Commercial Profit Model

Published: Jan 12, 2026, 4:19:23 PM

Updated: Aug 4, 2026, 11:35:00 AM

📈 Methodology, Formulas and Assumptions

Important: What you see here is a conservative, explainable baseline, not an inflated best case.

We believe in radical transparency. The results you see in our profit model are not magic numbers. They are the output of a financial model that combines:

Your shop’s inputs (Visitors, Margin, AOV, Return Rate, etc.), and
Our benchmarks for how guided, AI-assisted shopping impacts behaviour.

This page explains:

The logic and formulas behind each pillar
The constants used in the model
How we validated our assumptions using independent, non-competing sources
Why our numbers are intentionally conservative
A short summary to tie it all together

1. The Core Variable: Engagement Rate (E)

Everything starts from one question:

“How many of your visitors will actually talk to the AI?”

We call this the Engagement Rate (E).

The rate depends strongly on visibility and placement:

Strategy	Visibility	Model Assumption	Value Used (E)
Floating Widget	Low (corner icon only)	Around 4% of visitors will engage.	`0.04`
Embedded Advisor	High (within product list)	Around 18% of visitors will engage.	`0.18`
Hybrid (Both)	Maximum coverage	Around 22% of visitors will engage.	`0.22`

Business logic: The AI can only improve the metrics of a customer who actually uses it. All uplift is applied only to the share of traffic captured by E.

These engagement values are chosen in the low-to-mid range of what interactive tools (quizzes, advisors, guided-selling flows) typically see when they are visible in the main journey, and slightly above the 2–5% “generic interaction” benchmark for normal site elements.

Baseline Calculations (Before AI)

We first compute your current baseline, without any AI influence:

TotalOrders        = Visitors * ConversionRate%
TotalRevenue       = TotalOrders * AOV
NetMargin%         = Margin / 100
COGS%              = 1 - NetMargin%
ReturnRate%        = ReturnRate / 100
DepreciationRate%  = DepreciationRate / 100
RetentionShare%    = RetentionShare / 100

These become the reference point against which we calculate uplift and savings.

2. Pillar I: Acquisition (Increasing Sales)

Logic An AI-guided customer is a more confident customer. If the AI answers questions about fit, compatibility, use case or materials instantly, more people pass the “last doubts” barrier and complete their order.

Independent research on personalization shows that companies that get it right typically see 5–15% revenue uplift, and in some cases up to 25%, mostly by lifting conversion. (McKinsey & Company)

Our Benchmark

For the subset of users who engage with the AI, we assume a:

35% increase in conversion rate (vs similar users who did not use the AI).

This is applied only to the engaged segment, and then “diluted” across your whole traffic.

Formulas

GlobalConversionUplift = EngagementRate * 0.35

ExtraOrders            = TotalOrders * GlobalConversionUplift
ExtraRevenue_Sales     = ExtraOrders * AOV
NetProfit_Sales        = ExtraRevenue_Sales * NetMargin%

For typical engagement rates (18–22%), this translates to a global uplift of roughly 6–8%, which stays comfortably within the 5–15% uplift range reported for well-executed personalization.

3. Pillar II: Returns (Cost Savings)

Logic Returns are not just annoying; they are expensive:

You lose the margin on the original sale
You pay for handling and logistics
The product often loses value (opened packaging, damage, aging)

If the AI gives a better pre-purchase recommendation (fit, size, compatibility, use case, materials), the chance of a “regretful purchase” drops.

Studies on e-commerce show that returns account for tens of millions of tonnes of CO₂ per year and a large share of retail emissions, because items travel back and forth and are often repackaged, discounted or discarded. (cleanhub.com)

Our Benchmark

For users who engage with the AI, we assume they are:

30% less likely to return their purchase.

We then scale this by engagement and add a small throughput factor.

3.A Volume Calculation

ReturnReductionRate = EngagementRate * 0.30
EngagementFactor    = 1 + EngagementRate

CurrentReturns      = TotalOrders * ReturnRate%
PreventedReturns    = CurrentReturns * ReturnReductionRate * EngagementFactor

ReturnReductionRate handles “fewer returns per order”.
EngagementFactor reflects that slightly more orders are placed thanks to confidence, so we capture that effect in the same block.

