Return on Investment : Why CPG Leaders are using Image Recognition for Perfect Store Execution

Blog Cover Image in blue- Calculating Return ion Investment when using Image Recognition for Retail Execution for Perfect Store creation

A perfect store is one that provides a seamless shopping experience to the consumers. Implemented easily with the help of retailers and sales reps, it helps build a positive brand image which in turn results in profitability for both the retailer and the CPG manufacturer. This concept was born in the world of CPG companies and is termed officially as “Perfect store” by Unilever, “Golden outlet” by P&G , “RED – Right Execution Daily” by Coca- Cola and “Flawless Execution” by PepsiCo.

Over the past few months, we’ve written in detail about the intricacies involved in setting up a perfect store – why the need for the perfect store concept was felt, how to execute and measure its effectiveness & the KPIs involved. We’ve also mentioned how one of the best ways to achieve the goal of that elusive perfect store is through image recognition or “computer vision” technology.

But there is one question that often comes up, and rightly so – Does image recognition for perfect store execution generate significant return on investment for CPG and retail brands? In this blog, we address this skepticism, and detail how Image Recognition technology perfects your retail execution, thus increasing sales in the long term.


Only creating guidelines for the purpose of perfect retail execution is not sufficient. Its successful compliance via auditing  is necessary. CPG & retail brands can conduct retail audits by four ways –

Different methods of implementing perfect store execution through retail audits
  1. They have their own sales reps that routinely visit outlets to ensure perfect store guidelines are being met
  2. They outsource the retail execution audit to third party auditors
  3. They make use of the point of sale data provided by their retail partners.
  4. They make use of AI based Image Recognition solutions for retail execution – however this option is only now gaining traction.

But mostly the audit is conducted by filling surveys or capturing images that are later analyzed. The process is riddled with inaccuracies emanating from human errors/bias. Some glaring issues faced are –

Issues Faced by CPGs when implementing Retail Audits
  • Limited Manpower : Consumer Goods Manufacturers are not able to obtain a complete picture of retail execution – their products are present in many outlets, but retail auditing takes time, so brands are able to audit only a few outlets due to limited manpower. 
  • Finite Skillset : Calculating KPIs might not be the core skill of most sales reps, especially complicated but important KPIs like Linear Share of Shelf. This results in data gaps and misinterpretations, resulting in major loss of accuracy. This leads to errors.
  • Time Consuming : Such intensive data collection, KPI calculation and insight derivation is time consuming. Sales reps often have to compromise in making real time improvements in-store, that can augment sales. 
  • Conflict of Interest – A conflict of interest arises if the reps are measuring their own performance. This means if the brands wish to run a bonus/incentive driven perfect store programme, they’ll have to make do with the unverified data they are provided.
  • Lack of Instant Insight generation – Most importantly, it takes time to collect data, interpret it and then devise it to achieve a course of action. By the time a remedy is obtained, ground realities have changed, making the insights obsolete. Thus effectively, no value has been added to better the outlet in accordance with the perfect store guidelines.

Filling surveys, solely using SFA apps and capturing images to be analyzed later, all fail to address the core issue at hand – which is, are the insights generated, timely acted upon to ensure the perfect store guidelines are met ?


Let there be a chain of retail outlets called SuperMart. Now, this chain has a total of 500 outlets. And the total sales generated, per year per outlet is $10 M. Here’s how image recognition makes retail execution feasible for SuperMart – 

Infographic - Perfect Store Execution - Return on Investment through Image Recognition

Man-hours saved – It saves 20-30 man hours per week for scanning 8000+ SKUs. Resulting in potentialsSavings of around $30k – $50k per year per outlet.

Incremental Sales – For every 3% improvement in OSA, retailers can experience a 1% increase in monthly sales. Our image recognition solution ShelfWatch can help reduce OOS rate by up to 6%, meaning a sales lift of 2%. Which means $200K per year per outlet. 

With ShelfWatch, the chain can recover up to $210k per year per outlet OR $105 mn potential benefit for the entire chain annually.

Below we discuss how these numbers are achieved.


The core issue mentioned above is – “ Are the insights generated, timely acted upon to ensure that the perfect store guidelines are met ? “

Computer vision is able to address this issue because it’s designed to operate in a dynamic environment – a trait which is inherent to the retail sphere. 

When it comes to RoI generated by using image recognition for retail execution, one needs to understand this dynamic environment – that is, the definition of RoI varies from brand to brand and the region they are selling their products in. RoI can be generated in terms of man hours saved, accuracy of data collected and the timely generation of insights to address lapses in retail execution. 

