You are the CMO of Bearly Awake, an innovative watch distributor that carries an exclusive set of avantgarde watch brands . Bearly Awake runs a state-of-the-art Actual vs Plan reporting (to run the analysis, type "Watch retailer " in the app).
Multi-tier bar chart
See how a dimension - here complication - plays out in the two periods on a metric - here amount.
Bearly Awake sells four lines of watch complications. Chronographs are the largest segment but are lagging vs plan. Smart Watches, the smallest segment, are growing double digit.
See how a metric, here revenue, interacts along a pair of dimensions, here country and complication.
Breakdown by complication is similar by country.
See how two metrics - here sales and margins - play along a given dimension, here complication.
Margin rate is very different by complication.
GROSS MARGIN BY COMPLICATION
In terms of margin, Smart Watches are beating the plan, while Chronographs are doing badly due to higher than expected unit costs,...
...in all choices of materials.
Chronograph sales just lagged plan in terms of sales, but, due to higher than expected unit costs, delivered an unexpectedly large negative margin.
Chronographs represent less than half the revenues but 85% of the costs.
The breakdown of complications by material is fairly similar.
Margin rate is fairly similar by material.
AMOUNT BY MATERIAL
Steel-gold watches, the most sold material, topped the plan in sales,...
...but underdelivered in terms of margin...
...due to a significant increase in COGS...
..and a reduction in average unit prices.
The breakdown of sales by brand is similar across countries.
Margin rate is aligned across the top two brands, lower for the others.
See how a dimension - here brand - plays out in the two periods on a set of metrics - here amount, COGS and gross margin
The three smaller brands underdelivered in terms of margins: Ward due to lower volumes, Auriol and Marvin due to higher unit costs.
BY BRAND DOT PLOT
Margin rate is across most countries,...
...with all main countries experiencing a similar pattern of COGS increasing more than sales, leading to a fall in margin.
The breakdown of sales by channel is similar across countries.
...as is the breakdown by customer type.
The choice of material seems tied to local preferences.
Taking the three most important markets (Belgium, Turkey and Switzerland) as an example:
- Belgium has one material (Steel-Gold) in common with both Turkey and Switzerland, and one "exclusive" material (Carbon) that is sold neither in Turkey nor in Switzerland
- Turkey has one material (Steel-Gold) in common with both Turkey and Switzerland, and two "exclusive" materials (Gold and Steel) that are sold neither in Belgium nor in Switzerland
- Switzerland has one material (Steel-Gold) in common with both Belgium and Switzerland, and two "exclusive" materials (Bronze and Ceramic) that are sold neither in Belgium nor in Turkey.
Repeat customer segment has a higher margin rate,...
...with all customer types tied in a similar pattern of COGS increasing more than sales, leading to a fall in margin.
Product lines are also reasonably aligned in terms of margin rate,...
....with women watches on plan in terms of sales but considerably less profitable than planned.
Sales variance analysis
See aggregated sales variance or split variance between its components (volume, price, mix, drivers,..).
Actual sales are 5% over plan with some issues of price integrity, given the impact of some price reductions,...
Waterfall chart format
Whenever possible, charts are built in accordance to the IBCS standard. Green is "good", red is "bad", white is "plan", grey is "previous period" and black is "actual".
...and of a possibly deteriorating sales mix.
Steel-gold watches had a 20% reduction of average price...
...and show a considerably more skewed to the left pricing distribution compared to plan,...
...across all channels.
This is due to a drop in prices and to the impact of customers buying a less expensive "mix" of steel-gold watches.
The negative mix change hit all complication types of steel-gold watches...
...showing a significant change in their price distribution profile.
Margin variance analysis
See aggregated margin variance or variance split by component (volume, price, cost, mix,...).
Margins are 10% below plan due to higher than anticipated unit costs and lower than anticipated pricing.
Bearly Awake's margin rate is below plan, impacting margins negatively by 9M.
WHAT CHANGED IN SALES
FIRST RESULT COMBINATION
Variable dimension variance - Sales
Variable dimension variance helps see the interaction between the change in sales across the different dimensions. This first waterfall shows three facts we have already noticed - the growth of Smart Watches, the fall of Ward and Magrette brand watches, and the growth of Steel-Gold pieces - in one consistent picture.