3.B Financial Savings (Three Layers)

We then turn prevented returns into money saved:

Margin Protection The profit you do not have to give back:
```
Saved_Margin = (PreventedReturns * AOV) * NetMargin%
```
Operational Savings Direct handling costs (your input: CostPerReturn):
```
Saved_Ops = PreventedReturns * CostPerReturn
```
Depreciation Savings The cost of goods that lose value when opened, damaged or repackaged:
```
Saved_Depreciation = (PreventedReturns * AOV * COGS%) * DepreciationRate%
```

Total Return Savings

TotalReturnSavings = Saved_Margin + Saved_Ops + Saved_Depreciation

This does not try to capture all environmental or brand effects of fewer returns, only the concrete financial and basic CO₂ impact.

4. Pillar III: Loyalty (Recurring Revenue)

Logic A helpful, fast and human-like experience builds trust. Trust leads to repeat purchases.

Multiple independent surveys show that:

Around 80% of consumers are more likely to buy from brands that offer personalized experiences. (Epsilon)
Personalization leaders generate significantly more revenue from repeat customers than their peers. (McKinsey & Company)

Our Benchmark

For users who successfully engage with the AI, we assume:

A 50% increase in retention probability (they are 1.5× more likely to come back).

This uplift is only applied to the share of revenue that already comes from returning customers, and again is scaled by the engagement rate.

Formulas

LoyaltyUplift        = EngagementRate * 0.50

RepeatRevenueBase    = TotalRevenue * RetentionShare%
ExtraRevenue_Repeat  = RepeatRevenueBase * LoyaltyUplift * EngagementFactor
NetProfit_Repeat     = ExtraRevenue_Repeat * NetMargin%

For typical engagement rates, this translates into roughly a 10–11% uplift in repeat revenue at shop level, which fits within the broader “5–15% revenue uplift” band for strong personalization.

5. Sustainability Impact

Logic We model two mechanisms:

Fewer returns → less transport and packaging
Faster answers → less “wandering around the website” → fewer heavy page loads

5.A Digital CO₂

Independent analyses estimate that the average web page emits around 0.36 g CO₂ per page view, with heavier, media-rich sites significantly above that. (Website Carbon)

E-commerce category and product pages tend to be on the heavier side (images, scripts, tracking). We therefore use:

4.3 g CO₂ per engaged user as an approximate saving, interpreted as “avoiding at least one extra heavy page view” per guided session.

EngagedUsers   = Visitors * EngagementRate
DigitalCO2_g   = EngagedUsers * 4.3

5.B Physical CO₂ (Transport and Packaging)

Research on parcel logistics and packaging finds that:

A typical home-delivered parcel emits around 0.27–0.30 kg CO₂ in last-mile delivery. (Nature)
A common e-commerce cardboard box can add roughly 0.2–0.3 kg CO₂ over its life cycle, depending on size and production method. (Consumer Ecology)

Our assumptions:

TransportCO₂ per prevented return: 0.5 kg
PackagingCO₂ per prevented return: 0.3 kg

This is conservative compared with full “product + waste + processing” estimates for returns, which can reach several kilograms of CO₂ per item. (ScienceDirect)

Formulas:

TransportCO2_g   = PreventedReturns * 0.5 * 1000   // 0.5 kg → grams
PackagingCO2_g   = PreventedReturns * 0.3 * 1000   // 0.3 kg → grams

PhysicalCO2_g    = TransportCO2_g + PackagingCO2_g

TotalCO2Saved_kg = (DigitalCO2_g + PhysicalCO2_g) / 1000

6. Summary of Constants (Benchmarks Used)

These are the core multipliers embedded in our code, derived from our own experience and aligned with independent research ranges.

Metric	Code Value	Business Meaning
Conversion Uplift	`0.35` (35%)	Extra conversion probability for an engaged user.
Return Reduction	`0.30` (30%)	Lower return probability for an engaged user.
Loyalty Uplift	`0.50` (50%)	Higher retention probability for an engaged user.
Engagement (Floating)	`0.04` (4%)	Typical usage when only a corner widget is visible.
Engagement (Embed)	`0.18` (18%)	Typical usage when the advisor is embedded in the core journey.
Engagement (Hybrid)	`0.22` (22%)	Usage when both floating and embedded placements run together.
Digital CO₂	`4.3 g` per user	Estimated CO₂ avoided from reduced extra browsing.
Transport CO₂	`0.5 kg` per return	Average transport emissions per prevented return shipment.
Packaging CO₂	`0.3 kg` per return	Average packaging emissions per prevented return.

