Cannibalisation Rate: The Working FMCG Number Most TPO Reports Quietly Skip

Walk into any TPO post-event review and you will see the same chart. A bar showing the promoted SKU's volume during the event, towering over its baseline. A "lift" number in big green text. A polite round of nods. The event moves to the "replicate" pile.

What that chart almost never shows is what happened to the rest of the brand portfolio while the spotlight SKU was on promotion. The other six biscuit SKUs the brand sells. The non-promoted pack sizes. The premium variant. The flavour cousins.

Most of them dipped. Some by a little, some by a lot. And every unit they lost is a unit the promoted SKU "won" from your own shelf, not from a competitor and not from a new shopper.

That is cannibalisation. The cannibalisation rate is the working number that puts a percentage on it.

Why this matters more than you think

A senior commercial director reading the event report sees a 2.5x lift on the promoted SKU and signs off on the next event. The category buyer sees the same number and asks for another one. Both decisions are made on the SKU-level lift. Both decisions are biased upward because nobody subtracted the volume drop on the rest of the portfolio.

The honest brand-level uplift, after cannibalisation, is often half of what the SKU-level report claims. Sometimes less. And the trade dollars that funded the event were paid against the inflated number, not the honest one.

How this complements Van Heerde's thirds rule

The academic decomposition you meet in the Van Heerde thirds-rule concept splits promo volume into three roughly equal buckets: subsidised base (volume you would have sold anyway), brand switching (volume pulled from competitors), and category expansion plus stockpiling (new demand and pulled-forward purchases).

That decomposition is rigorous, peer-reviewed, and the right academic frame. But there is a problem when working teams try to use it on a Tuesday morning post-event review. The "brand switching" bucket bundles competitive switching together with own-brand switching, and from the manufacturer's P&L perspective those two are very different animals. Competitive switching grows your share. Own-brand switching just moves units between your own SKUs at trade cost.

The cannibalisation rate is the working metric teams actually compute to separate those two. Run it in parallel to the Van Heerde split. The Van Heerde number tells you the academic mix; the cannibalisation rate tells you how much of the "switching" was an own goal.

The arithmetic is simple. The discipline is in defining the comparison window and the comparison set correctly.

The headline formula

Cannibalisation Rate = (Volume drop on non-promoted own SKUs) / (Volume gain on promoted SKU)

Both numbers are measured against the same baseline window, in the same store set, over the same calendar weeks the promotion ran.

What goes in the numerator

Every SKU in the brand portfolio that was NOT on promotion during the event window. Sum the baseline-versus-actual gap for each of them across the event period. The number is usually negative because the rest of the portfolio dips when one SKU is shouting from the shelf.

If a sister SKU happens to be on its own promotion at the same time, exclude it from the numerator. You cannot cleanly attribute its movement to the SKU you are studying.

What goes in the denominator

The gross volume gain on the promoted SKU, baseline versus actual, over the same window. Use the controlled-store baseline, not last year's sales. Last year is contaminated by whatever happened in the prior calendar.

A worked example

A premium chocolate biscuit at $4.99 runs a 25 percent off TPR for 2 weeks across 200 stores.

Baseline volume on the promoted SKU was 4,000 units across the window. Promoted volume was 7,400 units. Gross gain: +3,400 units, or +85 percent on the SKU's own line.

The brand sells five other biscuit SKUs in the same stores. Their combined baseline across the same 2 weeks was 12,000 units. Their actual was 9,360 units. Drop: 2,640 units, or 22 percent below baseline.

Cannibalisation Rate = 2,640 / 3,400 = 78 percent

That promotion does not look like a promotion. It looks like the brand paid trade money to move 2,640 units sideways from its other SKUs into the discounted SKU, and only genuinely added 760 units of brand-level volume.

The brand-level uplift is 760 units, not 3,400. That is +18 percent on the brand portfolio, not +85 percent on the SKU. The SKU-level event report would have shown the +85 percent. Most do.

Reading the result

Practitioner thresholds give you a colour code for the post-event report:

• Below 20 percent: healthy. The promoted SKU is genuinely pulling new volume. Replicate the mechanic.
• 20 to 35 percent: typical. Most FMCG events sit here. Worth running but worth interrogating, especially if the promoted SKU and a sister SKU sit at similar price points.
• 35 to 50 percent: concerning. A meaningful share of the lift is sideways movement. Question the choice of promoted SKU.
• Above 50 percent: the brand is paying trade dollars to shuffle volume between its own packs. Stop or redesign.

