Grocery Market Navigator — Executive Summary

Supply Chain Outlook

Sales
The next six months are expected to be softer than the previous 12 months. The Sales TTM YOY% was 1.9%, but the next six months are expected to be only 0.7%.

Units
The sales decline is driven primarily by a drop in expected units from -0.4% in the previous 12 months to -1.6% in the next six months.

CPI FAH
Food inflation (CPI FAH) is expected to increase in the next six months by 2.2% which is lower than the previous forecast of 2.7%. The decline is almost entirely due to the decline in PPI Food Manufacture.

Farm and Food Manufacturing Prices
Farm and Food Manufacturing are the most exposed to energy prices and both will experience the largest decrease in prices. Farm prices were forecasted to increase to 6.9% in the next six months, but the revised energy and import prices pushes this forecast down to a much more manageable 3.2% YOY. The Food Manufacturing forecast came down from 3.6% to 2.5% YOY and is why CPI FAH dropped about 0.5 ppts.

Next Six Months
With lower oil prices, supply chain price growth will slow across the board and help unit and sales growth, which were both declining ahead of the conflict. The last couple of months have seen US sales YOY rebound to 1.8% and 2.3% while units lagged at -0.2% and -0.7%. The model is pointing to more downward pressure on both units and sales due to soft Real Disposable Income and Housing Prices. The long run trend for grocery sales is ~3% YOY and the forecast is coming in at 0.7% based on a combination of 2.2% CPI FAH and -1.5% units YOY. The model is likely missing to the low side but also confident that we will be lucky to hit 2% YOY growth in 2026.

Why this agreement will hold
If the conflict would have gone on much longer, things would have turned pretty ugly. Oil reserves around the world were getting precariously low, and with that buffer evaporating, oil prices would have likely moved north of $125/barrel. In addition, food manufacturers and grocery retailers can only compress margins for several months, and their reserves were also being tested. Then you role in wage pressure and things get really ugly. And all of this would not be ideal in a mid term election year. The administration was motivated to get a deal done and this is also why the agreement to hold.

* Adjustments to the Model
Unfortunately, we cannot do an apples-to-apples comparison for units and sales with last month because mass and warehouse sales and units were added into this month's model. I was holding off because you get a cleaner read on grocery items from supermarkets, but in the last few months mass and warehouse started to break away from supermarkets, so even though we get some non-grocery items into the mix, it was time. The supply chain data and models, however, did not change.

Causal Chain (TTM | 6-Mo Fcst)

↑

↑

↑

PPI Food Manufacturing

↑

OUTLOOK

PPI Farm Products – 6 Month YOY Outlook

+2.8% | +3.2%

ttm | 6-mo fcst

Before the conflict, farm product prices trended sharply lower peaking at 15.4% YOY in Feb-2025 and bottoming at -7.5% YOY in Jan-2026. As expected farm product prices have started to rebound and hit 4.4% YOY growth in Apr-2026. The pre-conflict deflation was driven by declining oil prices and falling food import prices but both of are now reversing.

Over the six-month forecast window, PPI Farm Products is projected to average +3.1% year-over-year, a modest step up from the trailing twelve-month average of +2.8%, but that headline comparison understates the true pricing pressure now building at the farm gate. Oil prices are the dominant driver by a wide margin, with crude averaging a 35.4% gain over the input window and contributing +4.50 percentage points to the forecast — a direct transmission from oil into fertilizer, diesel, and agrochemical costs that farmers cannot avoid. Natural gas prices are working in the opposite direction, averaging down 10.0% and subtracting 0.74 percentage points, providing a partial offset through lower energy inputs for irrigation and drying. Critically, the relatively moderate-looking YoY readings — particularly the dip to +1.4% in July 2026 — reflect a base effect distortion: PPI Farm Products swung from a peak of +8.5% in October 2025 to a trough of -4.7% in January 2026, and that sharp prior deflation created a low comparison base that is currently cushioning the headline numbers; absent that deflationary episode, the underlying price rebound driven by oil would be registering as significantly more alarming in year-over-year terms than the forecast figures suggest.

