What is the objective?
The onboarding process is a preliminary step in the Datup experience, where we provide you with all the necessary information about the implemented methodology to strengthen your knowledge and facilitate the process of understanding the results cubes when they are delivered. This article was developed to enable you to leverage, enhance and improve your demand planning processes.
We recommend the following reading order:
- How-to: Interpret the result cubes.
- How-to: Using Datup forecasts.
How-to: Interpreting Results Cubes
Below, you will find a detailed reference of the cubes and the fields that compose them, in order to facilitate users' interpretation and use in the continuous improvement of demand planning processes, in terms of accuracy and automation.
Qrank: Item Sorting Cube
It consolidates the classification of each item according to the ABC, FSN and XYZ rankings. The first refers to the profitability or revenue of the item; the second to the speed or movements in sales and the third to the stability of the item for a forecasting process.
Name Column | Description | Interpretation |
|---|---|---|
Item | Unique identification of the items or references making up the product portfolio | |
Target Revenue | Sum of income or profitability related to the item | |
RevenuePercent | Percentage of income or profitability related to the item | Values between 0 and 1 (0-100%). The higher the value, the higher the income and profitability of the item. |
| ABC | ABC classification of the item | Items A represent up to 80% of revenues or returns. Items B and C represent 15% and 5%, respectively. |
| Target Frequency | Percentage of periods (dates) with sales or demand movements of the item | Values between 0 and 1 (0-100%). The higher the value - the greater the movement of the item |
| FSN | FSN classification of the item | F items have movements at least 75% of the periods analysed. Items S and N have maximum 50% and 25% of dates with activity, respectively. |
Stability | Index of variation of historical sales and demand with the stability of the forecast item | Values from 0. The higher the value, the more difficult the forecast. |
| XYZ | XYZ classification of the item | Items X are of high stability and therefore favours prediction. Items Y and Z are of medium and low stability. |
| Ranking | Union of ABC, FSN and XYZ classifications, according to the criteria of income/profitability - movement and stability. | The items of highest value to the business are AFX, AFY and AFZ. The items of lowest value to the business are CNX, CNY and CNZ. |
| Other variables | Additional attributes of the item. For example description, category, line, brand, subcategory, etc. | Allow filtering or aggregation by criteria higher in the item hierarchy |
We leave you the following video to reinforce what you have learned:
Qfcst: Forecast Cube
Gather forecasts and backtests (evaluations of historical data against model predictions) for each item and dates in the forecast period, suggesting the most likely sales or demand quantities for the operation, manufacturing, supply or S&OP processes.
| Name Column | Description | Interpretation |
|---|---|---|
| Date | Historical date and forecast. | |
| Week | Number of the week of the year. | |
| Item | Unique identification of the items or references that make up the product portfolio. | |
| Target | Actual observation or actual historical behaviour of a particular item for a specific date. | |
| SuggestedForecast | Number of units forecast for a particular item and specific date. | This value should be taken as the most likely amount of sales or demand to be observed in the forecast period or date. |
| SuggestedInterval | Forecast interval associated with the number of units forecast. | Allows you to identify whether the most likely quantity of sales or demand is at your point forecast or closer to an under-forecast (Up60, Up80, Up95) or over-forecast (Lo60, Lo80, Lo95). |
| NextSuggestedForecast | Number of over-predicted units for a particular item and specific date. | In S&OP sessions it allows to propose a higher most likely forecast value - by considering the SuggestedForecast to be conservative. |
| NextSuggestedInterval | Forecast interval associated with the number of over-predicted units | It identifies whether the most likely amount of sales or demand, above, is closer to an under-forecast (Up60- Up80- Up95) or over-forecast (Lo60- Lo80- Lo95). |
| BackSuggestedForecast | Number of units under-predicted for a particular item and specific date. | In S&OP sessions, it allows proposing a lower most likely forecast value by considering the SuggestedForecast as optimistic. |
| BackSuggestedInterval | Forecast interval associated with the number of under-predicted units. | Allows to identify whether the most likely quantity of sales or demand - below - is closer to an under-forecast (Up60, Up80, Up95) or over-forecast (Lo60, Lo80, Lo95). |
ForecastNaive | Number of units predicted, according to the naive forecast. That is, the forecast is equal to the last actual observation. | Backup forecast for those items and dates with errors greater than the control value (e.g. 50%) and the comparative error measure MASE is greater than 1. |
ForecastPoint | Number of predicted units estimated by the model for the point interval. | Central forecast value generated by the model for a specific item and date. The forecast has a 50% probability of being located at the Forecast Point. |
ForecastLo95 | Number of predicted units estimated by the model for the Lo95 interval. | Lowest possible forecast value generated by the model for a specific item or date. The forecast has a 95% probability of lying between Lo95 and the Forecast Point. |
