Business Lists and Business Data from Responsiva.

1. Data List Owners

List Owners not only own the source of the file, but they maintain, update and refresh it. This update process is typically from a team of outbound telephone operatives, calling through the file, updating each record manually on an ongoing basis. Other source files (such as Companies House) rely on the business to update any change of details, or financial information, via their annual returns and legal obligations. Single source files (list owners) don't cover the entire business universe; some focus on the corporates, whilst others are richer in the SME workplace. Sourcing data directly from a list owner will mean missing out on those prospects which the owner doesn't have within their file.

2. Data List Amalgamators

Most list owners lease their data through List Amalgamators, who merge several lists together to make their own version of a single business universe. It is these universal files that represent the best sources of prospect data. Typically they have the breadth of data volume (spanning approximately three million business data records), and the depth of field coverage such as multiple contact names, email addresses and additional profile variables.

3. Data List Brokers

List brokers are more prevalent than the list amalgamators. They re-sell business data directly from a relationship with the list owners or the amalgamated files. Ultimately all parties in the chain receive their royalties; the customer pays the list broker, the list broker pays the amalgamator, who in turn pays the list owner.

Typically the customer pays a similar price regardless. The brokers, amalgamators and owners haggle behind the scenes for fair rates, data quality, service lines and long-term relationships.

List amalgamators (and experienced list brokers) supply many millions of records per year, enabling them to ally with quality list sources at a fair price for the end customer. This is how they can compete in the same marketplace as the list owners. Not so dissimilar to the insurance brokerage industry.

One particular advantage of an experienced list broker or amalgamator is that they can provide an unbiased service, because they don't own the list sources. They recommend using the sources appropriate for their clients, at the most attractive price. And (if needs be) they can switch between data sources to find the right business list for the end customer. The reputable brokers and amalgamators will also refuse to work with unethical list sources, or those of a substandard quality. So with a list owner you will almost always be told "our data is the best" etc, and this gives you no guarantee that their file is of a reasonable quality. At least with a list broker and list amalgamator, some quality checking has taken place, and the company will (or should) only re-sell the files that genuinely are of a high standard.

There is also the matter of a wider scope from amalgamated lists; the best ones being those which span the full coverage (or at least 90% coverage) of all available businesses within the UK. No list owner has this full coverage from a single data source, so the end customer will generally get a wider market coverage from a list broker or amalgamator.

Think of the business universe as a Venn Diagram.

Telephone directories and Companies House data represent sizable circles within this universe, and other list sources overlay their contribution. Lists overlap with other lists, and most have their own unique entries.

List sources rarely have a common business ground with each other, so it becomes the task of the list amalgamators to pull several source lists together and merge them as a single file. This process is regularly referred to as an update.

Updating is no simple task; not only must it be performed every time an updated file is received from source, but it also needs to respect duplicate records (across multiple sources). If a record is in two lists, it is probable that each list will contain some unique information. The update amalgamates them as a single record, and keeps the unique information from both sources.

Data Structure

A data record is the term given to all information stored against a single database entry. So if the data were stored in a spread sheet, a record is all the information held within a single row.

Business data is usually stored at a premise level. i.e., one record (or data row) for each business location. This is best explained with examples;

1. A supermarket chain with one head office, four regional offices and 500 supermarket branches will have... 505 records.
2. A sole trader working from home will have... one record.
3. A doctors surgery with five G.P.'s and a practice manager should have... one record.

The third example is perhaps the best place to start. Let's say there are 30,000 G.P.s working from 10,000 practices. If you would like to market to every GP in the UK then each practise (i.e., each record) should in theory contain the name of every GP within that practice. The data can be expanded upon supply, so that you can prepare 30,000 letters. Bearing in mind that these will reach just 10,000 addresses.

Likewise if you wanted to market just to the practices (or practice managers) then you can address your marketing material to the 10,000 practice managers. I would suggest that the latter would be the more efficient route to market, though this does depend on the offering.

So in this case it is better to amalgamate all GP's at a single practice as a single record.