Forecast Distribution Simulation

This distribution is built on 1,000 simulations using holdout errors from the rolling validation period. Small differences between the point forecast and the simulated median may occur as a result.

The median forecast is +4.8% with a forecast distribution of +/- 5.8%. We would expect Ppi Farm Products to fall between -0.2% and +11.4% over the next six months.

INPUTS

PPI Farm Products – Inputs

Forecast Decomposition — 6-Month Average Contribution

Variable	Avg Input (YoY%)	Contribution (ppts)	% of Total
Oil Prices	+35.4%	▲ 4.50 ppts	+81%
Natural Gas Prices	-10.0%	▼ 0.74 ppts	-13%
Food Import Prices	-0.5%	▼ 0.32 ppts	-6%
Structural Offset (model constant)	+1.0%	▼ 0.27 ppts
Total Forecast		▲ 3.17 ppts	100%

Oil Prices (Lead 1)

Oil prices are forecast to average +35.4% year-over-year over the next six months, a trajectory that points sharply higher and positions crude as the single most consequential variable in the farm input cost outlook. That surge alone is expected to contribute +4.50 percentage points to PPI Farm Products movement, accounting for 81% of total input cost pressure across the forecast period.

Natural Gas Prices (Lead 2)

Natural gas prices are forecast to average -10.0% year-over-year over the next six months, a declining trajectory that will continue to pull farm input costs lower. This drag accounts for a contribution of -0.74 percentage points to the overall input cost forecast, representing 13% of total input movement — a meaningful but secondary force behind oil, which remains the dominant driver in this cycle.

Food Import Prices

Food import prices are expected to average -0.5% year-over-year over the next six months, a modest deflationary reading that is nonetheless trending toward less negative territory as the forecast period progresses. This drag subtracts an estimated 0.32 percentage points from the PPI Farm Products forecast, accounting for roughly 6% of total input movement — a relatively contained influence compared to oil, which remains the dominant force driving farm cost pressures through this cycle.

MODEL PERFORMANCE

PPI Farm Products – Model Performance

24-Month Holdout

Over the 24-month holdout period, the model produced a mean absolute error of 6.43 percentage points, meaning that on average the forecast landed within roughly 6 and a half points of the actual PPI Farm Products reading. For a index that can swing 20 or more points in a volatile year, that is a workable baseline for directional planning, though it does signal that precise point estimates should be treated with appropriate caution.

Holdout Errors Over Time

Rolling errors across the holdout ranged from 3.0 to 11.1 percentage points, a spread that reflects how sensitive this model is to the pace of change in its inputs rather than any structural flaw. Errors peaked around January 2025, which aligns with a period when oil prices were moving quickly — the model's one-month lead on oil and two-month lead on natural gas mean that sharp, fast-moving swings in energy markets can outrun what the lagged structure captures in real time.

OLS Regression Results
R-squared: 0.668			Adj. R-sq: 0.658
F-statistic: 64.42			Prob(F): 6.58e-23
No. Observations: 100			Df Residuals: 96
Variable	Coef	Std Err	t	P>\|t\|	[0.025, 0.975]
const	-0.2579	0.999	-0.258	0.797	[-2.240, 1.724]
oil_prices_lead1_yoy	0.0960	0.025	3.805	0.000	[0.046, 0.146]
nat_gas_prices_lead2_yoy	0.1092	0.021	5.188	0.000	[0.067, 0.151]
import_index_yoy	0.6391	0.262	2.440	0.017	[0.119, 1.159]

Omnibus: 0.125		Prob(Omnibus): 0.940		Durbin-Watson: 0.410
Skew: -0.000		Kurtosis: 3.015

Model Performance

The model's R-squared of 0.668 means it explains roughly two-thirds of the historical variation in PPI Farm Products, a reasonable fit for an agricultural index influenced by weather, trade policy, and other factors that no energy-input model can fully anticipate. All three inputs are statistically significant, with t-statistics of 3.8 for oil, 5.2 for natural gas, and 2.4 for the import price index — each well above the 2.0 threshold that signals a reliable relationship. The Omnibus probability of 0.940, comfortably above 0.05, confirms that the model's residuals are normally distributed, meaning errors are not systematically skewed in one direction. The Durbin-Watson statistic of 0.41, however, sits well below the ideal value near 2.0, indicating meaningful autocorrelation in the residuals — consecutive forecast errors tend to run in the same direction, which suggests the model may be slow to adjust when the index enters a sustained trend up or down.