ForecastLo80 | Number of predicted units estimated by the model for the interval Lo80. | Second lowest possible forecast value generated by the model for a specific item or date. The forecast has an 80% probability of falling between the Lo80 and the Forecast Point. |
ForecastLo60 | Number of predicted units estimated by the model for the interval Lo60. | Third lowest possible forecast value generated by the model for a specific item or date. The forecast has a 60% probability of falling between the Lo60 and the Forecast Point. |
ForecastUp60 | Number of predicted units estimated by the model for the interval Up60. | Third highest possible forecast value generated by the model for a specific item or date. The forecast has a 60% probability of falling between the Forecast Point and Up60. |
ForecastUp80 | Number of predicted units estimated by the model for the interval Up80. | Second highest possible forecast value generated by the model for a specific item or date. The forecast has an 80% probability of falling between the Forecast Point and Up80. |
ForecastUp95 | Number of predicted units estimated by the model for the interval Up95. | Highest possible forecast value generated by the model for a specific item or date. The forecast has a 95% probability of lying between the Forecast Point and Up95. |
WMAPE | Measure of forecast error, Weighted Mean Absolute Percentage Error. | It allows to evaluate the prediction performance of the model for each item. It is usually evaluated in control bands: 0-30% (excellent), 30-50% (acceptable), +50% (poor). |
MASE | Measure of comparative forecast error, Mean Absolute Scaled Error. | Allow to evaluate the prediction performance of the model for each item by comparing the error of the proposed model against the error of the naive forecast (Forecast Naive). Errors less than 1 show that the proposed model performs better. |
sMAPE | Measure of forecast error, Symmetric Mean Absolute Percentage Error | It allows to evaluate the prediction performance of the model for each item. It is usually evaluated in control bands: 0-30% (excellent), 30-50% (acceptable), +50% (poor). It is advisable to use weighted error measures (e.g. WMAPE). |
Ranking | Union of ABC, FSN and XYZ classifications, according to the criteria of income/profitability, turnover and stability. | The items of highest value to the business are AFX, AFY and AFZ. The items of lowest value to the business are CNX, CNY and CNZ. |
The forecast intervals are the forecast impact zones, measured as the deviation of the forecast from the actual observation. These zones are delimited by: Forecast Point, Forecast Lo95, Forecast Lo80, Forecast Lo60, Forecast Up60, Forecast Up80 and Forecast Up95.
We leave you the following video to reinforce what you have learned:
Qstrgy: Strategic Forecasting Cube
It summarises the most relevant forecast results for each item to guide the continuous improvement processes of demand planning. Firstly, it allows to identify the highest performing items, the items above a control error value (e.g. 50%), the best performing items by a naïve forecast or proposed model, and the forecast interval that best describes a recurrent under- or over-forecast.
| Column Name | Description | Interpretation |
|---|---|---|
Item | Unique identification of the items or references making up the product portfolio | |
| Ranking | Union of ABC, FSN and XYZ classifications, according to the criteria of income/profitability - movement and stability. | The items of highest value to the business are AFX, AFY and AFZ. The items of lowest value to the business are CNX, CNY and CNZ. |
| WMAPE | Weighted Mean Absolute Percentage Error - Weighted Mean Absolute Percentage Error | It allows to evaluate the prediction performance of the model for each item. It is usually evaluated in control bands: 0-30% (excellent), 30-50% (acceptable), +50% (poor). |
| MASE | Measure of Comparative Forecast Error - Mean Absolute Scaled Error | Allow to evaluate the prediction performance of the model for each item by comparing the error of the proposed model against the error of the naive forecast (Forecast Naive). Errors less than 1 show that the proposed model performs better. |
| SuggestedInterval | Forecast interval associated with the number of forecast units | Allows you to identify whether the most likely quantity of sales or demand is at your point forecast or closer to an under-forecast (Up60, Up80, Up95) or over-forecast (Lo60, Lo80, Lo95). |
| Other variables | Additional attributes of the item. For example description, category, line, brand, subcategory, etc. | Allow filtering or aggregation by criteria higher in the item hierarchy |
We leave you the following video to reinforce what you have learned:
How-To: Uses of Datup Forecasts
In the following, we demonstrate how to get the most out of Datup forecasts to increase accuracy levels and decrease lead times in demand planning. To do this, the basic, intermediate and advanced journey and usage scenarios will be described.
These uses will allow the analyst to take into consideration various fields of the results cubes, which will better guide decision making with respect to sales, demand, supply or similar behaviour.
Basic Use: Suggested Forecasts
This use guides better decision making for:
Know the volume of demand required by customers for each item for one or more future months.
Propose in S&OP sessions demand volumes according to optimistic, conservative or pessimistic sales conditions.
Assess the certainty or uncertainty of demand forecasts.
Prioritise portfolio forecasting by top-selling items, turnover and demand variation.
