Today we cover what is the fairest data of all.
We need to set this in context first. The key question to ask is does the datum impact any of the business drivers in question. If not, who care?
Now that you have isolated the umpteen odd data elements that impact your business, how to validate the data. This is important because any analysis you make or decisions you take will depend on how good and relevant this data is.
I suggest you follow the following steps:
1) Is the data there?
If it is missing, that data will not help analysis. Here one must be careful to distinguish between "no data" versus "missing data".
2) How to populate missing data?
If data is missing, how to get it? Logic suggests that one asks a Customer. Organizations often have an illogical fear of talking to Customers. We will discuss this separately under the context of Customer Service.
3) Is the data correct?
This brings us to data that is there. If there is any doubt that the data correct, lets ask the customer and validate it.
4) Is the data current?
Any good CRM or Database should maintain last time data regarding this customer was updated and by whom. Some data changes over a short duration and others change over a long duration. Reviewing the date will tell us if we need to revisit and get this updated.
5) Is the data standardized?
For good analysis data should be standardized. In some future blog, I will cover my experience in the use of reference tables and such things.
Follow this for your most valuable segment of customers first, implement programs that help and attract them to you and you will have a better ability to weather economic uncertainties. Of course, you will also need to figure out how to keep them with you during these time.
Oddly enough, this will happen when Customers see that we treat their data with care, keep it current and manage it like a precious resource.
I would not do business with organizations who seem to sell your data to any Tom, Dick or Harry. Will you?
Thursday, September 24, 2009
Tuesday, September 22, 2009
There are Responses and then there are Responses
Today, we will cover something simple.
This thing has been bugging me since we started assigning responses to our promos.
It starts with the realization that one needs to track responses as uniquely as possible and assign these responses to the specific promo to measure effectiveness.
That is well and good.
So we assign Unique Identifiers to the records and unique Source Codes and Cell Codes to make that data as granular as possible. We hope that this information is printed or attached somewhere so it comes back to us with the same integrity with which it left.
When the response comes back we hit several brick walls that prevent us from assigning the response flags effectively.
1) The unique codes were conveniently ignored at the vendor fulfilling the campaign and now we are left to match names and addresses and emails to figure out who actually responded.
2) Data Entry folks are not rewarded for correct entry than quick entry so a significant portion of the response now has to be again matched back using steps described in 1.
3) Finally, the offer to A was forwarded to B and B used it. Do we credit A for the response or B and if we wish to credit B how do we we did not send any promo to B?
I have talked about this the last 15 years to dozens of Clients and to organizations where I owned processes and databases. But I failed utterly to change any process because of the number of seemingly imponderables.
The most significant one, it seems, is an organizations inability to change their existing process to add the three new pieces of data to the output files that would drive fulfillment processes at third party locations.
The next significant is the organizations inability to change a manual process to a automated process based on scanning labels containing this data which would clearly be more accurate and fast.
The last one is an organizations inability to see response beyond the single dimension of promo to response. I had to by most part ignored any response where I could not match the offer to the response.
In this day and age of the internet, I wanted to bring this third limitation to the fore for discussion because this dimension is the most important. You send and offer and someone responds, that is obvious and straight forward. But here we have a situation where someone was sent an offer and someone else has responded. Who is this responder how did he or she get this offer? This can be interesting but wait, who is the offer to? This is even more interesting.
Here is a person we should look through "Loyalty" lens. This person did not only respond, and yes, it should count as his response but he gave it to someone as an ambassabor of the brand.
We need a way to structure and look at response dimensionally and our CRMs need to enable us to do this. As we start collecting this type of data and analyzing their impact on our business, I think we will learn more about rewarding relationship with our customers that go beyond bean counting and traditional cross-sell/up-sell.
I would like to see almost 6 dimensions where response can be captured:
1) Individual level
2) Couple Level (identifying husband/wife together or cohabiting partners)
3) Family including children
4) Friends
5) Company Coworkers
6) Company
I am sure if we start working with this, we might think of other layers but this would be a good start.
Will this work? Is it to hard to do this?
What do you think?
This thing has been bugging me since we started assigning responses to our promos.
It starts with the realization that one needs to track responses as uniquely as possible and assign these responses to the specific promo to measure effectiveness.
That is well and good.
So we assign Unique Identifiers to the records and unique Source Codes and Cell Codes to make that data as granular as possible. We hope that this information is printed or attached somewhere so it comes back to us with the same integrity with which it left.