But where branches of a larger organisation are concerned, it is better to store the data at a branch (or premise) level. A small farm in a shire county may be looking to supply their harvest to the local supermarkets, some of which have a certain degree of autonomy in supporting local agriculture. So why would you wish to contact the head office and go through the lengthier process and red tape.

The inclusion or exclusion of company branches from a business marketing list is probably the most common complaint from data purchasers. i.e., they want to target decision-makers, not branch managers with no decision-making capability. This is a completely separate subject which is covered elsewhere within this book. But in straightforward terms, branches can be excluded by a number of de-selection criteria. And a good b2b data supplier will establish your requirements in advance, ensuring your file is not peppered with prospects which are unfit for purpose.

Updates & Duplicates

The updating process reveals anomalies; recognising two business data records as being the same entity is not an exact science.

Multiple Business Owners

Two distinct businesses would have unique company names and registration numbers with Companies House. But they could have the same postcode, telephone number and proprietor contact name. i.e., one business owner running two companies from the same premise. For example;

Most data amalgamators will recognise these companies as two distinct businesses.

So ask yourself; would you want to write to, telephone or email this proprietor twice? Or more?

Any data provider should be able to de-select these records for you. It may be worth asking next time.

The previous example represents quite a dilemma for the list amalgamators. Many marketing initiatives would want to target both businesses, even though the message is directed at the same person. It could be to market different products or services (relating to the two business types at the premise), or the matter may be more legally-driven around the distinct legal business entities.

So the list amalgamators are right to keep these records in the universe, even though many SME marketing initiatives hinge around appointment-setting with a single person.

Cross-file Duplicates

Other duplicate entries occur when two list sources are merged together and each contains the same business but with different spellings or data capture errors. For example; a single character difference with the postcode, two different telephone numbers (perhaps for different divisions of the company) and a different company name spelling.

To the eye it is possible to recognise these two companies as one and the same. The two list sources host the records as two 'different' company names; perhaps one received the legal trading entity name and the other the trading name. Fuzzy matching will review a percentage confidence level when comparing the two company names. However this company is essentially called "ABC", which contains just three characters. So when the lengthier words "Logistical Services" are added to the company name, just 3 of the 21 characters match! That's a very low percentage match rate, so the companies would not be considered as the same.

Coupling this with the two different postcodes (which could relate to a data capture error, or the fact that some companies operate with multiple postcodes for different divisions) then there becomes a decreased likelihood that these two records could be considered the same.

A third variable which is commonly used for deduplication (matching) is the telephone number. This is a reasonably accurate and definitive matching field because one would expect a telephone number to be unique to each company. The anomalies occur though when (a) companies have multiple telephone numbers for different departments (sales, service, finance etc) or (b) multiple companies share a central reception desk and telephone number. So the telephone number is not a perfect business data matching field, and will generally only be used in conjunction with (and support of) other fields such as the company name, address, industry classification, postcode and company registration number.

Intra-duplicates

Some list sources contain intra-dupes; duplicate records within their own file. These need to be recognised and reduced to a single entry also. These can occur from either of the reasons above, or simply through bad data list management.

Real Businesses?

There are other anomalies to consider; the business universe isn't compiled of just planets and pretty stars. Two asteroid belts to navigate are the not-for-profits and sole traders.

The not-for-profit region is huge. Quite often the businesses considered as not-for-profit spreads well beyond charities and into government, schools, doctors, hospitals, helplines, housing associations and even care homes. There are more. Despite each running to a pre-defined budget (and most are in fact expected to turn a profit), many are considered too channelled with their finances, leaving limited or no opportunity to market a service not already established within their operation.

Places of worship are also typically considered unsuitable targets, unless the marketed offering is specifically relevant to them.

So when a marketing manager wishes to exclude "not for profit" businesses from their target data list, in reality many of these aforementioned classifications should be excluded too. As with all aspects to a specification, the primary cause for a data list containing undesirable prospects is because...

• The marketing manager (data purchaser) didn't say

And

• The data supplier (list broker) didn't ask

Sole traders vary from the genuine business to the part-time "hobby" business. Many sole traders are employed by another organisation, running their micro business as a hobby or for a little extra income each month. They may perform a home hair-dressing service one or two evenings per week, or some cabaret show at weekends (magician, singer, dancer, comedian etc). The same could be applied across most business sectors, and the question may arise as to whether you should discriminate the full-time business owner from the part-timer?

