Discuss (0)

Print

RSS

Home » Salescast » Here

Understanding the pricing

Once the free trial period is over, Salescast requires a monthly subscription to keep working. We charge by the number of forecast data points processed by Salescast. On our pricing page you will find a pricing calculator that will let you simulate your expected monthly Salescast usage cost for classic forecasts. This section provides more details from a practical viewpoint.

Classic forecast pricing

Let's start with an example. Let's assume you have 200 products to be forecast 10 weeks ahead, twice per month, then the total number of forecast points is 200*10*2 = 4,000, which results in $38/month.

Then, Lokad's pricing is non-linear, and provides a major discount when the amount of forecasts increases. For example, if the number of monthly forecasts is 4,000,000 (compared to 4,000), the monthly subscription is only $3800/month, instead of $38000/month. The forecast volume has been multiplied by x1000 while the price has increased only by x100. In other words, the price per forecast has been divided by 10 with the volume growth.

As you start testing your own scenarios with our pricing calculator, you might end-up with ridiculously high subscription costs. We believe this situation is typically caused by a slight misunderstanding, the nature of which is lying in the rather unique technology used by Lokad.

Most classical forecasting systems rely on naive forecasting models (moving average, linear regression, exponential smoothing, etc.) and, as a result, recomputing the forecasts for say the next 52 weeks every day is not an issue because the computational processing power involved is extremely low anyway.

With Lokad, you get forecasts delivered in fully automated manner - all forecasting models being auto-configured by Lokad itself. This extra-accuracy comes with a drawback: the actual computation of the forecast requires a lot of computing resources.

In practice this means you won't want to recompute everything on a daily basis, and in particular, not to do it for some far distant future. For example, let's consider a pattern when a company wants to recompute forecasts on a daily basis for 52 weeks ahead for each product. This would generate about 52*30 = 1560 forecasts per product per month, which is quite a lot.

In fact, there is no reason to update the forecast for 52 weeks ahead on a daily basis. Indeed, one extra day of sales has little or no impact on the forecast that goes one year ahead. Instead, we typically suggest a weekly 4-week ahead forecasts, and once a month, to update the 52-week ahead forecasts. This way we have 4*4.3 + 52 = 70 forecasts per product per month, that is to say more than 20 times fewer forecasts than with the first method.

This constraint is inherent to any non-trivial forecasting method. In case you would be tempted by a Lokad's competitor who claims to deliver millions of forecasts for pennies, then be aware that you're going to end-up with a moving average variant. There is no free lunch: more accurate forecasts do require more intensive computations.

Quantile forecast pricing

In March 2012, Lokad has introduced quantile forecasts as a radically new and better way of computing very accurate reorder points. The pricing of quantile forecasts is similar to the one of classic forecasts except it comes with a twist: instead of charging per forecast, we charge per forecast value.

This pricing has been motivated by two extremes:

• Slow mover's: Items associated with reorder points lower than 3 were typically not worth to be forecast with our pricing for classic forecasts. With a lower price on slow mover's, our quantile pricing does not put an incentive to exclude slow mover's from the forecasting scope.
• High rotations: When financial stakes are high, it's worth trying to invest a lot more computing resources to improve the forecasting accuracy. With a higher price on high rotations, Lokad can invest more, use even more advance models, and deliver more accurate results as well.
  

Then again, our pricing is non-linear and provides a massive volume discount on high rotations. The cost of quantile forecasts will not skyrocket if your business grows.

The Lokad team is here to assist you in optimizing your consumption. Just send us an email.