How to implement Shelf Recognition tool, organize the processes and win the market?

Nutricia Poland is definitely widely recognizable on the Polish market, especially among parents of young children. The products of the company, with its brands such as Bebiko 2, Bebilon 2, Bobovita, accompany children in the early stages of their lives. Nutricia is a part of a larger group of Danone companies with its beginning of eLeader cooperation dating back to 2015/2016.

Handling conversations with our potential customers we pointed out how often we come across the same questions and doubts about Shelf Recognition AI talking to our Sales Representatives. The most frequently asked questions are: What difficulties wait ahead while implementing Shelf Recognition AI system?  What should we be aware of? How to prepare for the process? Is it really as scary as it seems? Is AI capable of replacing a human factor?

We have decided to ask these and many more questions to Sebastian Zapała – the sales performance manager of Nutricia Poland. That was one of the first implementations of out Shelf Recognition AI system, so the point of view presented by our guest, who paved the way for further implementations, is one of the key hints in dispelling concerns for those planning to start their journey with Shelf Recognition system.

Talking to Sebastian Zapała are Paweł Majsiej and Rafał Melanowicz.

---

Paweł Majsiej: We started our cooperation around 2015/2016. At that time you were definitely  the pioneers on the market as the first ones implementing the Shelf Recognition system to your selling processes.

Sebastian Zapała: My adventure with the position of the sales performance manager started exactly at that point when the Image Recognition system was implemented in Nutricia Poland and we chose eLeader to supply this solution. The Shelf Recognition tool was only at the early stages of its development and, at that time, not many companies were ready for the automatization of business processes with the help of Shelf Recognition tool. Now I can admit that this project was one of the hardest and most challenging implementations in my business career.

Was it more of an innovation drive or business necessity?

I do not believe in innovation for its own sake. Business need is, and should be, the trigger. Also, we, personally, faced a certain challenge. We wanted to ensure that the exposition of our products on shelves always met our standards. The second need was more transparent and reliable system informing about product availability on store shelves. Back in time the most basic and commonly used method of availability check were surveys performed manually by Sales Representatives. This kind of process was, in itself, very erroneous. The paradox was that no one was capable of measuring the extent of error. In the case of Image Recognition, the process is much easier. This is a transparent form of verification as it involves the actual photos that are always available to prove the state of things.

We happen to call the Shelf Recognition “a tool that takes off the burden of the Sales Representative to be judge in his own cause” – that functions as an impartial mediator that supports and watches the Sales Representative to prevent moral dilemmas over reporting the results.

Precisely.

Has the way of using Shelf Recognition in Nutricia changed over the years?

The Sales Reps provided us with two main indicators. One of them was the availability of products on store shelves, the other – the share of shelf. The first one was checked with every single visit. Unfortunately, the results were received with a certain delay. The other indicator, due to its complexity, was measured quarterly.  And suddenly, from these two indicators, the number rose up to  nine! That was the beginning of Perfect Store&Visit module that we currently use.

There has also been a change in technology. Our goal was to provide the Sales Rep with the results right during visits. We also wanted to get answers to our basic questions: what is my availability? What is my share of shelf? What are the remaining parameters? Having those data during visits allows the representative to react to them straight away. Fifteen years ago, technological advances could not yet provide us with such thorough outlook. The fact that, at this stage, we have the results at our disposal during visits really facilitates the process.

Does the possibility of improving the exposition during visits affect its quality? Can it be somehow measured or checked that this exposition works better?

The very fact that the reporting was based on photos meant that Sales Representatives really needed to take care of the exposition. They knew that products needed to face the client because that guaranteed that the system would recognise the products. Therefore, the parameters, they were responsible for, were diligently reported and assessed. This marked the first benefit – the need to take a photo. This factor, that the photo is the basis for the indicator measurement, meant that the shelf was properly taken care of and therefore well displayed. The Sales Representatives learnt that pretty quickly.

Initially, we expected that the implementation of Shelf Recognition would additionally shorten the time of visits. However, it was not the case at first. It appeared that merchandising actually takes more time. We, then, concluded that before the shelf must have been prepared not as well as we had wished.