This scenario focuses on suggested forecasts, automatically generated and selected by the platform. First, you must locate the date and the item of interest, by means of the columns Date e Itemrespectively. The column SuggestedForecast shows the suggested forecast for the date and item under analysis, based on historical demand behaviour (Target); considering seasonality, levels, trends and even relationships with other items in the portfolio.
In addition, the column NextSuggestedForecastwhich presents an alternative scenario to the suggested forecast when there are indications or a substantial increase in demand for the item is anticipated for the specific period. The NextSuggestedForecast is often very useful for S&OP sessions, where the trading counterpart often arranges higher values for forecasts of certain items, subject to marketing campaigns or higher consumption strategies. Instead of opening the space for speculation for such a possible higher level of demand, the platform offers an alternative value that contemplates such an over-demand scenario. On the other hand, the solution also presents an alternative value for the under-demand scenario, where the trading partner foresees a discouragement in the consumption of the item for future periods. The field BackSuggestedForecast contains the amount of demand to be considered.
The countryside WMAPEThe error expected for the forecast period, based on the errors obtained by testing the backtestsThis is to simulate the demand forecast for previous periods, where the actual behaviour is known, hence it is possible to measure performance. By default, Datup performs 5 of these simulations, or backtests in each iteration of the models.
Finally, it is suggested to take into account the column RankingThis allows us to determine the items with the highest consumption, turnover and stability in demand. In other words, it identifies the items of highest and lowest value for the business.
Intermediate Use: Suggested Intervals
This use guides better decision making for:
All the advantages of the basic journey.
Know which items are recurrently over-predicted with respect to their historical demand to prevent excessive inventory losses.
Know which items are recurrently under-predicted with respect to their historical demand to prevent out-of-stock losses or stock-outs.
The second scenario accompanies the suggested forecasts with suggested ranges, allowing the demand analyst to know which items have been consistently under- or over-forecast in the past. The column SuggestedInterval indicates the forecast interval associated with the suggested forecast in column SuggestedForecast. Accordingly, the suggested upside forecasts (NextSuggestedForecast) and downward (BackSuggestedForecast) are also accompanied by their suggested intervals NextSuggestedIntervaly BackSuggestedIntervalrespectively.
The forecast intervals SuggestedInterval, NextSuggestedInterval y BackSuggestedIntervalare not only calculated for the periods to be forecast, but also for each forecast period. backtest or simulation. It is precisely in this way that it is possible to determine what is the forecast interval, and hence the most likely suggested forecast, based on its repetition in the observed demand history.
As a practical matter, the demand analyst or planner can determine whether an item is susceptible to over-demand if the suggested range SuggestedInterval has one of the following values: Lo95, Lo80 o Lo60. On the other hand, the item is susceptible to under-demand if the SuggestedInterval presents either Up60, Up80 o Up95. In both cases, the under- or over-demand is established with respect to the point forecast. ForecastPoint which lies in the middle of all forecast intervals.
Using forecast intervals allows the demand planning process to anticipate and avoid overproduction or oversupply events, as well as stock-outs or stock-outs.
Advanced Use: Naive Forecasting and Forecast Intervals
This use guides better decision making for:
All the advantages of the intermediate journey.
Examine naïve forecasts, i.e. averages or last observations, to plan demand on items with high variation or intermittency and increase their accuracy.
Examine the suggested forecasts for each item and period to see how high or low demand may behave in different market circumstances.
Evaluate the MASE error to determine in which cases it is advisable to use naïve forecasts to increase planning accuracy.
The third and final scenario involves the naïve forecast and all the associated forecast intervals estimated by Datup. This collection of values allows the analyst to know all possible scenarios of high and low forecasted demand.
The forecasts point ForecastPointlower 95 ForecastLo95, lower 80 ForecastLo80, lower 60ForecastLo60, higher 60 ForecastUp60, higher than 80 ForecastUp80 and above 95 ForecastUp95 The total number of points Datup estimates for each date and item to consider the different scenarios of high, medium and low demand. The algorithm automatically selects the scenario, and in turn the most likely value. However, the cube results are presented in an open way to the analyst to enable possible additional exercises.
For its part, the naïve prognosis in the columnForecastNaive calculates the forecast value by means of a rolling window average. The size of the window coincides with the number of periods to forecast. For example if you want to forecast 4 weeks, the rolling window calculates the naive forecast by taking the average of 4 weeks at a time from history. Datup compares the performance of its model against the naïve forecast, as there are items whose stability in the demand history suffers from high intermittency and/or random sharing that prevent their forecasting through traditional or advanced statistical methods. The value of the MASE shows the result of this comparison. Values close to 1 (above or below) confirm the superior performance of the Datup model, while values well above 1 favour the use of the naive forecast. ForecastNaive. If this is the scenario, in the bucket, the naïve forecast will be the value suggested as the forecast for the period evaluated.