This discrimination is not straightforward, because there is nothing within the business data universe to discriminate the part-timer from the serious business. However, other data fields within the records can help provide the data controller with an indication. For example;

A premise type appears with some list sources, enabling the business trading from home to be side-filed from those trading from shops and offices.

The length of trading (or residency at premise) will enable long-established firms to be segmented from the start-ups.

And the turnover band can help point towards those with insufficient cash-flow for your services.

From this table, only Record 1 has more than a £50K turnover, has been established for 3+ years and doesn't trade from home. What this table illustrates is that not all sole traders are the same; some make great prospects, whilst others are not so great. And it is through the consideration of multiple variables that we get a clear picture and can segment the good from the bad

Record number 4 (as a good example) illustrates a sole trader, working from home, with a very low business income. This record is the most illustrative of an individual running a "hobby" business for extra monthly income, and therefore the business most likely to be operated by an individual who has some other form of employment.

How this contrasts to record number 2 (who has a similarly low income) is that record 2 trades from an office premise rather than from home. This is indicative that record 2 is perhaps a more seriously dedicated business. Of course this will not apply in all cases; there are numerous "serious" companies who trade from home. Driving Instructors, white van (builders / plumbers / electricians), architects, book-keepers and many other classifications have strong prevalence in trading from home. Even your truly (author of this book and successful business data broker) operates from home; for no other reasons than it keeps operating costs down, is convenient and there is no commercial justification to trade from an office. That is unless pride drives the business owner to rent an office so as to appear more professional?

If micro businesses were my target market (and using the table above as four examples) I would be inclined to consider all four record types. But the companies trading from home would be allocated a smaller percentage of the overall selection pool. For example; select 20% of the overall data list to include companies trading from home, with the remaining 80% being companies trading from offices, shops and other premise types. This is covered later, during the "Data Specification" chapter of this book.

Grey Data

From the examples above, there are different interpretations of duplicates and not-for-profit organisations. There are also ways of segmenting sole traders, to ensure only the strongest prospects are chosen for a marketing campaign.

Perhaps if these sole traders were not the target market today, they may represent an opportunity for the future. Organisations which target companies with a turnover of at least £250,000 would exclude all four records from the table above. The first prospect (record 1) appears to be the closest opportunity of the four however:

The turnover may be as high as £200,000, and is known to be a respectable six figures. And (by virtue of being a sole trader) the profit could be considerably higher than a higher turnover company which employs a team, and is therefore burdened with a larger salary or premise-related overheads.

This is where data becomes grey.

If the requirement is a turnover of £250,000 or more, are there records outside this remit that are worth considering? And are there records inside this remit that should be excluded?

Defining a sole trader is also open to interpretation. An employee count of 1 is common, but is zero employees not the same thing? A sole trader may have a silent business partner, declaring a second person within the business. And they may have an admin or temp support, taking the count to 3. An employee count of 0 - 2 is generally a good compromise for specifying sole traders.

Employee count is also a grey variable, not least because it is one of the more fluid variables within the business universe. If a company has 100 employees today, it would take a huge jolt (serious loss or expansion) to vary this number by 20% or more. But where a company with five employees is considered, many considerations are needed to interpret the true meaning.

Firstly the date of last update. If the company record was updated last month then it is more likely to have five employees than a business which has not been verified for a year.

For most verification calls, it would be the person answering the telephone within the business that declares there are five employees. This could be a genuine answer, or it could reflect that there are just two full-time directors, supported by three temporary or part time employees.

A data field is the term given to the variables within the database. In spread sheet terms, this is best identified as the columns. Typically the first column will be a URN (pronounced as the three letters U - R - N, rather than as a vessel for storing ashes or tea) , which is a unique reference number enabling quick reference back to the business data universe. The second column (or field) usually hosts the company name. And so on.

In general terms there are four types of data field;

Sections 2 & 4 are the most interesting, and will form the basis of the rest of this chapter

In simple terms, section 2 (profile fields) can be appended to your business database by a list broker, whereas section 4 fields ("bespoke data") needs to be appended by you, the data controller, via a suite of marketing initiatives such as telemarketing.