For many parameters we had to work for longer periods of time.  Here, as well, we faced issues of contracts with business partners. Dutiful implementation of individual exposition standards is
a lengthy process, however, the effect is measurable and noticeable.

Should we, then, be concerned about the fact that the time of visits has not been shortened when actually the time is more satisfactorily managed?

Not at all. In the first phase, the time of visits might have been the same, whereas, in a couple of years, we notice time savings at over 20%. The reporting system clearly helped us to achieve such results.

We have reduced the reporting time, at least, by half. From dozen of minutes of check to several. Previously, the check of shelf shares, by means of manual surveys and checking availability of products on shelves, counting the so-called faces, lasted over an hour. Now, it is a matter of taking a photo and counting the results by the algorithm and, at the same time, we, not only, have our exhibition checked  but also the entire category! All this in one merchandising task.

Did the Sales Representatives feel any improvement in their competences within this change?

We make jokes about our Sales Representatives that they have changed from the pollsters to photographers. Still, we remember that taking good photos is a key competence for our Field Force.

Though, Sales representatives now have a much better understanding of what a store looks like. They have access to historical data on many parameters, so they are equipped with information about how business has been developing in their stores over the years. We also began to focus more on shaping business relations with our partners rather than support models in which Rales representatives are the main sources of information about the situation in the store.

Does it expand the scope of reporting also in terms of business knowledge?

It surely does. We have many more parameters at our disposal that we can observe throughout months and even years. We also have many more indicators that we measure and we pay attention to. However, the amount of data must be approached with caution. They must be managed in a thoughtful way, so as not to get lost in their multitude.

One of the most difficult challenges when implementing this type of system, when we are already in awe of what we can measure, is a realistic assessment of what we really need. It’s easy to overheat it. And that was our first mistake.

If I were responsible for such implementation one more time, I would divide it into stages. You need a clear vision about the number of indicators you want to measure and step by step achieving the  goal you set. This kind of implementation is a lengthy and challenging process for the whole organization.

Rafał Melanowicz: What effect does better exposition of “baby food” have on the sales in general?

We look at shelves as our customers do. If something is missing from the photo, it means that our customers miss that too. Even if the products can be found on the shelf, but they do not face the customer, we can say that they are not there.

There is a so-called moment of truth in the purchasing process: the moment when the customer stands at the shelf and chooses which product to reach for and add to the basket. This moment could be won only if the product is well displayed. For clients for which we gather data on each of their stores, together with Perfect Store indicators, we see a correlation and we know that the better the shelf parameters are, the better the sales, to put it simply. This is confirmed in numbers.

Has the share of your products in the Baby Food category changed over the 6 years of using Shelf Recognition? Is it in line with the strategy?

The shelf share has grown and it is in line with the strategy. As a market leader, we must be fair to planograms or other arrangements with our business partners, and our share of shelf often reflects our market position.

It is worth mentioning that while analysing share of shelf we notice every single change on the market. If we come across any drop, we can quickly analyse the cause of it. Is it a consequence of absence from the market shelf? Is it connected with a new player on the market? Is it caused by increasing the share of products of our competitors? We have access to data on the situation with the shelf and, on the basis of those data, we can take quick and deliberate business decisions.

So, as eLeader, can we say that our solution helped Nutricia achieve the goals you aimed?

Of course. We have worked together to come up with such solution. If it was not for the eLeader solution, and the possibilities it offered, we would not be able to succeed. eLeader is a flexible and intuitive app that helps people perform their tasks for each store in an efficient manner.

Paweł Majsiej: I think it is worth emphasizing that the implementation of such a solution, without a conscious and engaged client – a partner who knows what he wants to achieve, will not solve problems.

We can go even further: technology provides certain information, however, a conscious business decision of what to do with it is the key.  The Sales Representative takes a photo, the recognition result is low, what would be your first conclusion? The tool does not work as it does not successfully recognize the products. However, if you take a closer look at the photo, you notice products turned back, hardly visible and covered. Some people may react that the products are there but the solution fails. Still, the question is: do I want my products to be displayed in such a manner? If I look at the products in the picture in such a way, the consumer will see them in a similar way. Will they be willing to pick my products that are unnoticeable and disorganized on the shelf?