Smart watch complications and steel-gold pieces drove growth. Ward and Magrette brands disappointed.
The first row result shows Smart Watches up 9M vs plan.
The third row result shows Magrette down 3.8M.
Every dataset is different
This use case is based on a fake dataset. Test run the app with your data to confirm the advantages of the variable dimension variance approach in your specific use case.
At the aggregate level (aka including Smart Watches), Magrette sales were barely (-1%) under plan. However, Magrette's performance is very different when split by complication: Smart Watches are up 3.5M (included in the 9M first row result) while the other complications are down by 3.8M (shown in the third row result). Variable dimension variance picks this up.
The fourth row result isolates the 4M increase in Steel-Gold sales, net of Smart Watches and net of Ward and Magrette sales.
How to filter
Set the first filter on the "complication" column and "exclude" Smart Watch. Set the second filter on the "brand" column and "exclude" Ward and Megrette. Set the third filter to keep only Steel-gold. Or simply set the "data to plot" widget to "row 4".
Plotting this slice of the dataset by dimension shows that something big happened at the Auriol and Marvin brand level.
Show small multiple plots
After setting the appropriate filters set the "Run" widget to "Variable dimension bridge". Set the "Choose dimension for small multiples" widget to brand. Hit the 🚀 Submit button.
Slicing the 4M increase by brand reveals two Auriol and Marvin sub-segments that are doing very well and that merit to be investigated further.
You want to get the combination that grew the most. Set the "Let me fine tune parameters" widget to True, and then set the "Variance amount weight" widget to 0.20.
Alternatively, you might want to drill down on the fourth row result of the first waterfall and apply variable dimension variance on that slice of data.
After a bit of parameter fine-tuning you get this arguably even more insightful view: Steel-Gold Auriol watches with moon phases did particularly well (+700%). A Pelton reference, the article 8-2, was wildly successful.
The fifth row result shows an impressive performance (+322%) by the Marvin brand, even net of our top performing Smart Watch and Steel-Gold sales. Marvin is Bearly Awake's brand that has best performed versus plan (+385%).
Even net of our top performing Smart Watch and Steel-Gold sales, everything is pointing North for the Marvin brand.
No patterns to uncover
In this subset of data, changing dimensions does not give further insights as opposed to simply sticking to a given dimension.
In this case changing dimension seems to end up just showing an item in a different dimension that also has a 300-something growth rate.
Variable dimension analysis overkill
If all the dimensions have the same growth rate, variable dimension variance might struggle to deliver additional information compared to a simple slice and dice. And it is cognitively more challenging.
The chart above - with variable dimensions for every row result - is not necessarily more insightful than the chart below. In this case changing dimensions is not a useful exercise.
For each set of parameters, the app can generate up to ten alternative sets of results. To see a different alternative set of results use the "Return alternative results" widget.
Nevertheless after a few tries an interesting alternative combination with a much higher growth rate - chronograph men 600% - emerges from the app.
SECOND RESULT COMBINATION
See another angle
This second waterfall shows that Steel-Gold and Bronze pieces did fine, while Chronographs did not.
Steel-gold is driving sales (+11%), while chronographs and standard complications are not doing so good.
In this alternative set of results, Steel-Gold "comes first" (instead of coming "after" Smart Watches) and is 11% over plan. The 6.5M, 11%, increase in Steel-Gold is certainly less impressive that the 3.5M, 83%, increase in Smart Watches of the first result set.
On the plus side, however, the result surfaces a 2.9M, 12%, decrease of "non Steel-Gold" Chronographs, that is more than double the -5% overall unfiltered loss of Chronographs sales shown below.
Drill down 1
Drill down on the first "Steel-Gold" result row and see its elements.
Most of the gain in steel-gold watches comes from watches fitted with smart watch complications, that more than doubled sales.
Drill down 2
Drill down on the second "Chronograph" result and see its elements.
Magrette and Ward branded chronographs lost substantial ground, while Marvin and Auriol chronographs were well appreciated by consumers.
Consistency of results
Alternative result sets are always consistent. The different sets of results combinations show pretty much the same elements from slightly different angles, potentially helping to unlock insights.
Plotting result row data
By default the app plots the data of the entire dataset. It can also plot the data of a single result row - here the third, positive, "Bronze" result row - to help filter out patterns.
