Stop leaking margin to your market.

Outprice watches competitor pressure, models demand, and recommends price moves that protect gross profit. Every decision is reviewed before it reaches Shopify.

Get a pricing audit Watch a decision
Recommendation-only first No autopilot. No match-lowest. Built for Shopify merchants first
BSH-7016R · 04 / 28 / 2026 14:22
Brake shoes, rear, 1-ton diesel · OE-grade
Pending review
Current
$148.50
Recommended
$156.20
+5.2% GP
Cost floor
$112.40
38.5% margin guard
Competitor range (n=11)
$144.99–$172.00
3 candidates · floor / anchor / ceiling
Demand (vs 14d baseline)
1.18×
trailing, EWMA-smoothed
Confidence
92%
log-log OLS, 312 obs.
Action
Live decision · cycling

02 Audit

Find out where your catalog is leaking margin.

Send your store. We'll return a 5-page pricing opportunity audit. Competitor pressure across representative SKUs, what looks underpriced, what's risky to hold, where margin floors are at risk, and whether your catalog is actually a fit. No pitch attached.

Get a pricing audit

5 working days. Free for qualified merchants. We'll tell you if it's not a fit.

Pricing audit
Sample · {your_store}
5 pages
PDF
  • 01Competitor pressurerepresentative SKUs, range, EWMA-smoothed anchor
  • 02Underpriced SKUsvs cheapest authorized retailer, with $-impact estimate
  • 03Risky SKUs to holdwhere the market signal is thin or noisy
  • 04Cost-floor / margin concernswhere price + COGS leave no real room
  • 05Catalog fit assessmenthonest read on whether Outprice helps you
  • 06Recommendationsraise, hold, review, or exclude (per SKU)
If the catalog isn't a fit, we'll say so. No obligation.

03 Pricing decision flow

How a recommendation becomes a price.

One scan. One SKU. Five stages. Nothing ships to the store between stage one and stage five.

Stage 01 Scanning competitors
SerpAPI fan-out · authorized retailers weighted higher · timestamps logged
8 retailers found anchor $93.50 weighted by authority
Stage 02 Three candidates generated
bounded by hard cost-floor (low) and observed market ceiling (high)
Floor
$64.20
Anchor
$91.99
Ceiling
$108.00
Stage 03 Reasoning
Elasticity −1.42 from 412 prior orders. Demand steady. Competitor anchor $93.50, EWMA-smoothed across last 6 scans. $91.99 holds velocity, lifts GP $4.49 per unit, well above floor.
Stage 04 Recommendation
$87.50 from $87.50 +5.1% gross profit
Stage 05 Pending review
queued for the merchant. nothing ships to the store without explicit approval.
Queued · awaiting decision

04 Operator review

Nothing ships to your store without you.

Every recommendation lands in a pending review queue. Approve, modify, or reject. One-click revert on anything that didn't behave.

Product Current → Recommended GP Δ Confidence Market Action
BSH-7016R Brake shoes, rear, 1-ton diesel
$148.50 → $156.20 +5.2% 92% avg $158.40 (n=11)
WLD-2050S 0.030 in. mig wire, 33 lb spool
$87.50 → $91.99 +5.1% 87% avg $93.50 (n=8)
HDP-4140 Hydroponic pH down 1 gal, food grade
$32.99 → $31.49 −1.4% 89% avg $30.80 (n=14)
DSL-9120F Diesel fuel filter, HD class 8
$41.20 → $41.20 no change 54% avg $42.10 (n=4)

Click a row to read the reasoning. Every decision is logged. One-click revert on anything that didn't behave.

05 Safety / trust

The engine recommends. The merchant decides.

Recommendation-only first. Every change waits in the pending review queue until you approve it. Cost floors enforced before the reasoning layer runs. Move sizes capped. Risky SKUs excluded by default. Designed to keep every recommendation inside merchant-approved constraints.

$108 $92 $76 $64 × FLOOR · $64.20 $91.99 ↑ rec
  • 01
    Recommendation-only first

    Pilots start with the engine generating recommendations but never writing to Shopify. Write authority unlocks per-SKU only after the merchant has watched the engine's behavior on their catalog.

