## Contact

4dapoland@gmail.com

## czwartek, 7 kwietnia 2016

### How to properly calculate launch of new product ?

Ability to properly calculate launch of the new product is one of the key competency of commercial departments (marketing,sales, operations-demand planning) in organisation. Especially first weeks are crucial since no sales history is built, thus risk of under or over stock is really high. How to mitigate those risks and have launch well-prepared?

First of all - information based on which assumption are being created. We can divide them into 2 main groups: information regarding distribution of the product and rotation. More details are presented on the graph.

After receving all of those information you can either find a similar product and adjust its volume to current assumption accordingly or if there is no predecessor you can try to get data from competitor's products (e.g. based on Nielsen databases).

Finally, calculations can be made. Here is the example.

1. Each of the variable has its weight and price has the biggest (80%), then discount and advertising has 10 each.
2.% differencies are being calculated in colums H5:H7. Price indicatior has reversal sign. Higher price is actually worsed than lower.
3. With help of sumproduct funtion final hitrate is calculated (H8).
4. At the end of process volumes per each channel are multiplied by this hitrate and results are given in cells G11:G26.

This is how your volume is calculated :) For more information don't hesitate to contact me !

## niedziela, 14 lutego 2016

### There is no good forecast without well-understood past.

Almost every statistical software offers simple time series decomposition. Your past figures will be splited into 3 groups: season, trend and random factor. But is it enough to get to know what is influencing your sales ? Not really. How do you want to explain to your management board what is hidden beyond this random factor ? It might be crucial to know what is that...

Thus, I strongly recommend to build up econometrical model with as much variables as possible. There is a list with possible variables that you should include in modelling process:

1) Constant - obligatory - dummy variable.
2) Calendar effects: e.g. bank holidays, Christmas etc - dummy variable
3) Temperature - it could explain seasonality - quantitive / dummy variable
4) Out of stocks (OOS)- it could explain significant drop in sales history - quantitive/dummy variable
5) Stocking after OOS - it could explain picks after significant OOS - quantitive/dummy variable
6) Promotions -  it could explain picks in sales history, could be dependent on promo discount - quantitive / dummy variable
7) Drop after promotions - it could explain significant drop in sales history after stocking for promotion by customer - quantitive/dummy variable
8) Cannibalisation - dependancy of another sku's sales. Rather quantitve variable.
9) TV campaing - measured by e.g. GRP's. ammount. Rather quantitive variable.
10) Distribution - measured by ammount of POS (points of sales) that analysed product is delivered.
11) Price, price per KG/liter. It could measure price elasticity. Quantitve variable.
12) End of the month - very often practise in FMCG companies. Explanation of this variable will help you understand how efficient this mechanism is. Quite similar to 6th variable. Dummy variable.
13) Beginning of the month - explains effect of the sales drop after push in point 12. Dummy variable.

and many others, specify for particular segment. Analysis of examplary variables and expected effect is prerequisite to build up advanced econometrical model.

More content soon :)

## środa, 10 lutego 2016

### Forecast Accuracy - how to measure it correctly.

One of the most important KPI's measured by company is forecast accuracy (FA). If the forecast is correct (in given target), then:

a) logistics costs are being minimized i.e. production and storage costs.

b) CSL is high. Forecast accuracy is positively correlated with customer service level. The higher FA the higher CSL. And we know how customer's satisfaction is important...

but how to measure FA in the way that its result will be showing current situation at company and contributing to good decision making process ? Here is an example:

Apparently we have sold as much as we forecasted, meaning FA should be 100%. Indeed it is, great. But if we look at the details, per sku, total FA is...37%. So instead of sold as much as we wanted, many attention must be paid in order to clarify what happened.

## wtorek, 2 lutego 2016

### Logistics regression in business - key to success.

When (at which level) supporting customer with discounts becomes unprofitable?
When (at which level) TV campaings are not working, not reaching extra viewers?
etc.

Having right answer to those questions may contribute to huge savings at your company. To achieve so, simple linear function won't be enough. It's crucial to be acquainted with logistics regression, the so-called s-curve chart. It's purely non-linear function and I recommend to use more advanced software e.g. SAS, SPSS, R. I won't get into details right now, just show how useful this function is.