OUTLOOK

PPI Food Manufacturing – 6 Month YOY Outlook

+3.1% | +2.5%

ttm | 6-mo fcst

Prior to the conflict, food manufacturing prices followed farm products lower. PPI Food Manufacturing peaked at 4.9% YOY in Oct-2025 and dropped to 0.8% by Feb-2026, driven by falling farm input costs, declining oil prices, softer manufacturing wages, and weak food import prices. But food manufacturing pricesare reversing course as expected but not as significantly as farm products.

Over the next six months, PPI Food Manufacturing is forecast to average +2.5% year-over-year, a step down from the +3.1% recorded over the trailing twelve months, though that comparison obscures a more important story about underlying pricing pressure. Manufacturing wages are the dominant force, contributing +3.94 percentage points on an average input gain of +4.2%, and oil prices add a further +0.72 points on a +25.0% average rise, directly elevating energy-intensive processing and transport costs embedded throughout the food manufacturing supply chain. PPI Farm Products contribute an additional +0.56 points, while Food Import Prices, falling an average of 2.3%, subtract 0.47 points, providing a modest counterweight. Critically, the +2.5% forecast average is being held down by a base-effect distortion: PPI Food Manufacturing peaked at +4.9% year-over-year in September 2025 before collapsing to +0.9% by January 2026, and it is that depressed trough which is acting as the comparison base for the early months of the forecast window — meaning the muted readings in May and June (+1.6% and +1.0%) reflect arithmetic against a low base rather than any genuine easing of cost pressure, and as the window rolls past that trough the series accelerates sharply to +3.9% by September before pulling back, a trajectory that is far more representative of the actual pricing environment manufacturers are navigating.

Forecast Distribution Simulation

This distribution is built on 1,000 simulations using holdout errors from the rolling validation period. Small differences between the point forecast and the simulated median may occur as a result.

The median forecast is +2.4% with a forecast distribution of +/- 1.2%. We would expect Ppi Food Mfg to fall between +1.2% and +3.6% over the next six months.

INPUTS

PPI Food Manufacturing – Inputs

Forecast Decomposition — 6-Month Average Contribution

Variable	Avg Input (YoY%)	Contribution (ppts)	% of Total
Manufacturing Wages	+4.2%	▲ 3.94 ppts	+69%
Oil Prices	+25.0%	▲ 0.72 ppts	+13%
PPI Farm Products	+3.0%	▲ 0.56 ppts	+10%
Food Import Prices	-2.3%	▼ 0.47 ppts	-8%
Structural Offset (model constant)	+1.0%	▼ 2.29 ppts
Total Forecast		▲ 2.46 ppts	100%

PPI Farm Products (Lead 1)

PPI Farm Products is forecast to average +3.0% year-over-year over the next six months and is trending higher, feeding directly into food manufacturing cost pressures. That average contributes +0.56 percentage points to the PPI Food Manufacturing forecast, accounting for 10% of total input movement — though because this rebound is being measured against a depressed trough, the true pricing pressure at the farm level is meaningfully stronger than the year-over-year figure suggests.

Food Import Prices (Lead 3)

Food import prices are averaging -2.3% year-over-year across the forecast period, a deflationary read that is nonetheless moving in a less favorable direction as the period progresses. This drag subtracts 0.47 percentage points from the PPI Food Manufacturing forecast, accounting for 8% of total input movement — though executives should note that this figure understates true pricing pressure, since it is measured against the depressed trough of a V-shaped cycle in which PPI Food Manufacturing itself bottomed at 0.9% year-over-year in January 2026, meaning the low comparison base is artificially cushioning what would otherwise appear as a significantly steeper rebound.