This set of results gives relevance to the good performance of Bronze pieces (result row three), and to the problems of Standard watches (result row four).
Interestingly, while only "green" markers appear when plotting the result row three data (Bronze Material)...
...and only "red" markers appear when plotting the result row four data (Standard Complication),...
...variable dimension analysis drilldown surfaces "counter trending" elements in both cases.
Split sales variance
This waterfall shows variance split between its volume & mix and its price component.
THIRD RESULT COMBINATION
Steel-gold watches grew in sales, but had a significant negative price variance.
Breaking up the second - "steel-gold price - result row shows the negative price variance especially touches chronographs and men watches...
...mainly in Auriol, Ward and Marvin brands, that grew significantly in sales but sacrificed prices...
...by over a third.
The first row result above shows the positive 13.7M volume component of "material = Smart-Gold" variance. The app switches dimension for the "brand = Magrette" forth row result.
The standard "fixed dimension" variance calculation sliced by material with separate price variance and volume variance entries returns this chart. Note that rows 1 to 3 are identical to the variable dimension plot.
After five results the standard approach leaves us with 5.2M of "unexplained" variance balance. The balance with variable dimension variance is 1.2M less. This is generally, but not always, the case. Variable dimension variance arguably also provides additional insight by pointing to the Magrette volume loss.
Another important advantage of variable dimension variance is the possibility of drilling down on a given variance component - for instance price variance of Steel-Gold pieces - to try to identify the root causes.
See all variance components of a combination
To see the non-filtered initial value of all variance components of a given combination, filter on the combination and run fix dimension variance. The example chart on the left shows price and volume variance for Bronze.
A disadvantage of variable dimension variance is that it only shows the top-ranked components of each combination. In our example both volume and price variance are shown for Steel-Gold, but only volume variance is shown for Bronze and Magrette, and only price variance is shown for Gold.
Variable dimension variance - Margin
Variable dimension variance helps see the interaction between the change in margins across the different dimensions.
WHAT CHANGED IN MARGINS
FORTH RESULT COMBINATION
Steel-gold chronographs are the major culprits of the loss of margins vs plan...
See another angle
This second margin waterfall shows that steel-gold smart watches did fine, while the other steel-gold pieces did not.
...while steel-gold smart watches returned increasing margins.
While it is immediately clear that Steel-Gold did not do well in terms of margins while Smart-Watches overperformed,...
...variable dimension variance "isolates" the negative performance of Steel-Gold Chronographs from the positive performance of Steel-Gold Smart Watches.
Split margin variance
This waterfall shows how Steel-Gold margin variance splits between its volume, price and unit cost components.
FIFTH RESULT COMBINATION
For steel-gold watches the increase of margins due to higher volumes did not suffice to offset the reduction of margins due to lower prices.
The negative impact on margins of the 4M higher unit cost of Chronographs was disproportionally tied to woman watches.
The first row result of the fifth waterfall shows the negative 7.2M price component of "material = Smart-Gold" margin variance. The app switches dimension for the "complication = Chronograph" third row result showing a negative 4M unit cost component.
The standard "fixed dimension" variance calculation sliced by material with separate price, volume and cost variance on margin entries returns this chart. Note that rows 1 to 2 are identical to the variable dimension plot.
After five results, the standard approach leaves us with -6.4M of "unexplained" variance balance. The balance with variable dimension variance is 2.1M less. This is generally, but not always, the case. Variable dimension variance arguably also provides additional insight by pointing to the Chronograph unit cost loss.
Another important advantage of variable dimension variance is the possibility of drilling down on a given variance component - for instance cost variance of Chronograph pieces - to try to identify the root causes.
SIXTH RESULT COMBINATION
Finding patterns in data
'Mparanza uses variable dimension variance to help find patterns in the data that could be missed with a traditional slice and dice approach.
Poor performance in terms of margin of chronographs was due to higher unit costs and lower volumes. Steel-gold watch prices suffered due to lower prices.
Switching the first result of the sixth waterfall from "material=steel-gold" to "complication=smart-watch" surfaces a negative volume variance of chronographs that was not immediately evident in the previous analysis since it was hidden in the negative variance of steel-gold pieces.