1.12. Another Event-Response

1. Agency wants to run a campaign         CAMFAIGN REQUIREMENTS (IN)
                                    SUGGESTED CAMFAIGN (OUT)

Turn to Chapter 3.6 and look at the Piccadilly current physical model, specifically process 3.1 PLAN CAMPAIGN (Figure 3.6.5). It contains the data flows and stores that the system uses to respond to this event. Their definitions are in the data dictionary. The interview below with Dollis Hill describes the response to the event. Use the information from it and the current physical model to build your event-response models.

When you interviewed Stamford Brook, he mentioned that Dollis Hill is the sales executive assigned to the agency with a new campaign for liqueur chocolate bars. Let’s talk to Dollis and find out more about how she planned the campaign.

“I had a phone call from my contact at Totteridge & Whetstone. They’re the agency with the account for the liqueur chocolate bars. They want to advertise their product on television, and will spend £250,000 with Piccadilly. The remainder of the £1,000,000 budget is to be spent with some other television companies.

“During my phone conversation about the campaign, they said they want to advertise from the beginning of December to the end of February. This chocolate has an alcoholic filling, and so their target audience is adults. They also told me how many television rating points they want from the Piccadilly area. From the ratings, they will know how many adults have seen their commercials.

“The commercials they’re going to use sound really interesting. Each one is thirty seconds long.

“To begin, I allocated the campaign a unique number (this is for identification purposes), and recorded their requirements in my campaign file. Next, I planned a campaign for them. I looked at the breakchart to analyze the available time. For each break, I found out how many seconds were available for sale and at what moveability rate. Then I identified which times would be suitable by looking at the predicted ratings on the programme transmission schedule. Remember, I want breaks close to the programmes that have the highest predicted ratings for adults, those most likely to buy the new chocolate bar.

“When I had identified the suitable amount of time, I used the ratecard to price each spot. The price for a spot depends on the moveability rate.

“Next, I went back to the campaign requirements the agency had given me, and compared the priced time against their requirements. I finally selected a mixture of spots that I thought would satisfy the media buyer at Totteridge & Whetstone, and I recorded my suggested campaign in my file. Then I phoned the agency and told my contact my suggestions for the campaign. She’ll call back this afternoon with the decision.”

If you need any more background to build the models, re-read Stamford Brook’s statement in Chapter 1.6 Selling the Airtime.

The first step is to model the current system’s response to the event Agency wants to run a campaign. Your model should isolate all the processes that can be connected to the response. Make sure that each process has all the data that it needs to do its job. Show all the appropriate current physical data stores on this model.

Your first-step model is intended to reflect the way the business is currently being done. However, you know there are some features that are present in the model because of the existing implementation. If you can easily recognize them, you may choose to leave them out of your model. If you are undecided whether a process is essential or implementation dependent, then play it safe and include it in your model. The second step will take care of it.

When you are happy with your physical event-response model, compare it with the sample in Chapter 3.11 (Figure 3.11.1). The discussion there will give you some feedback on the first step before you plunge into the essential waters in the next step.

The second step in event modeling is to refine your model so that it reflects the essential view of the system. This view excludes anything to do with the implementation, as implementation-dependent features are not part of and often obscure the essential policy.

Finally, the model is refined by removing any processing or data that exist because of the implementation, and replacing the physical data stores with their essential equivalents. There are several ways to do this. Perhaps the easiest way to refine your model is to start by replacing physical data stores with the appropriate essential data entities and relationships. The first-cut data model that you have already built will help you identify them.

Look at the data dictionary definitions of the incoming and outgoing data flows for this event-response. For each data element in a flow, ask, “Which entity is described by this element?” You will find the entities in the first-cut data model, or, in some cases, you may need to create a new entity if no suitable one exists. Make each entity a data store in your event-response model in place of the existing physical files.

Alternatively, you could look at the definitions of the physical files, and determine which entities can be combined to make the logical equivalent of each file. Draw those entities on your model. When you have replaced the physical files, relate the entities that you have used to each other to make the event-response data model.

Another alternative is to build an event-response data model first. Remember that the event-response data model is a collection of stored data that is private to the particular event-response. The data for the data model come from the data flows that are input to and output from the event-response. Use the entities and relationships to replace the physical files.

Another approach that we should mention is instead of building a detailed model of the current event-response, you build a model showing only one bubble and all the data flows and physical data stores. Then, instead of refining the model by eliminating implementation-dependent processes, you refine it by replacing the physical data stores with the essential data stores that correspond to the appropriate entities. If the single bubble, known as an essential activity, is too complex for one mini specification, you level downward to decompose the single bubble into essential processes that are small enough to specify in detail. Once you have captured the event-response’s private data by building the event-response data model, then the essential processes are those that store or retrieve these data.

Whichever method you choose, annotate the event-response data model with the CRUD operators. In other words, show which entities and relationships are created, referenced, updated, or deleted by this event-response. Doing so will help you later on when you are ready to correlate all of your event-response models.

If you want to break the task into smaller stages, check your progress frequently against the models in Chapter 3.11. Do this exercise in the way that is most comfortable to you; but whatever you do, don’t wait until you have finished everything before checking. You must ensure you are developing the models along the right lines before you get too far.

Your final concerns at this stage include entering any new or altered data items in the data dictionary and writing mini specifications for each of the processes in the essential event-response process model. Remember, the work that you’ve done already on your event-response data model will simplify the task of writing your mini specifications.