Manufacturing Wages (Lead 1)

Manufacturing wages are forecast to average +4.2% year-over-year over the next six months, a rate that is heading higher and that, given the V-shape in PPI Food Manufacturing — which bottomed at 0.9% in January 2026 after peaking at 4.9% in September 2025 — is almost certainly understated by the low comparison base left behind by that trough. Wages contribute +3.94 percentage points to the PPI Food Manufacturing forecast, representing 69% of total input movement and making labor the single dominant cost pressure grocery executives need to track in this cycle.

Oil Prices (Lead 4)

Crude oil prices are forecast to average +25.0% year-over-year over the next six months, a level that continues to climb and is transmitting directly into food manufacturing input costs across the supply chain. That increase contributes +0.72 percentage points to the PPI Food Manufacturing forecast, accounting for 13% of total input movement — and because PPI Food Manufacturing itself bottomed at just 0.9% year-over-year in January 2026 before rebounding, that depressed base is acting as a cushion on current readings, meaning the true pricing pressure building in food manufacturing is meaningfully stronger than the headline year-over-year numbers suggest.

MODEL PERFORMANCE

PPI Food Manufacturing – Model Performance

24-Month Holdout

Over the 24-month holdout period, the model produced a mean absolute error of 1.61 percentage points, meaning the average forecast deviated from actual PPI Food Manufacturing readings by that margin. For a series that has swung from deflation into positive territory in recent years, that level of precision is operationally useful — executives can treat the forecast as directionally reliable with a roughly plus-or-minus 1.6-point band around the central estimate.

Holdout Errors Over Time

Rolling 12-month errors ranged from 0.5 to 2.7 percentage points across the holdout window, with the largest errors clustering around July 2024. That peak aligns with a period when oil prices and farm input costs were moving quickly in both level and direction — exactly the conditions that stress any regression-based model, since large or fast-moving inputs compress the lag structure and can outrun what historical coefficients anticipate. The tighter errors at either end of the window reflect more stable input environments where the model's relationships held cleanly.

OLS Regression Results
R-squared: 0.900			Adj. R-sq: 0.894
F-statistic: 168.38			Prob(F): 2.65e-45
No. Observations: 100			Df Residuals: 94
Variable	Coef	Std Err	t	P>\|t\|	[0.025, 0.975]
const	-2.8487	0.610	-4.673	0.000	[-4.059, -1.638]
ppi_farm_products_lead1_yoy	0.1790	0.016	11.428	0.000	[0.148, 0.210]
import_index_lead3_yoy	0.2582	0.062	4.194	0.000	[0.136, 0.380]
mfg_wages_lead1_yoy	1.0926	0.174	6.276	0.000	[0.747, 1.438]
oil_prices_lead4_yoy	0.0267	0.005	5.513	0.000	[0.017, 0.036]
covid	2.3717	0.623	3.807	0.000	[1.135, 3.609]

Omnibus: 4.071		Prob(Omnibus): 0.131		Durbin-Watson: 1.390
Skew: 0.438		Kurtosis: 3.260

Model Performance

The model's R-squared of 0.90 means it accounts for 90 percent of the historical variation in PPI Food Manufacturing, a strong fit for a macroeconomic price series with this many moving parts. Every input — farm products, import prices, manufacturing wages, oil prices, and the Covid indicator — carries a t-statistic well above 2.0, confirming that each one is contributing genuine explanatory power rather than noise. The Omnibus probability of 0.131 sits above the 0.05 threshold, indicating the residuals are distributed normally and the model is not systematically misreading any portion of the data. The Durbin-Watson statistic of 1.39 is somewhat below the ideal of 2.0, suggesting a modest degree of positive autocorrelation in the residuals — worth monitoring, though not severe enough to invalidate the forecasts.