Profile Fields

These are the criteria which a prospect list can be selected by, and will ultimately form the basis of your b2b data specification.

The most commonly used profile fields are as follows:

Postcode

There are four levels of postcode used, for which examples follow:
AREA: NW, W, SW, EC, WC
District: NW1, NW2, NW3
Sector: NW1 1, NW1 2, NW1 3
Full postcode: NW1 1AA, NW1 1AB, NW1 1AC

Company Size

There are three variables which can be used to distinguish a company's size:

TURNOVER. Typically this field is banded. E.g. <£50K, £50K - £250K up to £50M+

EMPLOYEE COUNT: This field is usually distinct, updated from the most recent verification call. The data controller will usually want to select an employee range, such as 5 - 100.

BRANCH COUNT: Companies can be selected (or excluded) by the number of branches they have. Typically this will be either (a) remove national chains, or companies with >10 sites or (b) include only companies with 3+ sites.

Industry Classification

There are several industry classification tables, ranging from the SIC code (Standard Industry Classification code) to the telephone directory classifications. In general, the directory classifications are richer, drilling deeper into the data. Examples include the segmentation of book-keepers from accountants, or the types (nationality) of food a restaurant serves rather than just the higher-level classification of; restaurants'.

Industry classifications can also be grouped into Business Groups (such as the Service Sector, Education, or Manufacturing), and the lower level Market Sectors. An example of this may include Marketing Companies or Business Consultants, which both reside within the Service Sector.

The important thing to bear in mind when selecting b2b data is that industry classifications may be applied as an exclusion rather than as a selection. Good examples of this are charities, government, and churches. Companies who sell into all SMEs may wish to exclude these industry types (and others) prior to purchasing their data list.

Premise Type

Perhaps the most overlooked of all b2b data selection tools is the premise type. This field gives no direct relevance to the industry classification, and looks only at the kind of building that the company trades from. Values include:

• Factories
• Warehouses
• Businesses trading from home
• Retail outlets
• Offices and head offices
• Education premises
• Medical premises
• Repair centres & workshops
• Holiday & leisure premises
• Places of worship
• Other business premise types

Some examples of why this field can be of more relevance than the industry classifications are;

Although there are approximately 2.2million businesses registered with Companies House, only one million records are active. There are two primary reasons behind the 1.2million inactive records;

Of the one million active companies, many are limited liability partnerships who are not required to report turnover figures. And of the limited companies who are required to file annual accounts, there is a cut-off of turnover at £5.6million. Companies filing accounts below this figure are not obliged to reveal their turnover figure for the purpose of public domain viewing. Almost all of these take the decision not to reveal their turnover.

In summary; around 100,000 - 150,000 companies are obliged to reveal their annual turnover.

This comprises less than 2% of the overall business universe of 3.2 million records!

Where Do The Turnover Figures Come From?

Answer: Data Modelling.

By analysing a company's industry classification, geographical location, employee size and number of years trading some very sophisticated data models have been developed. Through research and testing, these models have proven to be reasonably accurate, enabling a turnover "band" to be applied to most business records. Bands are typically <£50K, £50K - £250K up to £50M+

The strength of these data models sits with medium to large organisations with more than ten employees. For the companies with up to ten employees the model becomes awry, as the data is sensitive. i.e., length of years trading is usually low, employee size small etc. In other words, I would not recommend trusting the turnover bands for the companies having up to ten employees.

And here is the problem: about 90% of business records have up to ten employees!

So that would suggest the turnover model contains some serious anomalies for around 90% of business records. i.e., those with up to ten employees.

To make matters worse, it is at this end of the spectrum where an accurate turnover is most frequently required to be accurate.

Perhaps the weirdest aspect to comprehend is that the data model is actually robust (or certainly the one in particular that I am aware of). The data model works fine, but is restricted by the quality of information feeding into it.

So How Can The Turnover Field Be So Weak?

Selecting b2b data by modelled turnover alone will yield disappointment. Here is why;

Let's assume that the turnover data model is 95% accurate with companies having more than ten employees, but only 75% accurate with the companies having up to ten employees.

