[Peter Gulka]

... and my supervisor asked me to do this:

Quote:

think that the analytics package will be a great source of information for us going forward. We can apply this information to fund development activities, event planning and communications lists.

I would like the two of you to come up with and perform a list of data integrity tests that we can do to ensure that Blackbaud have used correct data fields in their analysis prior to importing the data into RE. The total giving numbers is an example of something that could impact the analysis and require scoring adjustment if incorrect or discredit the validity of the information if unexplained.

The "audit" should include verification of 80 to 100 individual constituent records to ensure that reported amounts are correct. The goal should be to take a good cross section of the data and test it. The desired outcome is to have demonstrated a very high rate of accuracy, and be able to accept the analytics information as highly reliable.

Have any of you tested the data you received like this? What approaches did you use to verify what they gave you for scores was accurate?

Did you just accept their model scores on faith?

[Matthew]

it seems to me that if you had the capacity to do the type of work it's going to take to verify the BBA data then you wouldn't need BBA to have done the work for you.....

[Doug Creek]

There really isn't any way to verify the data, other then doing what you just paid BB to do on those 80 to 100 constituents... I know when ours went out we had a very large duplicate consitutent problem (long before I got here) and when we merge constituents, we sometimes find out they had two sets of scores, one from a constituent with limited information, and one that is much higher from the constituent with more information. You always have to wonder what their score would have been if all the information had been in one record instead of multiple.

I don't think you really have any way to verify the data, and you kind of have to take it on faith. Its not a perfect science anyway. We've had many people that have low scores and have turned into some of our highest givers...

[slvelazquez]

In testing our BBA data against other results from the RFPs, we exported only the fields we needed from RE - then created a query of the same individuals in RE. Our PR then simply compared the data side-by-side in an Excel spreadsheet. It's a quick and dirty way to ensure that you're getting the right information on your test population.

Spring Velazquez
Director, Donor Relations
Widener University
One University Place
Chester, PA 19013

Widener University - Home

[dstepp]

I also did some side by side data comparison before importing BBA data to RE. Overall I am satisfied that BBA did a good job of getting the right data for the right constituents. Over the past several months I have found a few duplicates with disparate scores even though most of our model variables have nothing to do with RE data (although past giving type is heavily weighted in the model). I encourage you to attempt the comparison and a small sample of 80-100 records shouldn't take too long once you match up the data.

It's a guarantee that there is some slop in the results. This shouldn't concern you too much if your small sample looks ok. The model scores applied to individuals are statistics based on a population study and should be treated as such. I have used higher Annnual Gift Likelihood scores to narrow the size of our direct mailings by about 1/3 (saving great expense and increasing net revenue is the goal) but I don't exlcude lower scoring individuals if they have some other quality that suggests they might give (like having made a gift to the same mailing last year).

I did do a fair amount of training with our development staff so that they understand that a high scoring individual is someone they might consider contacting but that a lower scoring individual that they know is a good prospect should also be called.

Good luck.

[bjcadena]

Are you Using TRE also? or just the RE import file?
I just went through our data delivery and TRE part 1 training
I notice different score on some of my duplicate records as well as adult children that use their parents address still have higher scores sometimes.
Just wondering where you are at now, 4 moths later into the process,,,,,
Billie Jo

[Peter Gulka]

We did not get TRE as we have access to the Prospect module in Raiser's Edge. We just recently hired a new Annual Fund Officer, so we haven't actively used the data for a mailout or appeal yet. We are starting to slowly look at the values as we assign individual accounts to our Development Officers. These would be for larger value prospects that we don't want to send a general mailing to.

More news as it happens.

We ended up accepting the data on faith after some cursory evaluation. We are trusting BB that they know what they are doing.