you a place to actually run the three diagnostic questions: ACV, dollars, and units on one time series, cross-retailer comparisons in view, and a shared dashboard that sales and category leadership can read together instead of emailing PDFs back and forth. To be clear about what Scout does not do: it will not auto-flag a suppressed cell or auto-detect a backfill for you today. That discipline still belongs to the analyst. What Scout does is make the work compound. Figure out a Week-18 backfill once, and the annotation lives on the dashboard, not in a pivot table on somebody's laptop that nobody else will ever find.

Summary + further reading

A few things to carry out of this. SPINS data is projected from a sample, never measured directly, and the projection coefficients shift as the panel changes shape. Suppressed cells are not zeros, and reading them as zeros drags distribution, share, and lift downward. For any anomalous cell, run the three questions: did ACV move with sales, is the neighboring week unusually high, is the retailer direct-scan or distributor-flow. Before a suspicious number goes in a deck, validate it against implied velocity, a parallel distributor signal, and a quick call to your rep. And remember that historical data can be revised retroactively, so lock down your baseline reports when each cycle starts.

Related reading: What is SPINS data? · Syndicated vs. panel data