OUTLOOK

CPI Food-at-Home – 6 Month YOY Outlook

+2.4% | +2.2%

ttm | 6-mo fcst

Food inflation was also softening ahead of the conflict peaking at 2.7% YOY in Aug-2025, and hitting 2.0% in Feb-2026. That pre-conflict decline was driven by falling farm and food manufacturing costs working through the supply chain. The other key drivers — PPI Grocery Retail, stable around 3.0% YOY, and retail wages, trending up since Jan-2025 — were providing some upward momentum. The result was a slow, measured decline heading into the conflict. But we are starting to see, the recent higher farm and food manufacturing prices are starting to push food inflation higher.

Over the six-month forecast window, CPI Food-at-Home is projected to average +2.2% year-over-year, a modest step down from the trailing twelve-month average of +2.4%, with the path running from +1.8% in May to a peak of +2.5% in August before easing back to +2.3% in October. The single largest contributor is retail wage growth, which at an average of +3.7% adds approximately 1.71 percentage points to the forecast — a reminder that labor, not just commodity inputs, is a persistent structural cost embedded in grocery pricing. PPI Grocery Retail and PPI Food Manufacturing add a further 1.39 and 0.88 percentage points respectively, with the latter reflecting the upstream pressure that has built through farm inputs and oil-driven production costs now working their way through the food manufacturing supply chain. Critically, a structural offset of -1.83 percentage points is suppressing the headline number, capturing the real-world counterweights that the analyst guidance makes explicit: slowing retail wage momentum and margin compression by grocery retailers absorbing upstream cost pressure rather than passing it fully to shelf prices. The result is a forecast that looks stable on the surface but masks a supply chain that is still pushing hard — the relative calm in the YoY readings owes more to retailer margin sacrifice and moderating labor cost escalation than to any genuine easing of input cost pressure.

Forecast Distribution Simulation

This distribution is built on 1,000 simulations using holdout errors from the rolling validation period. Small differences between the point forecast and the simulated median may occur as a result.

The median forecast is +2.3% with a forecast distribution of +/- 0.7%. We would expect Cpi Fah to fall between +1.7% and +3.0% over the next six months.

INPUTS

CPI Food-at-Home – Inputs

Forecast Decomposition — 6-Month Average Contribution

Variable	Avg Input (YoY%)	Contribution (ppts)	% of Total
Retail Wages	+3.7%	▲ 1.71 ppts	+43%
PPI Grocery Retail	+3.5%	▲ 1.39 ppts	+35%
PPI Food Manufacturing	+2.7%	▲ 0.88 ppts	+22%
Structural Offset (model constant)	+1.0%	▼ 1.83 ppts
Total Forecast		▲ 2.15 ppts	100%

PPI Food Manufacturing (Lead 1)

PPI Food Manufacturing is forecast to average +2.7% year-over-year over the next six months, with the index expected to edge higher as upstream oil-driven cost pressures continue working through the production chain. That increase contributes an estimated +0.88 percentage points to the CPI Food-at-Home forecast, accounting for 22% of total measured input movement — though its pass-through to shelf prices is being partially absorbed by slowing retail wage growth and some margin compression among grocery retailers.

PPI Grocery Retail

PPI Grocery Retail is forecast to average +3.5% year-over-year over the next six months, with the index continuing to drift higher as upstream cost pressures from oil-driven farm and manufacturing inputs increasingly outpace the offsetting drag from slowing retail wage growth and margin compression. This measure contributes +1.39 percentage points to the CPI Food-at-Home forecast, representing 35% of total input movement and making it the single most consequential transmission channel between the supply chain and what consumers ultimately pay on the shelf.

Retail Wages (Lead 3)

Retail wage growth is expected to average +3.7% year-over-year over the next six months, holding broadly stable and continuing to exert upward pressure on food-at-home prices. That wage growth contributes an estimated +1.71 percentage points to the forecast — representing 43% of total input movement — making it the single largest driver of CPI food-at-home inflation in the period ahead.

MODEL PERFORMANCE

CPI Food-at-Home – Model Performance

24-Month Holdout

Over the 24-month holdout period, the model produced a mean absolute error of 0.84 percentage points, meaning the average forecast deviated from actual CPI Food-at-Home by less than one percentage point. For grocery executives tracking inflation to within a band that informs pricing and margin decisions, that level of precision is operationally useful, though it is not tight enough to treat any single monthly forecast as exact.