When the response comes back we hit several brick walls that prevent us from assigning the response flags effectively.
1) The unique codes were conveniently ignored at the vendor fulfilling the campaign and now we are left to match names and addresses and emails to figure out who actually responded.
2) Data Entry folks are not rewarded for correct entry than quick entry so a significant portion of the response now has to be again matched back using steps described in 1.
3) Finally, the offer to A was forwarded to B and B used it. Do we credit A for the response or B and if we wish to credit B how do we we did not send any promo to B?
I have talked about this the last 15 years to dozens of Clients and to organizations where I owned processes and databases. But I failed utterly to change any process because of the number of seemingly imponderables.
The most significant one, it seems, is an organizations inability to change their existing process to add the three new pieces of data to the output files that would drive fulfillment processes at third party locations.
The next significant is the organizations inability to change a manual process to a automated process based on scanning labels containing this data which would clearly be more accurate and fast.
The last one is an organizations inability to see response beyond the single dimension of promo to response. I had to by most part ignored any response where I could not match the offer to the response.
In this day and age of the internet, I wanted to bring this third limitation to the fore for discussion because this dimension is the most important. You send and offer and someone responds, that is obvious and straight forward. But here we have a situation where someone was sent an offer and someone else has responded. Who is this responder how did he or she get this offer? This can be interesting but wait, who is the offer to? This is even more interesting.
Here is a person we should look through "Loyalty" lens. This person did not only respond, and yes, it should count as his response but he gave it to someone as an ambassabor of the brand.
We need a way to structure and look at response dimensionally and our CRMs need to enable us to do this. As we start collecting this type of data and analyzing their impact on our business, I think we will learn more about rewarding relationship with our customers that go beyond bean counting and traditional cross-sell/up-sell.
I would like to see almost 6 dimensions where response can be captured:
1) Individual level
2) Couple Level (identifying husband/wife together or cohabiting partners)
3) Family including children
4) Friends
5) Company Coworkers
6) Company
I am sure if we start working with this, we might think of other layers but this would be a good start.
Will this work? Is it to hard to do this?
What do you think?
Friday, September 18, 2009
Indiscriminate Data Cleansing may be injurious to your companies financial health!
My recent experience of owning a customer database and actually living with it for a year and a half has given me some insight on data quality and need for data cleansing.
Typically when it comes to poor decisions in business, the often blamed entity is the data used to make those decisions. And it is true that over time data tends to become inaccurate. People move, things change, data goes missing. In the end conclusions become erroneous and models do not work.
Does that mean that the solution is to indiscriminately cleanup your entire database?
Many organizations try this at great cost. Sometimes the results can be profitable but in general it would not fetch a decent ROI. Often done as a mandate from the senior management, because money is there, no thought is put in regarding the best way to define this project and save money in the process.
I suggest a different approach.
Step 1: Segment your database. It does not matter how you segment it, objective is to find that 20% of your customers who are contributing 80% of the returns for the organization.
Step 2: Get your business leaders to define goals or business objectives targeting this segment for specific returns. For any company this would be the most important segment and if there is one project that must be approved and funded that year, it would be this one.
Step 3: Decide what existing and new data elements will be necessary to support the goals and objectives of the step above and target these for the cleanup phase.
In the next blog we will discuss further how to effectively cleanup the select data for your prime segment and do this at a budget your CFO would love.
Typically when it comes to poor decisions in business, the often blamed entity is the data used to make those decisions. And it is true that over time data tends to become inaccurate. People move, things change, data goes missing. In the end conclusions become erroneous and models do not work.
Does that mean that the solution is to indiscriminately cleanup your entire database?
Many organizations try this at great cost. Sometimes the results can be profitable but in general it would not fetch a decent ROI. Often done as a mandate from the senior management, because money is there, no thought is put in regarding the best way to define this project and save money in the process.
I suggest a different approach.
Step 1: Segment your database. It does not matter how you segment it, objective is to find that 20% of your customers who are contributing 80% of the returns for the organization.
Step 2: Get your business leaders to define goals or business objectives targeting this segment for specific returns. For any company this would be the most important segment and if there is one project that must be approved and funded that year, it would be this one.
Step 3: Decide what existing and new data elements will be necessary to support the goals and objectives of the step above and target these for the cleanup phase.
In the next blog we will discuss further how to effectively cleanup the select data for your prime segment and do this at a budget your CFO would love.
Subscribe to:
Posts (Atom)