7. Validity of Assumptions(Why Our Numbers Are Conservative)

Our commercial profit model is a scenario tool, not a promise. To keep it honest and useful, we follow three rules when choosing constants.

7.1 How We Choose Benchmarks

We start from independent research. We use neutral sources such as McKinsey, Epsilon, sustainability researchers, UN agencies and academic papers for:
- Personalization and revenue uplift
- Consumer preferences for personalization
- Digital carbon emissions per page view
- Parcel delivery and packaging emissions
- Environmental impact of returns
(McKinsey & Company)
We pick the low-to-mid point of the range. When studies show a range (for example, 5–25% revenue uplift from personalization), we choose a value in the lower half and then apply it only to the engaged segment, not to all traffic. (McKinsey & Company)
We limit the scope. We deliberately ignore several upside effects (higher AOV, lower CAC, word-of-mouth, brand lift) to keep the model simple and conservative. Only Acquisition, Returns, Loyalty and a basic CO₂ view are quantified.

7.2 Acquisition and Loyalty

McKinsey and others report that good personalization typically delivers 5–15% revenue uplift, with some companies achieving up to 25%, alongside 10–30% better marketing ROI. (McKinsey & Company)
Epsilon’s survey of 1,000 consumers found that 80% are more likely to purchase when experiences are personalized, and around 90% find personalization appealing. (Epsilon)

In our model:

We assume a 35% uplift in conversion rate only for users who engage with the AI, not for everyone.
With typical engagement (18–22%), this becomes a 6–8% uplift at site level, inside the 5–15% range.
We assume a 50% uplift in retention probability only for engaged users, which translates to roughly 10–11% higher repeat revenue overall.

Taken together, these numbers are on the cautious side of what independent research suggests is achievable with strong personalization.

7.3 Returns and CO₂

Independent reports estimate that:

E-commerce returns generate up to 24 million metric tonnes of CO₂ per year worldwide. (cleanhub.com)
Shipping and returns can account for around one third of e-commerce emissions. (Sustainable Brands)
For apparel, emissions from unused returned products can be many times higher than emissions from transport and packaging alone. (ScienceDirect)

Digital and logistics emissions:

Average websites emit about 0.36 g CO₂ per page view; heavy sites can be several times higher. (Website Carbon)
Last-mile home delivery typically emits around 0.27–0.30 kg CO₂ per parcel. (Nature)
A common e-commerce cardboard box has a footprint around 0.2–0.3 kg CO₂, depending on size and production. (Consumer Ecology)

In our model we therefore:

Use 4.3 g CO₂ per engaged user as an avoided “extra heavy page view” — safely within published ranges for media-rich sites.
Model each prevented return as 0.5 kg transport CO₂ + 0.3 kg packaging CO₂, even though full life-cycle analyses show that many returns have significantly higher emissions when you include wasted products and additional processing.

This means we are undercounting both the financial and environmental upside of fewer returns.

7.4 What We Do Not Count (By Design)

To avoid over-claiming, our model does not include several effects that could increase the numbers further:

Higher average order value from better matching and upsell
Lower customer acquisition cost because traffic converts more efficiently
Increased referrals and word-of-mouth from delighted customers
Long-term brand and trust effects that lift all future performance

Those are real effects, shown repeatedly in personalization and customer-experience research, but we keep them out of this model so that the numbers remain conservative and easy to justify.

8. Summary

To close, here is what this methodology section is really saying in simple terms:

You stay in control. All results are generated from your own inputs (visitors, margins, return rates, and so on), not from generic averages.
Impact is applied only where the AI actually acts. Uplift is calculated only for the share of visitors who engage with the AI, and then translated back to the whole shop, which keeps numbers realistic.
All benchmarks sit on the safe side. For conversion, returns, loyalty and CO₂, we take values from independent research and stay in the lower or middle part of the range.
We undercount upside on purpose. Extra effects like higher order values, lower acquisition costs and referrals are real, but not included in the model.
The goal is credibility, not hype. The model is designed so that you can show the logic to your team, your CFO or your board and say with confidence:

What you see here is a conservative, explainable baseline, not an inflated best case.