First of all, in order to understand in advance the effectiveness of the planned activities. You will no longer have to make decisions about choosing a particular strategy based only on common sense, experience and intuition.

A tool that you can use on your own, make different marketing scenarios into it and see what they will result in sales. We can provide you with a model in the form of a formula, an excel file with an interface, a dll / ocx library, or write a software product specifically for you.

In different ways. It took 2 hours to create the formula and optimize the coefficients for the anti-cold drug after collecting all the data and aggregating them. But with the model for the gas station, it took 2 weeks. In any case, knowing the category, we will provide you with the current timing.

The main limitation is that you can only accurately model what was previously there. For example, if you had only TV and Radio advertising, then the model will be able to predict with high accuracy only the effect of TV and Radio in the future, but not from the press, Internet or outdoor advertising. You also need a sufficient amount of data. It is not possible to build a qualitative model on a sample of only, for example, 10 dimensions.

Definitely not. But you can make assumptions. For example, see what was the contribution of what you want to use from competitors, apply to you by making adjustments. Or, for example, you want to advertise on TV, but you only had a Radio. In this case, you can take the dependence on Radio and assume that it will be better from TV due to the presence of an image, and add this new factor to the model. Of course, these are already assumptions that will have a slightly lower accuracy than the rest of the model.

Any that you have at your disposal. For example, for FMCG and pharma, this should traditionally be the sales of the brand and category, distribution, price, media activity of the brand and competitors, and the availability of promotions. It is good if there is additional Ad Recall data, if the communication message has changed over the previous years. And all this is monthly for 3 years, and, better, in general for 5 years. The more data, the more they are for a longer period of time – the more accurate the model is obtained.

It's okay, there are two ways:
1. We take a similar competitor brand that is closest in terms of positioning and price, build a model based on it, and adjust it based on common sense.
2. We go the way of "Knowledge-Purchase-Repeat purchase", based on the average category data.

Not always. This is often due to the lack of high-quality historical data in the form of a source. We always evaluate the quantity and quality of data before each project and will tell you if there is not enough data.

Pharmaceuticals, retail, gas stations, education, FMCG, medical equipment, cars.

We choose the type of models depending on the task. Most often, this is a multiplicative model. But it happens that this is a normal regression (as, for example, for a building materials store). If there is a lot of data, and we know for sure that the new inputs will not go beyond them, then we can use regression trees. We also use neural networks, but this is a last resort, since it is very difficult to explain the logic of their work and it is quite difficult to determine the optimal investment levels, although the prediction accuracy of neural networks is quite high. In general, multiplicative regressions are much easier to interpret, and it is easier to make optimization recommendations based on them.

Because in them all the parameters are set as a sum, which does not correspond to common sense. For example, you currently have 50% distribution. If it grows to 80%, then all other things being equal, sales should grow by about 50-60%. And, if it falls to zero, then you will have zero sales. But in additive models, the distribution (like price and all other factors) is given as a summand, and if it becomes zero, the sum of the other parameters will not give zero, which is incorrect. The same is true with the growth of distribution. Increasing it twice will not give an increase in sales in additive models and is close to the logical 90-100%.

We analyze how the buyer behaves, whether he makes repeated purchases, how often he does it, what factors affect the new and returned buyer, and so on. After that, we create a formula that would describe all this as closely as possible.

1. We collect data on sales, advertising activity, and so on.
2. We summarize them in a table and add them to the statistical software.
3. We begin to select formulas for the model, optimize the coefficients in them. We go from simple formulas to complex ones, introducing new variables.
4. Thus, we get to the point where entering new variables no longer improves the accuracy of the model.
We have reached our goal.

As practice has shown, if there is enough data, the model can be built, then our models are highly accurate and a lot of examples and tests have proved this.

For any of them. We build models for Russia as a whole, for individual regions, and for cities with millions of inhabitants, such as Moscow and St. Petersburg and even for foreign markets, for example, for the United States. In general, if you have a wide geography of sales, and it is not very important to make a forecast for each region separately, then it is better to make one model at the national level than to split it by city. Thus, the accuracy is higher due to the leveling of errors in data collection and forecasts.

We are confident, and practice shows that the model works very accurately if the new data is in the same range as in the simulation. For example, if you posted 200 to 600 TRP per month on TV, and you also post further, the response will be predicted very, very accurately. But if you want to place, say, 1200 TRP, then the forecast error will increase and we will take it into account in the coefficients. We guarantee high performance of our models. In addition, we develop the model-based strategy and tactics of the advertising campaign and provide financial guarantees for the execution of the pledged KPI's.

Yes, you do. Preferably quarterly. We do this for a nominal fee, because it is quite simple to optimize. If something extraordinary happens, for example, a new player enters the market with huge budgets for promotion, then in a few months it is advisable to model everything again, taking into account the new introductory ones, and we will help you do this.

The price depends heavily on the task, the amount of source data and the required forecast indicators. Please contact us, we will make the calculation. However, if we provide you with strategic and tactical media planning and advertising services, the simulation is free of charge.

We have extensive experience in communication and strategic planning of advertising and marketing, advertising placement, budget optimization, and analytics.