So why would the data be so badly inaccurate if you select companies having a £500K+ turnover? Firstly let's review the overall universe by employee size;

Then let's look at how many accurate and inaccurate records there are;

This means that there would be FIFTY times more inaccurate records for companies having up to ten employees. And as the data is massively skewed towards these companies (ten times more of them) the overall file will appear heavily peppered with "bad" data.

Another way of looking at it is to review a file of 200 records:

20 of these records will have more than 10 employees. At 95% turnover accuracy, this means that 19 of the 20 will be good.

But 180 records will have up to 10 employees. At just 75% accuracy, this means that 135 records will be good.

The total being that 154 records are good. 77% of the overall file.

A figure of 77% is much closer to the lower accuracy rate of the companies with up to 10 employees.

And that means that when you are selecting data by turnover, the overall data quality will tend towards the model's accuracy of the smaller companies; because they have a massively higher percentage population.

Turnover-Per Employee

If you require an accurate turnover field and are targeting businesses with up to ten employees (e.g., say your brief overlaps and targets 5 - 100 employees) then do not rely solely on a modelled turnover field. In most cases it will lead to serious disappointment.

The employee size for these micro organisations (which comprise nearly 90% of the business data universe!) is a stronger indicator, especially when coupled with business sector. For example; a service sector company will typically yield a higher turnover-per-employee ratio than a manufacturing company.

For companies with up to ten employees, I would recommend working to the following estimates;

Employees	Low	High
1 - 2	£50,000	£200,000
3 - 4	£100,000	£250,000
5 - 6	£200,000	£400,000
7 - 8	£300,000	£600,000
9 - 10	£500,000	£1,000,000

Despite the wide gap between the high and low estimates, even these figures will generate anomalies.

There are sole traders with a turnover in excess of £200,000, and companies with ten employees having a turnover less than £500,000. But from many years' experience (rather than data modelling) in both data and outbound b2b data telemarketing, these figures represent a reasonable estimate and should cater for the majority of the records you source.

Prospect data specifications can be ambiguous, so it is imperative to ensure they are accurate. In my opinion, it is the responsibility of the list broker to discuss the b2b data requirements with the customer and then take ownership of specifying the required data list in accordance with the customer's requirements. It would then be the customer's responsibility to agree to that specification, or make changes where required.

This goes back to my previous point:

• The marketing manager (data purchaser) needs to say what they want

And

• The data supplier (list broker) needs to ask questions

Having worked within the marketing data industry since the 1980's I have seen many examples of errors, which go some way to justifying why (a) customers need to work with a fully trained and competent account manager and (b) why it is always best to source data from a data 'person' (i.e., with open discussion) rather than from an automated / online data purchasing download system.

Typical examples include;

A. The Nurseries

A client who sold children's soft-play equipment wanted to target nurseries. He was supplied a data list, roughly equally split between the industry classifications "nurseries & creches" and "garden centres and nurseries". The problem was quickly spotted and rectified, though it is perhaps easy to understand why this issue occurred in the first place. But many automated downloading systems request advance payment before you download the data list. So a refund is not always granted once you press "buy", because data is essentially 'information' and once you have been given the information it is impossible to return it. Furthermore, some of the online ordering systems may appear user-friendly at the front end, but it is not always easy to get through to a person on the telephone (and get refunded) if an error such as this has happened.

B. SME's

One of my favourites is an appreciation for what constitutes an SME (Small to Medium Enterprise). If a client ever says to me that they wish to target SME's, I always respond by asking what they mean by this. This is coupled with a suggestion, such as "would 5 - 100 employees be acceptable?". Nearly always the client would come back to me with an alternative employee band, or turnover figures. And despite nearly twenty five years of data experience, my guess at what the client thinks an SME should be differs from their reality. Responses such as "No, there needs to be at least 50 employees. Say up to 1,000" are commonplace. Or "Sole traders are fine, so anything up to 50 employees will do". Both of these responses not only make a huge difference to the data available to that brief, but will also have a considerable impact on the marketing initiative's results.