Based on historical spendings and sales level, it's visible that at ca. 7 unit of money spent function reaches asymptot - is not growing any more. Thus, company shouldn't spend more that ca. 7 unit of money in order to have the best effect of this market activity.

## niedziela, 24 stycznia 2016

### Trap of simple methods in forecasting - moving average, exponential smoothing etc.

Majority of companies use rather simple methods of forecasting, especially in short term. Moving average, exponential smoothing are among them. However, there are some traps that those methods may fall into. Those methods are not following trend represented by product/product group in the past, but simply extrapolate their own, recent history. As a result some gap may occur:

Therefore: always when making forecast based on past weeks/months doublecheck trend of the material (if it has history of course) and try to reflect it in demand. I assume that extrapolating exponential smoothing or moving average after 3 weeks of sales to whole e.g. year is rather not so popular but just in case - be prepared to have a gap (red field) which might be quite costly for your organisation.

## niedziela, 17 stycznia 2016

### Profitability vs. common sense

More and more companies cut their costs in order to keep the profitability either growing or stable due to some negative changes on the market (see the post regarding beer market).

It makes sense, however in some cases looking only at P&L i.e. accounting costs may be very dangerous. Especially in the long-time horizon. That's why managers should also consider opportunity costs. (more -> https://en.wikipedia.org/wiki/Opportunity_cost)

Case: should we cooperate with non-profitable* sales channels / customers ?

*very often not as much profitable as we want

It, of course, depends on many factors, including:

1) How big is the channel?
2) Are comepetitors present there?
3) What is the trend of the channel ?
4) What is the trend of our category in the channel ?
etc.

Even though lot of P&L calculations shows lower profitability of being present in the channel company should go for it. I haven't seen P&L showing opportunity costs. Many managers don't even consider question "what will happen in a long run when we won't be present there ?". What is crucial is here and now. If competitors are present, if channel is growing then it's very likely that customers will be slowly switching to their products. In short run, i.e. 6 months - 1 year it can be invisible in sales data. Sooner or later we can expect drop in market shares.

That's why being a good manager requires not only a good accounting skills.

P.S
Cooperating with low proftiable customer can be treated as a good marketing -instead of spending money on GRP's (see 1st post)

## sobota, 9 stycznia 2016

### Quick overview: beer market in Poland.

It won't be a cutting edge analysis, just couple of reflections regarding beer market in Poland. Data are taken from Euromonitor.

Scope: 2009-2014.

Remember when in 2008 "Ciechan Miodowy" was available only in 2-3 shops in the city. Now there are maybe 2-3 shops where it is not. What, this particular and amazing growth of local beers, means for global coorporations and product quality ?

Looking at the market shares (volume)

conclusion is rather obvious. Consumer trends weren't correcly forecasted. People are much more keen on trying new flavours, quality, they want to be identified with small, local producer and lager for masses is not longer a choice. Ciechan Miodowy was a turning point on the Polish market, it might be treated as an icon of the change. As of the end of 2014 trends were as follows:

Grupa Żywiec (Heineken Group) -14%
Kompania Piwowarska (SABMiller) -8%
Local beers&others +25%

I didn't get into details regarding Carslberg (24.01 - I did. Growth is determined by Harnaś, one of the cheapest SKU on the market). Their shares grew by 34%. Anyway, the decline of top brands is unprecedented. Companies with such a huge market shares should drive the trends, not being adjusted to them. Their products weren't diversified, and being honest today it seems that only Grupa Żywiec introduced SKU's that can breakdown the negative trend. That is first conclusion.

Secondly, I'm a bit afraid of product quality. With such a decline you have couple of decision to be made in order to keep profitability in target:

*increase the price - easiest
*downsize packaging with the same price (increase of price per kg) - more complicated
*fire people, shutdown the factory - easiest, for a long run
*change the product quality e.g. decrease the ammount of hop. - more complicated on short run but in a long - easiest. Plus, it's invisible to customer.

Cheers!