A premium juice brand at $5.49 runs the same SKU through two different promotional designs in two consecutive quarters at the same retailer. Same shelf, same stores, same time of year. Only the design changes.

Q1 design: 30 percent off TPR on the premium SKU, no display

• Promoted SKU baseline volume: 5,000 units
• Promoted SKU actual volume: 9,250 units
• Gross gain: +4,250 units (+85 percent on the SKU)
• Other 4 SKUs in the brand portfolio: combined baseline 18,000 units, actual 14,400 units
• Combined dip: 3,600 units (minus 20 percent on the rest of the portfolio)
• Cannibalisation Rate: 3,600 / 4,250 = 85 percent
• Brand-level net uplift: +650 units, or +3 percent on the portfolio

The post-event SKU report showed +85 percent and was filed under "replicate". The honest brand-level economics were close to flat. The trade money funding the event delivered roughly nothing in incremental brand volume.

Q2 design: 15 percent off plus end-cap display on the entry-tier SKU

• Promoted SKU baseline volume: 4,000 units (entry tier sells less than premium at base)
• Promoted SKU actual volume: 7,200 units
• Gross gain: +3,200 units (+80 percent on the SKU)
• Other 4 SKUs in the brand portfolio: combined baseline 19,000 units, actual 18,050 units
• Combined dip: 950 units (minus 5 percent on the rest of the portfolio)
• Cannibalisation Rate: 950 / 3,200 = 30 percent
• Brand-level net uplift: +2,250 units, or +10 percent on the portfolio

The SKU-level lift was almost identical (+80 vs +85 percent). The brand-level economics were three times better.

Reading the comparison

Three things drove the swing.

First, the entry-tier SKU has fewer close substitutes inside the portfolio. Shoppers buying the entry-tier on promotion are mostly cost-conscious shoppers from outside the brand or shoppers expanding their consumption. They are not premium-tier loyalists trading down for the week.

Second, the display intercepted shoppers who were not in the juice aisle to start with, which is genuinely category-expansive volume rather than within-brand switching.

Third, the shallower depth meant the discount was not deep enough to pull premium-tier loyalists across to the entry-tier purely on price, which limited the cross-tier cannibalisation.

The Q1 design looked great on the SKU report and was an own goal at brand level. The Q2 design looked similar on the SKU report and was three times better at brand level. The cannibalisation rate is the metric that surfaces the difference. Without it, the team would have replicated Q1.

Three things conspire to keep the cannibalisation rate out of post-event reports across most TPO teams.

One: the data is hard to assemble

The SKU-level lift comes free from the syndicated panel. The brand-level lift after cannibalisation requires you to pull every sister SKU's actual versus baseline for the same window, sum them, and net the dip against the gain. That is a 30 minute analyst task per event. Multiply by 200 events a year and the cost is real.

Most teams skip it. The lift number is on the dashboard by Friday. The honest cannibalised number never gets built.

Two: incentives are misaligned

Category buyers and account managers are graded on shipments, not brand-level incrementality. A buyer who delivers a 2.5x lift on a promoted SKU hits a target. A buyer who reports the same event as +18 percent brand net does not. Guess which framing wins inside a quarterly review.

Three: nobody wants to be the bearer of bad news

Once you start reporting cannibalisation rates, a meaningful share of the calendar that everyone has been congratulating each other on suddenly looks underwater. Renewal conversations get harder. Trade negotiations get harder. The first team to put cannibalisation in their scorecard takes the political hit while everyone else still claims the inflated SKU-level numbers.

How to design promotions that minimise cannibalisation

You do not always pick the promoted SKU. Buyers do, retailers do, joint business plans do. But where you have a vote, three principles reduce cannibalisation rate without giving up volume.

First, promote the SKU with the fewest close substitutes in your own portfolio. A unique flavour or pack size cannibalises less than the middle SKU in a tiered range, because there is less for shoppers to switch from.

Second, promote at the entry tier, not the premium tier. Premium SKUs cannibalise heavily from the rest of the brand because brand-loyal shoppers trade up temporarily during the event and then trade back. Entry-tier promotions tend to win shoppers from competitors or from category exit.

Third, avoid promoting two own SKUs into the same window. If two of your SKUs are on promotion in the same store at the same time, they cannibalise each other in addition to the rest of the portfolio. The SKU-level lift on each looks fine. The combined brand-level economics are usually awful.