  • 02
    Hard cost-floor check

    Every candidate is compared to the per-SKU floor before the reasoning layer runs. If a candidate breaks the floor, the candidate is dropped. Not flagged. Dropped.

  • 03
    Capped move sizes

    Single-move maximum is hard-capped: 3% in early phases, 5% later, configurable down. The engine cannot move faster than your tolerance, even when the math says it could.

  • 04
    Exclusion lists

    SKUs and categories you mark as off-limits never enter the candidate set. Premium / branded / MAP-restricted items are excluded by default until you say otherwise.

  • 05
    SPC circuit breakers

    If demand drops anomalously after a price change, the engine auto-reverts and pauses recommendations on that SKU until a human reviews.

  • 06
    Pending review on every change

    Every recommendation lands in the queue first. Approve, modify, or reject. Nothing ships to the store without your explicit approval. The merchant is the final decision-maker.

06 Why not a repricer

Match-lowest chases velocity. Outprice optimizes gross profit.

Honest comparison. We're not naming names; the patterns are the patterns.

Question
Match-lowest repricer
Outprice
What does it optimize?
Lowest visible price within a margin floor.
Gross profit per SKU, given elasticity, demand, and competitor distribution.
Can it raise prices?
Structurally no. Beat-by-X is one direction.
Yes. When elasticity and demand support it, the engine raises and explains why.
Thin-data SKUs?
Same rule applies. No concept of data quality.
Bayesian shrinkage to category priors. Holds when the signal isn't there, with the reason logged.
Reaction to flash sales?
Chases the drop. Margin lost on every sale until the rule reverts.
EWMA-smoothed anchors. One outlier scan absorbed; only sustained moves get followed.
Who decides?
The rule. Pricing executes automatically.
The merchant. Pending review on every change. The engine recommends; you decide.
Audit trail?
Price changed at timestamp. That's the log.
Per-decision reasoning, market context, confidence, bandit arm, SPC state. One-click revert on anything.

07 Method

What's under the recommendation.

Nothing exotic. Defensible techniques you'd recognize from a quant research note. Six of them, applied per SKU, every scan. The methodology supports the recommendation; it isn't the homepage's main pitch.

loge(Q) ~ loge(P) · WLD-2050S · 0.030 in. mig wire
Demand curve · Bayesian posterior
Quantity / week (log) Price (log) →
Sample size · drag to see Bayesian shrinkage 312 obs.
n = 20 · thin n = 500 · rich
Effective β
−1.353
Confidence
92%
Recommended
$156.20
Posterior weight λ
0.16
thin data, model defers to category prior β = −1.00
Engine pipeline · per-SKU, per-scan
data flow
Cost data Order log Competitor scan EWMA anchor Bounded candidates Reasoning Recommendation + reason Pending review queue SPC circuit breakers Cost-floor guardrail · hard reject before anything ships auto-revert if anomalous
scroll horizontally to see the full pipeline →
Per-SKU log-log OLS elasticitylog(Q) = α + β log(P) + ε
We estimate how price-sensitive each SKU is from your real order history. SKUs with hundreds of observations get tight estimates; the rest get handled below.
Bayesian shrinkage to category priorsposterior = λ · prior + (1 − λ) · sample
Thin-data SKUs borrow strength from similar products instead of guessing. The shrinkage weight scales with how much data you have.
EWMA-smoothed competitor anchorsαt = γ · obst + (1 − γ) · αt−1
One outlier scan won't whiplash your prices. Recent observations weigh more than old ones; flash sales get absorbed, real shifts get followed.
Multi-armed bandit explorationThompson sampling, ε-decay
Intentional learning moves, never random. A small fraction of decisions explore alternative price points to keep elasticity estimates fresh; the rest exploit what's already known.
SPC circuit breakers3σ control limits on demand residuals
If demand drops anomalously, the engine auto-reverts and pauses. Not a soft warning. The price reverts and the SKU sits out of recommendations until reviewed.
Hard cost-floor guardrailscandidate < floor → reject
Cannot recommend below your floor. Period. Floor checked before the reasoning layer ever sees the candidate.