Speed of Automated Retail Audits –

If a brand needs a complete picture of their in-store execution, including relevant (if not maximum) retail outlets under the ambit of retail audit makes sense. In certain regions, affordable manpower is an issue. They need a solution that is not only accurate – but is also quick about it. The saving is in the form of time saved by sales reps, which is then devoted to activities that add value to the brand. 

For example – In one of our project’s with a top global beverages company, ShelfWatch provided instant image recognition for 800+ outlets. Out-of-Stocks and errors in Planogram Execution were addressed by the field reps in-store. The insights were generated with a turn-around time of less than a minute, cutting down the audit time by 50%.

Accuracy of Insights generated for the Perfect store Programme –

In regions like South Asia, retail audits are marred by the limitations that manual audits bring – bias, lack of skill and human error. The field rep may not be skilled to measure and calculate complex KPIs like Linear Share of Shelf. This is where Image Recognition steps in. It correctly identifies SKU and its many variants, calculates the KPIs and checks for compliance. 

For example – In one of our projects, ShelfWatch achieved 95%+ accuracy in On-Shelf Availability metrics. In another project, we achieved SKU-level accuracy of 98%+ in image recognition. For a leading Brewery, ShelfWatch calculated Price Display Detection – Its accuracy was 95% despite the fact that the price display detection was added later on to the project KPIs.

ACTIONABLE INSIGHTS leading to improvement in key perfect store KPIs –

Image Recognition Solutions when consistently used to monitor key KPIs, leads to their improvement. Here we cite some examples – 

  1. A Global Cleaning Products Company –  25% improvement in Core-SKU OSA within 3 months
  2. Top Food Products Company – 30%  improvement in planogram compliance in 2 months. 
  3. Global Food Products Company – Over 20% Improvement in purity, placement and planogram scores in 3 weeks 

You can read detailed case studies here.


Image recognition solutions can also offer an end-to-end integrated solution for perfect store execution. It does so by making use of  two features – the App and the Dashboard. A feedback loop is created in the following manner –

shelfwatch worklow image recognition for retail execution

Image recognition instantly highlights the missing SKUs on the shelf – thereby becoming almost like an assistant to the field rep.

Let’s say we are calculating On-Shelf Availability – What image recognition solutions do is create a robust redressal mechanism in the form of a virtuous feedback loop. It gives real-time OSA insights to both the sales rep and the CPG leadership. The sales rep is able to solve the issue immediately in the outlet, and simultaneously the AI also relays the issue to the CPG HQ. The brand leadership has a dashboard which has access to the OSA calculated, and they communicate with both the sales rep and the retailer to improve their OSA numbers.

Limited Resource Input –

Image Recognition often comes under crosshairs for the amount of input that CPG & retail brands have to provide their vendors. But that is not always true. At ParallelDots’ we have our proprietary one-shot image training technique where only one packshot of the SKU is required to train the AI. 

For example – A major brewery did not have a list of outlets and the planograms applicable in them. So we built them an AI model to automatically identify applicable planogram from a list of planograms for a particular outlet. There are many such other examples that indicate implementing image recognition for perfect retail execution is simple – its ease of use is just a function of the vendor you choose.

Affordable, Scalable and Consistent –

Manual Audits are inefficient and erroneous. And third party audits are too expensive to scale – thus reducing the frequency and consistency with which retail audits should be performed to effectively implement perfect store programmes. 

This is where image recognition solutions win – because of its affordable nature, they can be easily scaled to include more retail outlets in the ambit of the brands’ perfect store initiative. With frequency of visits from quarterly to say, bi-weekly – the sales reps are able to ensure that the retail partners follow the terms of trade.

Accurate assessment of Loyalty/ Incentive Programmes –

A perfect store card is generated based on compliance scores for every retail outlet. The retail partners and sales reps are ranked based on their performance on the perfect store card. 

Since we’ve established that image recognition solutions offer reliable, accurate insights – they can be relied upon to calculate percentage bonus for retail partners and sales reps. This helps in creating an incentive based perfect store programme – which means both the retailer and the sales reps are now invested in the health of the brand.

In conclusion, we can definitely say that retail execution is a highly dynamic process – where the variables change very quickly. In order to keep up with this rapid change, a smart solution is required. It is by using this smart solution, CPG brands and their retail partners can hope to create a perfect store.

Want to know more about other shelf KPIs? Read our next blog to find out.

To see how your own brand is performing on the shelves, click here to schedule a free demo of ShelfWatch.


Khyati Agarwal

Leave a Reply