Back to the earlier point; the marketing manager said what they want ("SME's"), and the list broker had to ask a question to clarify exactly what that meant. It's not difficult!

The important thing to remember when selecting data is that the fields are usually gathered and verified by telephone. This creates possible anomalies for numerous reasons;

• Changes to the company data over a period of time (since the last update)
• Data capture keying errors on the part of the data verification company
• Exaggerations or errors on the part of the company being verified (data can only be updated on the strength of what is advised)

So if a customer requires companies with 5 - 100 employees, consider that the companies with precisely 5 employees (and there are approximately 200,000 of them!) could for example be a husband & wife partnership, with two or three temps. When they were last called (and the employee size verified) they may round up to saying there are five employees, in the interest of giving a simple number or "bigging themselves up" a little.

C. London

What is "London"?

The City of Westminster? The square mile? The postcodes EC & WC? An area within the North-South Circular? The compass-point postcodes (N, E, SW, W etc)? Or is it the area within the M25?

The reason I ask it this way is because customers frequently have a different view on what constitutes the London area. Typically I would suggest the M25 area to be correct, but I never assume this and always clarify with the customer first. This is all part of the specification process, and back once again to asking questions to clarify the true requirements.

An Example Specification

The b2b data specification needs to cover all bases, and accurately clarify what the client needs. Here is an example;

• London (postcodes WC & EC only)
• 10 - 50 employees
• All records must include a director or business owner (no managers / branch managers)
• All records must have a TPS-checked telephone number
• Exclude charities, government, medical and education sectors
• Exclude national chains (companies with 10+ branches)
• Select OFFICE premises only
• Select only companies with an accompanying email address; exclude generic email addresses ("info@" etc)
• Then: select and supply the 1,000 most recently verified records

This brief covers all bases and by operating to this same standard client complaints after the data is supplied are extremely rare.

Control Cells

A control cell is a selection of the data which is used to test the campaign results. For example; if your target market is accountants with 5+ employees, the eventual data selection (of 1,000 records) may be as follows:

• Accountants with 5+ employees: 900 records
• Accountants with <5 employees: 100 records

Although the latter selection is not within the desired brief, this 10% of the file acts as a sense-checker. When the campaign has finished, and the results are reviewed, ask the question: did that 10% (the control cell) yield a similar percentage of enquiries / responses / sales?

Although in many cases control cells do not yield as well as the main selection, occasionally they do throw up pleasant surprises. One possible example is that larger businesses tend to prove harder to reach the business owner through a telemarketing campaign, whereas the smaller companies (with less than 5 employees) can yield a greater connection ratio to the business owner. For this reason alone, the smaller companies may prove more fruitful. The sale potential may be smaller, but with the increased connectivity the overall sales may prove higher.

Always consider a control cell, even if your ultimate conclusion is not to have one.

Request Samples

In addition to providing a data specification, good data list brokers will also supply a few samples. So if you aren't supplied some as standard, then request some. These samples are not intended for use in a test campaign, though they can be. Their main purpose is to illustrate how the data is supplied, the fields and format of the data supply and (most importantly) for the customer to sense-check that what is being supplied meets with their requirements.

For example, the client may respond to the brief & samples with "I see the samples contain an estate agent; actually, we don't want those either".

Request A Breakdown

The client has the right to say "I need to sense-check the full 500 records you are proposing to send me, so please would you email me an industry classification breakdown?".

Or; "Can you email me a postcode breakdown?"

What this will achieve is to let the client see the kinds of business (or their geographical spread) that will comprise their new data list. How many accountants, solicitors, graphic designers etc, and where these businesses are located. And from this they can pick out any additional undesirables to ensure their eventual list is exactly what they need.

Data Accuracy

One of the most frequently asked questions by clients is "how fresh is your data? / How frequently is it updated?". Ironically this is probably not quite the question the client should be asking. How would you feel if you bought a data list that was updated last month but (in-between then and now) all of those businesses had gone bankrupt and shut down? Of course the probability of this is virtually zero, but it raises the point that data "freshness" is not quite what you're looking for.

The true question to ask is "what are the guarantees regarding dead data?"

The answer may be different according to the different methods of contacting the data.