Holdout Errors Over Time

Rolling 12-month errors ranged from 0.4 to 1.0 percentage points across the holdout window, with the widest errors occurring around January 2025. That timing aligns with what this model is most exposed to: when oil prices and food manufacturing costs move sharply or reverse direction quickly, the lagged input structure can temporarily fall behind the data, widening the gap between forecast and outcome before it corrects.

OLS Regression Results
R-squared: 0.890			Adj. R-sq: 0.885
F-statistic: 191.26			Prob(F): 1.54e-44
No. Observations: 100			Df Residuals: 95
Variable	Coef	Std Err	t	P>\|t\|	[0.025, 0.975]
const	-2.3594	0.383	-6.159	0.000	[-3.120, -1.599]
ppi_food_mfg_lead1_yoy	0.3581	0.031	11.643	0.000	[0.297, 0.419]
ppi_grocery_yoy	0.5163	0.031	16.556	0.000	[0.454, 0.578]
retail_wages_lead3_yoy	0.3804	0.104	3.647	0.000	[0.173, 0.587]
covid	1.2763	0.521	2.449	0.016	[0.242, 2.311]

Omnibus: 0.640		Prob(Omnibus): 0.726		Durbin-Watson: 1.062
Skew: 0.169		Kurtosis: 2.740

Model Performance

The model's R-squared of 0.89 means it accounts for 89 percent of the historical variation in CPI Food-at-Home, which is a strong explanatory fit for a macroeconomic price series with this many moving parts. Every input in the model — food manufacturing PPI, grocery-specific PPI, retail wages, and the COVID indicator — carries a t-statistic well above 2.0, confirming that each variable is making a statistically meaningful contribution rather than adding noise. The Omnibus probability of 0.726 sits comfortably above 0.05, indicating that the model's residuals are normally distributed and that no systematic pattern is being left unexplained in the errors. The Durbin-Watson statistic of 1.062 is below the ideal of 2.0, suggesting some positive serial correlation in the residuals, which means consecutive forecast errors tend to lean in the same direction and is worth monitoring as new data arrives.

OUTLOOK

US Grocery Units – 6 Month YOY Outlook

-0.4% | -1.5%

ttm | 6-mo fcst

US Grocery Market Units have been trending down YOY since early 2025 and bottomed out at -3.0% just ahead of the conflict.The last 12 months were down -0.5%, but the first half of the year was much stronger than the last half. The higher CPI FAH number in Apr-2026 helped to keep unit growth down about 1.4% YOY.

Grocery unit volume is forecast to average -1.4% year-over-year over the next six months, a meaningful deterioration from the trailing twelve-month average of -0.4%, with conditions worsening progressively through the period before a brief partial recovery in September and then renewed pressure in October. The dominant force pulling units lower is food-at-home inflation, which is expected to average 2.1% and subtract 1.69 percentage points from unit growth — a direct consequence of elevated oil prices feeding through farm input costs and food manufacturing expenses into shelf prices that consumers are increasingly unwilling or unable to absorb in full. Real disposable income, averaging -0.6% over the period, adds a further drag of 0.24 percentage points, meaning households are simultaneously facing higher prices and shrinking purchasing power, a combination that reliably compresses transaction volumes. Home prices, rising at an average of 0.6%, contribute a modest offset of 0.24 percentage points, consistent with the wealth-effect channel that tends to support grocery spending among homeowning households even in inflationary environments, though this cushion is clearly insufficient to counteract the pricing and income headwinds. Executives should note that the -1.4% forecast average still represents a significant step down from the roughly 2% unit growth seen in late 2024 and early 2025, and sits well above the cycle trough of -3% registered in February 2026, suggesting the market is navigating a prolonged contraction rather than a brief correction.

Forecast Distribution Simulation

This distribution is built on 1,000 simulations using holdout errors from the rolling validation period. Small differences between the point forecast and the simulated median may occur as a result.