Methodology appendix, not a marketing slide.

08 Validation

Validated in simulation. Honest about it.

We measured this before letting it touch a live store. Two academic-grade datasets, 1,200+ seed runs, Wilson 95% confidence intervals, and a real-world reality haircut applied to every customer-facing number.

Dominick's Finer Foods
Academic dataset
6,200
UPC × store combinations
53M
scanner rows, weekly resolution
Olist Brazilian e-commerce
Public dataset
50
sellers, multi-category
8,000
decisions, simulated against historical pricing
Seed runs
1,200+
Confidence interval
Wilson 95%
Drawdown analysis
Worst-15% reported
Reality haircut
30–50% applied
0%
98% agreement with engaged-merchant decision patterns. Across the same SKU, same week, same competitor set, the engine and the operator agreed 98% of the time, with the engine catching margin opportunities the operator missed in the remaining 2%.
++0%
Post-haircut gross profit lift on commodity catalogs. Auto parts, hydroponics, welding consumables, HD diesel parts, industrial materials, specialty hardware. The defensible range. We say "post-haircut" because we mean it.
Per-category lift, post-haircut · Wilson 95% CIs
Figure 3 · validation
post-haircut GP lift % 0 −5% +5% +10% +15% Auto parts HD diesel parts Welding consumables Industrial materials Hydroponics Specialty hardware Aftermarket parts Premium / branded SKUs ↓ contained by review queue
n = 6,200 UPC × store · Dominick's Finer Foods scanner data + Olist · Wilson 95% CIs · post-haircut · 1,200 seed runs
Known weakness
Premium and branded SKUs, where price signals exclusivity rather than tracking elasticity. The pending-review-queue model contains this by design. The engine recommends; the merchant decides. Acknowledged weakness; not hidden.

09 Fit

Honest about who this works for.

First wedge: aftermarket auto parts, HD diesel, performance, off-road. The pricing dynamics there fit the engine cleanly. Adjacent commodity verticals come later. We'd rather lose the lead than disappoint the operator.

Works well

+ fit
  • Aftermarket auto parts, HD diesel, light-truck accessories, performance / off-road, heavy-duty, motorsports parts
  • Shopify merchant, 500–2,500 active SKUs
  • $50K–$500K / month GMV
  • Real per-SKU COGS data, kept current
  • 3+ visible competitors per SKU on Google Shopping for most of the catalog
  • Operators measuring gross profit, not GMV vanity
  • Margin-sensitive, comparison-shopper-driven categories

Not the right fit

− fit
  • Premium / luxury / DTC brands where price signals exclusivity
  • Single-brand or in-house-brand-only catalogs (no third-party SKU resale)
  • Heavily seasonal catalogs (one peak window dominates the year)
  • Merchants without per-SKU COGS data
  • MAP-restricted catalogs without a maintained exclusion list
  • Teams wanting blind match-lowest automation, no review
  • Catalogs under 100 or above 5,000 SKUs (current pilot scope)

Not sure? Email us. We'll tell you within a day, honestly.

10 Pilot ladder

Trust earned, not asserted.

Audit first. Shadow second. Recommendations only after fit is proven. Write-mode only after recommendations behave. Each phase commits less than the next, by design.

Phase 0
Audit
Read-only · analysis
Freequalified merchants
5 working days

5-page audit deliverable. Competitor pressure, underpriced SKUs, risky SKUs to hold, cost-floor concerns, fit assessment. We'll tell you if it isn't a fit.

Phase 1
Shadow
Recommend, don't write
$1,500prepaid
14 days

Engine generates recommendations daily. You mark would-approve / would-reject. Nothing ships to Shopify. Calibration window for floors and exclusion lists.

Phase 2
Recommend
Write on per-SKU approval
$1,500/ month
30 days

Pending review queue active. Write-mode opens per-SKU when you approve. 3% maximum single-move cap. SPC circuit breakers live on every change.