The b2b data industry norm is as follows:

For telephone numbers and postal mailing addresses, most data companies guarantee a 98% accuracy rate. This means that you would expect to have less than 2% dead telephone numbers or postal mailing returns (sometimes referred to as "gone-aways"). And a guarantee which maintains the spirit of this agreement would mean that any excess of this 2% figure would be refunded to the client without question. With good quality data lists a refund is rare, but can happen sometimes even with the best quality lists.

Clients do need to report the dead data figures quickly however. If you bought a list 11 months ago then it is inevitable that the volume of dead records will increase. Most data companies will therefore stipulate that the guarantee only applies for 30 days from the date of supply.

Some companies offer replacements (or even a "2 for 1 replacement") on all dead records, whilst others stipulate that the guarantee only kicks-in once the 2% threshold has been passed.

The reason for this latter stipulation (i.e., a 2% threshold) is that virtually all files of (say) 1,000 records will include one or two bad records. In reality, probably 100% of files contain at least one dead record; companies close every day and it is just not possible in today's market to maintain every record every day. So a fully verified record today could be a closed business tomorrow. Without the 2% threshold, data list brokers would be writing out credit notes and refund cheques for "one record here, and one record there" for every single order they supply. It would be crazy.

But what some data companies can provide is a goodwill business ethic. If the client says "hey, we had a few dozen dead numbers last time" then the data company will most likely want to maintain good relations and add an equivalent volume of free records to the next order. It's all fair play.

In practice, most of the best data source files have less than 2% dead telephone numbers, or mailing addresses. 1% is probably closer to the norm. But 2% is usually chosen as the best figure because numbers do fluctuate, and the 2% threshold allows a reasonable degree of "catch-all" and manages the client's expectations fairly. There are some b2b data list sources within the direct marketing industry which have a considerably higher dead data rate. Some even as high as 20%. No self-respecting list broker will touch or supply these sources, but the most unfortunate fact is that many unsuspecting customers buy these files because they are usually cheaper. I will cover this point in the next chapter; when all things are considered, they are in fact not cheaper at all!

Email Addresses

Email data has a higher dead (or "bounce" rate).There are two types of bounces;

1. Hard Bounces

Most frequently, the hard bounces are caused by:

• The email has been deleted because the employee has left the business
• The domain/URL has been shut down (possible company closure)
• Data capture errors in gathering the email address

In general terms these bounces can be regarded as a data issue, with the list broker being required to take some responsibility for them.

Generally, hard bounces can be expected to be in the region of 5% to 10%. When sourcing an email data list, always ask for the guarantee on hard bounces; there are lots of files out there which can sting!

2. Soft Bounces

These are the bounces which the email data list provider has no control over. Usually they relate to spam filters, which can vary in their rejection of your marketing email depending on the system you are conducting your marketing campaign from.

Typically the soft bounces will add a further 5% to 10% in email rejections.

Cheap Data Sources

What does the word "cheap" really mean to you? In my opinion it means, quite simply, that you pay less. But in buying data you should consider all the costs associated with the campaign. And even by keeping the algorithms simple, you can very quickly establish the difference between "cheap" data and the more expensive files.

Let's start with telemarketing. You need to source 1,000 records, for a campaign lasting eight days. The telemarketer is expected to call 125 prospects per day, which is a reasonable expectation. As your telemarketer is an in-house employee, you are probably paying around £100 - £150 per day for them, considering salary & holiday/sick pay, insurance, office space, equipment, telephony charges, management costs and all other associated overheads.

You have been quoted by two data list brokers as follows;

• 1. 1,000 records with a 98% guarantee on telephone numbers: £200 + vat
• 2. 1,000 records with no guarantee: £100 + vat

File 2 seems like such a bargain! But let's say it is only 80% accurate. Here are the numbers:

Item	File 1	File 2
Data Sourcing:	£200	£100
Telemarketer (say £125 per day for 8 days):	£1,000	£1,000
(*1) Total cost:	£1,200	£1,100
Dead records:	20	200
(*2) Records connected to (i.e., good records):	980	800
Cost per connection (*1) / (*2)	£1.22	£1.38

The cheaper data source in this example equates to paying 16p more per connection.

