Simulation Adjustment Examples

The following examples show what you can expect when creating events and how they are going to affect your forecasts.

No Simulation Adjustments

When you open Simulation Adjustments for the first time, you see collected data on a grey background for the past two weeks and simulated data for the coming four weeks. The simulated data is based on the average traffic over the previous two weeks for the same day. For example, the simulated traffic for Wednesday 21 September 2016 is based on the average traffic on Wednesday 7 September and Wednesday 14 September.

Future Simulation Adjustments

You can enter future events for an overall increase in traffic or increase in traffic for a specific category. The estimated effect is distributed equally over the dates in the defined time span and shows up as green bars in the graph. Any forecasts for the affected dates take both black and green parts of the bar into account, unless no sample traffic is available from which simulated traffic can be generated (see Considerations for an example).

Past Simulation Adjustments

You can enter past events for an unusual overall increase in traffic or an unusual increase in traffic for a specific category. The estimated effect is distributed equally over the dates in the defined time span and shows up as blue bars in the graph. The estimated effect in the past is then deducted before averaging for the simulated traffic. The estimated effect is seen on the simulated traffic as the striped black part. Any forecasts only take the black part of the bar into account, unless the deducted amount of the past event exceeds the amount of sampled traffic (see Considerations for an example).

An example in numbers shows exactly what happens in Simulation Adjustments. Your normal traffic is 1000000 for each day. A royal wedding occurred on 9 August and you had some special videos prepared as well. Due to this event you saw an increase in traffic of 500000 each day between 9 and 11 August.

Looking at simulated traffic without a past simulation adjustment, the simulated traffic for 16 to 18 August is 250000 more than the other days of the week and this pattern is repeated each week. The simulated traffic is calculated as the average of the sampled traffic from the past two weeks for the same days in the week, and then this pattern is repeated for all coming weeks:

If we put in the past simulation adjustment, then the data looks as follows:

Name Estimated effect Description Start date End date Category
Past Event 1500000 Royal wedding occurred. 09 Aug 2016 11 Aug 2016  

When we have entered a past simulation adjustment, the calculation of the simulated traffic takes this into account:

Name Estimated effect Description Start date End date Category
Past Event 1500000 Royal wedding occurred. 09 Aug 2016 11 Aug 2016  

Cumulative Simulation Adjustments

After entering future and past simulation adjustment that affect the same dates, you see their cumulative effect on the graph.

For example, a past event entered for 15 September says there was an extra 1600000 traffic. This affects the simulated traffic on 29 September. Then a future event was entered which adds 3000000 traffic between 27 and 29 September, which equates to 1000000 for each day. This also affects the simulated traffic for 29 September. The cumulative effect of the past and future event results in an added effect of 200000 traffic.

Recurring Event Simulation Adjustments

A new season of a hugely popular show is going to start in one week, airing live every Thursday evening for the next 6 weeks and made available on your on-demand channel on Fridays. You estimate that this show is going to generate 300000 extra traffic over the weekend, including Friday evening. You put these future events in as simulation adjustments, so you can start selling the expected extra ad inventory.

Name Estimated effect Description Start date End date Category
Future Event 300000 Episode 1 19 Aug 2016 21 Aug 2016
Future Event 300000 Episode 2 26 Aug 2016 28 Aug 2016
Future Event 300000 Episode 3 02 Sep 2016 04 Sep 2016
Future Event 300000 Episode 4 09 Sep 2016 11 Sep 2016
Future Event 300000 Episode 5 16 Sep 2016 18 Sep 2016
Future Event 300000 Episode 6 23 Sep 2016 25 Sep 2016

After one week, the first future event becomes a past event, so simulated traffic is adjusted to remove the first event's effect . The combination of the current sample traffic, which has increased as predicted, the past event and all future events, ensures that the simulated traffic continues to have the same expected increase over the weekend including Fridays. In this simplified example, we assume that your future adjustment was correct in total but that the distribution was a bit different than 100000 each day.

Name Estimated effect Description Start date End date Category
Past Event 300000 Episode 1 19 Aug 2016 21 Aug 2016
Future Event 300000 Episode 2 26 Aug 2016 28 Aug 2016
Future Event 300000 Episode 3 02 Sep 2016 04 Sep 2016
Future Event 300000 Episode 4 09 Sep 2016 11 Sep 2016
Future Event 300000 Episode 5 16 Sep 2016 18 Sep 2016
Future Event 300000 Episode 6 23 Sep 2016 25 Sep 2016

After two weeks, the next future event also turns into a past event, so simulated traffic is again adjusted accordingly. This past event together with the previous one now balance out each of the next future events completely, which explains why there are no longer Added Effect bars on the graph. However, the show's popularity has picked up even more, and your second estimate was lower than the actual increase in traffic (450000 extra traffic in total).

Name Estimated effect Description Start date End date Category
Past Event 300000 Episode 1 19 Aug 2016 21 Aug 2016
Past Event 300000 Episode 2 26 Aug 2016 28 Aug 2016
Future Event 300000 Episode 3 02 Sep 2016 04 Sep 2016
Future Event 300000 Episode 4 09 Sep 2016 11 Sep 2016
Future Event 300000 Episode 5 16 Sep 2016 18 Sep 2016
Future Event 300000 Episode 6 23 Sep 2016 25 Sep 2016

Because there was another increase in traffic after the second episode in this example, you should increase the past event between 26 and 28 August to 450000 and all subsequent future events also to 450000. If your actual traffic goes down, then compensate in the opposite direction. It is recommended to always do this, so your simulated traffic is closer to what is really going to happen.

Fast forward to after the last episode has aired and the following weekend has passed as well. The popularity increased significantly and the finale generated a lot more traffic. 900000 extra traffic was recorded over the last weekend including the Friday, so the last event was adjusted accordingly. You can now notice that the simulated traffic is getting compensated and has gone back (black bar) to about what the traffic was before the show started. When the next two weeks have past, all past events have no more influence over the simulated traffic, and you should see the normal trend of your traffic again without the need to adjust it.

Name Estimated effect Description Start date End date Category
Past Event 300000 Episode 1 19 Aug 2016 21 Aug 2016
Past Event 450000 Episode 2 26 Aug 2016 28 Aug 2016
Past Event 450000 Episode 3 02 Sep 2016 04 Sep 2016
Past Event 600000 Episode 4 09 Sep 2016 11 Sep 2016
Past Event 750000 Episode 5 16 Sep 2016 18 Sep 2016
Past Event 900000 Episode 6 23 Sep 2016 25 Sep 2016

Was this article helpful?