Let’s try to combine these threads. Aren’t you tempted to go further and analyse product labels for their identification at the consumer level before launching a product?

Sure thing. We look at it and, from my experience, I can say that, very often, we learn about the change in the packaging – of our or the competitor’s products – from the Shelf Recognition analysis. Even the slightest change in the packaging layout can make a product unrecognizable. It happens in the industry that marketing teams are not being always informed about the change in the price labels. In such cases, an image recognition solution can be helpful in streamlining, for example, the process of introducing new products.

We do not yet have enough experience to build a product visualization or the structure of its packaging so that it is ideal for recognition by Shelf Recognition. But maybe this should be our next step.

You mentioned that, at the beginning of the introduction of Shelf Recognition, the change was a phenomenon that caused some resistance from your employees. Do you notice that, just like in the 18th century, during the industrial revolution, Luddites destroyed weaving machines because they felt that it would be a threat to their employment, so here there are fears that the machine replaces human labour, and this may lead to massive layoffs?

I think, now, after years, the Sales Representatives appreciate this change. We directed at the development of their competences. Technology will never replace a human factor in full. I always assume that technology often serves the people in the field because it saves their time and allows them to focus more on building and developing business relationships in the store. They have the extra time to spend on active work because reporting is simplified. There are some other, not less, valuable benefits: more effective visits, better route plans. All of the aforementioned may only result in reaching more stores and clients.

Do employees, having different workplace experiences and lack of such solution in their previous companies, perceive automatic audit as a support or rather a way of control?

The feedback, I came across, is rather positive and leans towards support. This is also how we try to approach it and communicate it to our employees. Sn automatic audit is a supplier of information for us – the organization. One of the entities for whom we collect this information are Sales Representatives. They also use feedback from Shelf Recognition analysis. The results provide them with list of products to be ordered or are the basis for business conversation with the store staff.

Rafał Melanowicz: Let’s assume that I had KPIs that were built on the basis of manual surveys. I showed my owner, management board and supervisory board information about shares and availability. Shelf Recognition appears and – I assume – the shares have fallen by 5%. How can this manager, who has been relying on this manual data for 5 years, get out of this inconvenience?

There is an English saying “Don’t kill the Messenger”. It means that we should never kill the person bringing information to us. You can either take the information and use it or ignore someone telling you an important truth about your business. What you do with the given information is up to you in the end.

What if I’m a listed stock-market company? Let’s assume I just have to announce that the moment I introduced Shelf Recognition my market share dropped by 5%. You know how the stock price reacts. What should I do?

We can approach it carefully, process-based. Let’s double the trouble (two recognition tasks) – manual and with the help of Shelf Recognition system. Let’s compare them afterwards. Let’s see which system is better, which one is less erroneous. We have acted in such a way within several business projects and we have found out that Shelf Recognition tool recognized product Sales Representatives failed to mark as present. People may often be wrong as well as machines. Such error may be a two-way street. Products may be marked as present if they are not or not marked when they should have been. In the case of Shelf Recognition we can always rate the tool and fix it. In case of manual surveys it is less effective. Shelf Recognition error margin is 2.5%. If I am aware of this margin, I can work with this information. What is the error threshold for a manual survey though? I am unable to check it. What you do with this information is up to you.

Many companies ask Sales Representatives to document with a photo the manual state of reporting. I talked to several people responsible for manual analysis of the photos and most of them admitted that the process is simply lifeless. Gigabytes of pictures placed on discs left there and no one has the time to analyse them.

That’s what I’ve always wondered – why? You can take the time to visually check a random selection of one hundred photos to ensure that the survey is accurate. In my opinion, the reality is that no one is doing it, or is quickly giving up on it. This is a difficult job – browsing through hundreds, thousands of photos. It’s an inefficient process and unproductive work. Why should a man do it when there is a system that can do it for him? Why take a survey and photos when photos alone can be satisfactory?

---