The median forecast is -1.4% with a forecast distribution of +/- 0.4%. We would expect Units Mkt Trend to fall between -1.7% and -0.9% over the next six months.

INPUTS

US Grocery Units – Inputs

Forecast Decomposition — 6-Month Average Contribution

Variable	Avg Input (YoY%)	Contribution (ppts)	% of Total
CPI Food-at-Home	+2.1%	▼ 1.69 ppts	-78%
Real Disposable Income	-0.6%	▼ 0.24 ppts	-11%
Home Prices	+0.6%	▲ 0.24 ppts	+11%
Total Forecast		▼ 1.69 ppts	100%

CPI Food-at-Home

CPI Food-at-Home is forecast to average +2.1% year-over-year over the next six months, a rate that is headed higher as oil-driven input costs continue to push through to shelf prices. That acceleration is the single largest drag on grocery unit volume in this forecast, subtracting 1.69 percentage points from unit growth and accounting for 78% of the total input movement pulling units lower.

Real Disposable Income

Real disposable income is forecast to average -0.6% year-over-year over the next six months, a trajectory that points to continued deterioration in household purchasing power. This drag subtracts 0.24 percentage points from the grocery unit forecast, accounting for 11% of total input movement across the model.

Home Prices

Home price growth is expected to average +0.6% year-over-year over the next six months, a rate that is gradually moderating from its recent pace. This modest appreciation contributes +0.24 percentage points to grocery unit volume growth, accounting for 11% of total input movement in the forecast.

MODEL PERFORMANCE

US Grocery Units – Model Performance

24-Month Holdout

The 24-month holdout mean absolute error is 0.01 percentage points, meaning that on average across the holdout period, the model's forecasts for US grocery unit growth differed from actual outcomes by just 0.01 percentage points. That is a remarkably tight result for a macro-driven model operating over a two-year window. It gives executives a high degree of confidence that the model's forward projections are tracking the real dynamics of consumer purchasing behavior rather than fitting noise.

Holdout Errors Over Time

Rolling forecast errors across the holdout period ranged from 0.3 to 0.8 percentage points, with errors peaking around November 2025. That timing is consistent with what this model is built to capture: when oil prices are moving sharply or reversing direction quickly, the upstream input chain transmits those shocks through farm costs and food manufacturing with a lag that can temporarily outpace the model's signals. Fast-moving inputs create the widest forecast gaps, and the late-2025 spike in errors almost certainly reflects a period when one or more key drivers shifted faster than the trailing data could fully encode.

OLS Regression Results
R-squared: 0.881			Adj. R-sq: 0.876
F-statistic: 175.90			Prob(F): 5.17e-43
No. Observations: 100			Df Residuals: 95
Variable	Coef	Std Err	t	P>\|t\|	[0.025, 0.975]
const	22.9287	1.110	20.650	0.000	[20.724, 25.133]
cpi_fah_log	-0.8002	0.070	-11.459	0.000	[-0.939, -0.662]
rdi_log	0.3761	0.126	2.994	0.004	[0.127, 0.625]
home_price_log	0.4013	0.068	5.944	0.000	[0.267, 0.535]
covid	0.0368	0.007	5.283	0.000	[0.023, 0.051]

Omnibus: 117.344		Prob(Omnibus): 0.000		Durbin-Watson: 2.118
Skew: 3.708		Kurtosis: 29.965

Model Performance

The model's R-squared of 0.881 means it explains 88.1 percent of the historical variation in US grocery unit growth, which is a strong result for a consumer demand model relying on macroeconomic inputs. All four independent variables — food-at-home CPI, real disposable income, home prices, and the COVID indicator — carry t-statistics well above 2.0, confirming that each input is statistically significant and contributing genuine explanatory power rather than coincidental correlation. The Omnibus probability of 0.000 is below 0.05, which indicates the residuals are not normally distributed, a flag worth monitoring as the skew reading of 3.708 suggests the model's errors tilt in one direction during outlier periods. The Durbin-Watson statistic of 2.118 sits close to the ideal value of 2.0, indicating no meaningful autocorrelation in the residuals and that the model is not systematically leaning on recent errors to generate its next forecast.