Let's repeat this example where an outsourced telemarketing company is used. These can charge around £300 per day for their telemarketing services;

Item	File 1	File 2
Data Sourcing:	£200	£100
Telemarketer (say £300 per day for 8 days):	£2,400	£2,400
(*1) Total cost:	£2,600	£2,500
Dead records:	20	200
(*2) Records connected to (i.e., good records):	980	800
Cost per connection (*1) / (*2)	£2.65	£3.13

Now the cheaper data source costs an extra 48p per record. If this figure were applied back to the data, the true data cost is not £100 for the 1,000 records (i.e., 10p per record) but £580 (i.e., 10p + 48p per record). Another way of looking at this is that you would be paying nearly three times as much for the cheaper file.

The same example applies to direct mail. Typically a letter can cost about £1 apiece to print, pack & post. Potentially the costs can soar if an expensive marketing pack or brochure is being despatched.

Rather than illustrate this with another table as per the above examples, perhaps you would like to consider the maths based on your typical mailing fulfilment costs, adding in the costs of design and production of your marketing literature.

Simply put, don't cut corners with "cheap" data!

Why Is Cheap Data Cheap?

Given the above examples, cheap marketing data is usually not so cheap. In fact most of the time it ultimately proves to be more expensive.

If you consider what constitutes the full business data universe of around three million records in the UK, how do you think it is maintained? The most respected list sources are telephone verified on an annual basis, which equates to three million calls per year. Working on the premise that (say) 125 businesses are called each day by the average telemarketer, this means that to call the full file of three million records will take... 24,000 days!

If we work on the premise that the average employee works for 240 days per year (allowing for sickness, holidays and a normal 5-day working week) then we are left with the magic number of a team of 100 full-time employees.

Then add to this team: team leaders, managers, directors, HR and a back-room analytical (data quality) team. We are probably now considering an overall team structure of 150 staff. The data needs to be "sold" in order to make an honest profit from this commercial venture, so add in a sales team (with company cars), account managers, their internal managers & directors, followed by project managers, analysts, project managers and solutions architects for the larger client data-driven solutions. The organisation requires marketing & PR, with more HR support, a finance department and an IT infrastructure.

So a company with 250 or more employees is not unreasonable. These require office space, desks, chairs, p.c.'s, telephony, personal development plans and so on.

And that is just to maintain the data.

Additional costs occur in sourcing the new businesses, such as from the BT OSIS file. Then there are company insurances, licenses and other commercial costs in sustaining such an organisation.

In other words, a company with running costs of say £10million is not an unreasonable figure in order to sustain and verify a UK-wide business database. And for such a company to stay in business, data sales need to exceed that £10million running cost.

Can they achieve this by selling 100,000 records on a disk for £100 a time? Absolutely not. Instead they charge a fair price, which is evaluated by establishing the total number of data sales, size of marketplace and essentially a figure which will return a reasonable sales figure for their own internal investment in maintaining the database.

So why are there companies out there who can supply 100,000 records for just £100?

You tell me!

If the data is less frequently verified and updated (or indeed never updated!) then the operational costs are reduced accordingly. As a customer you get to 'enjoy' a lower data cost, but would not see the data company's reduced internal costs, which effect the overall quality and accuracy of the data.

Or perhaps there is some illegitimacy to the data source; not all data lists are ethically sourced and verified. Some have historically been identified as 'stolen'. i.e., sourced once under strict license, and then illegally re-sold multiple times. Although it is not an easy task to police the data industry, there are methods (such as "seeding") and perpetrators can, and usually are, ultimately found and prosecuted. Unfortunately though, this takes time. Time during which many customers have already sourced and used illegitimate files.

So what?

I have already identified one of the issues with a cheap data source which is less regularly updated; ultimately it costs more when the marketing expenses to that data are considered.

But the worst potential cost comes from using copyrighted data, even if you were unaware that the data was stolen. "Handling stolen goods" is a prosecutable crime, even if you were unaware that the goods had been stolen. The same principles apply to data. As the data purchaser, you and your company are liable. So it is imperative that you check the data list broker's credentials at the point of enquiry.