Forbes calls data cleaning the most time consuming, the less enjoyable data science task, and that is absolutely relevant for marketers. Demand gen performance heavily depends on data quality, but data quality itself is not commonly included in marketing KPIs. Thus it often becomes „someone else“ job, with very limited resources to handle it. Another challenge is data often owned by different parties within an organization, and you might depend on some data that is beyond your control or even access.
If you have dedicated resources – perfect. If not – I suggest applying a lean approach to data quality. Even if you cannot have it all, you can still significantly improve the situation, fix broken parts and streamline the process. We can call this approach guerrilla, lean, agile or scrappy (to follow Nick Westergaard’s advice in his book Get Scrappy), but the main goal is the same: make the best of what you have. The key point is identifying and focusing on the most critical data issues that hurt your business process most of all. At the same time, taking simple and proven steps to make progress in that area.
Before starting, I would ask the following questions:
- Think of your internal consumers [sales, management] – what data do they need, when and where? What exact data points do they rely upon? This is where you need to start.
- Think of what you measure and how you report – what data (fields) do you use? What systems are involved, is it MAP, CRM or both? How are they synced? Is everything visible on both sides? Where do your reports live? Of course, data need to be cleaned up in both systems. However, start from the very end – if you report in CRM, you need to have it done first.
Tip 1. Not all data is equally important – prioritize data points
- Build a sort of Maslow’s pyramid of data quality – make a priority list from a business perspective. Take actions based on what is the most basic need of marketing.
- Remember interdependency – you often need A to get B (thus A is becoming the same priority as B)
- Timing: not all data is equally important the same time. E.g. MQL data is needed right now, while data required for reports might wait till the end of the quarter. Adjust your priority by important vs. urgent parameters.
Tip 2. Think systematic – do not put patches
- Try to automate whatever is possible. Subscribe yourself to smart lists; create campaigns to standardize values on Country, State, Industry, Job Title; set up auto-routing rules for leads, etc.
- When building operational campaigns, apply OHIO (Only Handle It Once) rule – you build a cycle, not just action. So in most cases, that should include:
- One batch campaign – to fix what we already have
- One trigger campaign [daily campaign; go through the campaign every time] – to fix what we will have
- One remove from flow campaign / flow step – to circle it back and be able to fix next time
Tip 3. Put junk into bins
- Qualify out automatically: create lists variations of XYZ, XXX, a, 1, asdf, home, none, na, n/a, self, etc.
- Set up automated routing/disqualifying irrelevant leads – students (by job title), competitors (by domain match), vendors and employees (by domain match)
- Create lists with non-standard values to find and fix them easily
- Use junk queues or special “owners” to bucket irrelevant leads and remove them from the lead flow
- Ideally, you would have an auto-reply mining tool like Siftrock. However, you can still export auto-replies into excel, do quick search for: “No longer,” “retired,” “resigned,” “last,” “left” and suspend/erase those records
Tip 4. Set up marketing environment right
- Automated removal of test and irrelevant leads from marketing campaigns (by domain, by status)
- Keep campaign naming convention and campaign hierarchy consistency
- Define precisely your metrics and communicate to your team. They need to know what data you use for reporting on their performance (dates, types, costs, results).
Tip 5. Embrace risks
- When performing any data transformation (integration/deduping/converting), test in several batches, give it time
- Read activity log line by line, check both Marketo and CRM
- Inform your team ahead and ask to watch
- When there is no decision yet – make it!
Resources and guidance:
Gold Nugget Data Policy – Six Steps To Trash Your Bad Data by Jeff Coveney.
Washington DC Marketo User Group discussion on Data Quality and Deduping.
Marketo tips on data cleansing.