Examples of models
These econometric models were created by us for our clients
and they work to this day
and they work to this day
Brand №1:
Pharmaceuticals
National sales model for one of the market leaders in the category of anti-cold products
Pharmaceuticals
National sales model for one of the market leaders in the category of anti-cold products
We have created an econometric model of sales of anti-cold medicine. The model is successfully used to predict the effectiveness of marketing activities
Influencing factors: sales volume, category, price, distribution, advertising (TV, radio, press, digital, competitors), shipments
Influencing factors: sales volume, category, price, distribution, advertising (TV, radio, press, digital, competitors), shipments
Comparison of model and actual sales
Average monthly error: 8%
R2 (correlation coefficient): 0.99
Error overall for the year:
2012 - 6,7%
2013 - 0,9%
2014 - 0,9%
2015 - 4,0%
R2 (correlation coefficient): 0.99
Error overall for the year:
2012 - 6,7%
2013 - 0,9%
2014 - 0,9%
2015 - 4,0%
Decomposition into new and loyal customers
The largest share of sales is accounted for by new customers.
At the same time, loyalty to the drug is extremely low.
At the same time, loyalty to the drug is extremely low.
Decomposition into influencing factors
The decomposition into influencing factors clearly shows the high impact of advertising on sales, especially television advertising
The model also takes into account the contribution of the drug price and its distribution, but they cannot be displayed on the graph, because they are not added to the total, but affect each of the factors separately.
The model also takes into account the contribution of the drug price and its distribution, but they cannot be displayed on the graph, because they are not added to the total, but affect each of the factors separately.
Response curves to advertising impact
Response curves determine the optimal volume of placement in each media, which achieves the maximum result in sales without overinvestment
Brand №2:
Gas station network
Fuel spill forecast model
in Moscow and St. Petersburg
Gas station network
Fuel spill forecast model
in Moscow and St. Petersburg
We have created an econometric model of the spillage across the entire gas station network for two regions. To maximize the accuracy, we had to create 4 models (Moscow, MR, St. Petersburg, LR) and combine them by region.
Influencing factors: air temperature, season, price per 1 liter of fuel, advertising (TV, radio, press, outdoor advertising, competitors), promotional campaigns
Influencing factors: air temperature, season, price per 1 liter of fuel, advertising (TV, radio, press, outdoor advertising, competitors), promotional campaigns
tT – average air temperature
R – the difference in the price of gasoline between the average for the category
N – power network N=F(Q,L)
Q – number of filling stations
L – average Strait at each station for the entire period
S-1 – the Strait in the previous month,
a, s, k, - const – coefficients of the model
AdStock is the accumulated effect of advertising
Promo – the presence of promotions, 1 (present) or 0 (no)
R – the difference in the price of gasoline between the average for the category
N – power network N=F(Q,L)
Q – number of filling stations
L – average Strait at each station for the entire period
S-1 – the Strait in the previous month,
a, s, k, - const – coefficients of the model
AdStock is the accumulated effect of advertising
Promo – the presence of promotions, 1 (present) or 0 (no)
Comparison of model and actual sales
Moscow and the Moscow Region
Average monthly error: 4%
R2 (correlation coefficient): 0.84
Error overall for the year:
2014 - 1,3%
2015 - 1,5%
2016 - 3,1%
Average monthly error: 4%
R2 (correlation coefficient): 0.84
Error overall for the year:
2014 - 1,3%
2015 - 1,5%
2016 - 3,1%
Comparison of model and actual sales
St. Petersburg and the Leningrad Region
Average monthly error: 4%
R2 (correlation coefficient): 0.95
Error overall for the year:
2014 - 1,3%
2015 - 1,5%
2016 - 3,1%
Average monthly error: 4%
R2 (correlation coefficient): 0.95
Error overall for the year:
2014 - 1,3%
2015 - 1,5%
2016 - 3,1%
Interface in Microsoft Excel for working with the model:
Brand №3:
Online store of repair products
Sales model
Online store of repair products
Sales model
We modeled all sales, including those generated by onlineactivity, but excluding online promotion. This inaccuracy in the source data determined the large error of the model.
TV, OOH, Radio sales - advertising activity
Sales-1 - sales in the previous month
a, k, const - model coefficients
Sales-1 - sales in the previous month
a, k, const - model coefficients
Comparison of model and actual sales
Average monthly error: 9%
R2 (correlation coefficient): 0.89
Error overall for the year:
2015 - 4,3%
2016 - 5,4%
R2 (correlation coefficient): 0.89
Error overall for the year:
2015 - 4,3%
2016 - 5,4%
Decomposition into basic (including those who came 100% from the Internet), advertising and loyal customers
The largest share of sales is accounted for by new customers.
While loyalty is relatively low.
While loyalty is relatively low.
Increase your ROI to +30%
and stop testing your marketing plans blindly
Contact us
By clicking the "Submit" button, I give my consent to the processing of my personal data, in accordance with Federal Law №152-FZ of July 27, 2006 "On Personal Data", on the terms and for the purposes specified in the Consent to the processing of personal data