Phase 3
Operate
Expanded with caps + review
$1,500/ month, ongoing
Month 4+

Per-SKU write authority across the active subset. 5% maximum single-move cap. Weekly review cadence. Exclusion list always honored.

Get a pricing audit The pilot starts after the audit. Not before.

All phases include cost-floor enforcement, max move caps, exclusion lists, and the pending-review-queue model. Recommendations are advisory; the merchant is the final decision-maker. Validation is on simulation; results are not guarantees. Best fit observed in commodity-shaped catalogs.

11 Questions

The ones worth asking before a pilot.

What's the difference between Outprice and a rule-based repricer?
A rule engine applies one undercut formula across the catalog. Outprice estimates per-SKU elasticity from your order history, anchors EWMA-smoothed competitor distributions, runs a multi-armed bandit for intentional learning, and lands every recommendation in a pending review queue. The engine recommends; the merchant decides. A rules engine has no concept of data quality, no concept of demand, and no off-switch when a flash sale runs.
Are the validation numbers real?
Yes, and labeled honestly. Validation ran against the Dominick's Finer Foods academic dataset (6,200 UPC × store combinations, ~53M scanner rows) and the Olist Brazilian e-commerce dataset (50 sellers, 8,000 decisions). 1,200+ seed runs, Wilson 95% confidence intervals, drawdown analysis. Every customer-facing number gets a 30–50% reality haircut applied before we publish it. The +5–10% post-haircut figure is the defensible range on commodity catalogs.
Will the engine drop my prices below cost?
No. Hard cost-floor guardrails are checked before the reasoning layer ever sees a candidate. If a candidate hits your floor, it's dropped, not flagged. SPC circuit breakers monitor demand after every change; anomalous drops auto-revert and pause recommendations on the affected SKU.
What does the pending review queue actually do?
Every recommendation lands in a queue. Each row shows the SKU, current → recommended price, GP delta, confidence, market context, bandit arm, SPC state, and a written reason. Approve, modify, or reject. Nothing ships to the store without your explicit approval. One-click revert on anything you change your mind on later.
My catalog isn't commodity-shaped. Is this a fit?
Probably not, and we'll tell you. Premium / branded SKUs are the known weakness; price there signals exclusivity, not elasticity. Highly seasonal catalogs and stores without COGS data are also outside scope. The fit list is the fit list. Email us if you're unsure; you'll get a real answer within a day.
What does "currently installs through Shopify" mean?
Shopify is the install vector right now. The engine reads order history through Shopify's API for elasticity estimation and writes price updates back when you approve them. Outprice runs as an embedded app inside Shopify admin. Permissions are limited to read_orders, read_products, write_products. The platform itself isn't Shopify-only forever; this is where the operators we serve are today.
How does the pilot ladder actually work?
Four phases. Phase 0 · Audit: free for qualified merchants, 5 working days, read-only analysis, ends with a 5-page deliverable. Phase 1 · Shadow: $1,500 prepaid, 14 days, the engine recommends but never writes to Shopify; you mark would-approve or would-reject so we calibrate floors and exclusions. Phase 2 · Recommend: $1,500 / month, 30 days, write-mode opens per-SKU when you approve, 3% maximum single-move cap. Phase 3 · Operate: $1,500 / month ongoing, expanded write authority across the active subset, 5% cap, weekly review cadence. Each phase commits less than the next. You can stop at any phase boundary.
Who's behind Outprice?
A solo founder, pre-launch, building deliberately with merchants we trust. The validation work (Dominick's, Olist, 1,200+ seed runs, Wilson intervals) was the gate before any pilot. If you want to talk to the founder before applying, the email at the bottom goes to a human.

12 Apply

If your catalog is built on margin, run it through this.

Not autopilot. Not match-lowest. A pricing control room that recommends, explains itself, and waits for you to approve. The audit comes first.

Request pricing audit

If your catalog has real COGS, competitor overlap, and hundreds of SKUs, we'll show where pricing may be leaking margin.

Outprice · pricing intelligence infrastructure · built for Shopify merchants first · pre-launch · validated on 6,200 UPC × store combinations · 1,200 seed runs · simulation results are not guarantees