OUTLOOK

US Grocery Sales – 6 Month YOY Outlook

+1.9% | +0.7%

ttm | 6-mo fcst

Sales were solid in 2024 and the first eight months of 2025, but the last six months have seen YOY sales drop precipitously. The long-run trend for grocery is about 3% YOY with 2% growth typically coming from CPI FAH and the other 1% from unit growth. CPI FAH reversed course in Apr-2026 which helped Sales growth tick up to a meager 1.5% YOY in April.

Over the next six months, US grocery sales growth is forecast to average +0.7% year-over-year, a meaningful step down from the trailing twelve-month average of +1.9%, reflecting a gradual deceleration that runs from +1.4% in May 2026 to just +0.2% by October. The dominant driver keeping sales in positive territory is CPI food-at-home inflation, contributing +2.15 percentage points as food prices continue to rise, propelled upstream by elevated oil prices feeding through to farm input and manufacturing costs. Working directly against that is a deterioration in grocery unit volume, which subtracts 1.45 percentage points from the forecast as consumers, squeezed by cumulative price increases, pull back on the number of items they purchase. The net result is a sales trend that remains technically positive but is being held aloft almost entirely by price, not by any genuine strengthening in consumer demand.

Forecast Distribution Simulation

This distribution is built on 1,000 simulations using holdout errors from the rolling validation period. Small differences between the point forecast and the simulated median may occur as a result.

The median forecast is +0.7% with a forecast distribution of +/- 0.9%. We would expect Sales Mkt Trend to fall between -0.1% and +1.6% over the next six months.

INPUTS & PERFORMANCE

US Grocery Sales – Inputs and Performance

Forecast Decomposition — 6-Month Average Contribution

Variable	Avg Input (YoY%)	Contribution (ppts)	% of Total
CPI Food-at-Home	+2.1%	▲ 2.15 ppts	+60%
US Grocery Units	-1.4%	▼ 1.45 ppts	-40%
Total Forecast		▲ 0.70 ppts	100%

US Grocery Units

Grocery unit volumes are expected to average -1.4% year-over-year across the six-month forecast period, with the trend continuing to decline. This contraction subtracts 1.45 percentage points from the sales growth forecast, making it the single largest drag on the outlook and accounting for 40% of total input movement.

CPI Food-at-Home

CPI Food-at-Home is forecast to average +2.1% year-over-year across the next six months, with the index trending modestly higher as oil-driven input costs continue to work through the supply chain into shelf prices. That +2.1% average contributes +2.15 percentage points to grocery sales growth over the forecast period, accounting for 60% of total input movement and making it the single largest driver in the model.

Forecast Errors

US Grocery Sales is a derived model, calculated as the product of US Grocery Units and Average Price (CPI Food-at-Home), and carries no standalone holdout MAE or OLS regression statistics of its own. Its forecast uncertainty is therefore characterized through the distribution of outcomes rather than a single error metric. The forecast distribution implies an uncertainty of approximately 0.9 percentage points, with the 90% confidence interval spanning -0.1% to 1.6%.

The median point forecast for the six-month average year-over-year change in US Grocery Sales stands at 0.7%, but the 90% confidence interval from -0.1% to 1.6% signals meaningful dispersion around that central estimate. In practical terms, executives should read this as: nine times out of ten, realized sales growth will land somewhere in that range, but the gap between the downside and the upside — a flat to slightly negative outcome versus growth approaching 1.6% — is wide enough to matter for budgeting and inventory planning. A 0.7% median is a reasonable working assumption, but the interval warns against treating it as a precise target.

Because US Grocery Sales is derived entirely from its two upstream components, the accuracy of this forecast depends completely on how well the Units model and the CPI Food-at-Home model each perform individually. Any forecast error in either of those inputs will propagate directly into the sales estimate, and errors that move in the same direction across both models can compound. Readers should consult the US Grocery Units and CPI Food-at-Home sections of this newsletter for R-squared values, holdout MAE, and the full set of statistical diagnostics that characterize model reliability.