“The problem is that AI is constantly changing right now. What you knew six months ago might not be relevant anymore – or it might be obsolete.” - Ernesto Valdes
AI is going to transform the way our revenue organizations operate, not to mention boost profits. But when we try to implement the amazing use cases we’re seeing, we quickly realize that AI will be much harder to execute than we thought.
With new tools, stale and missing information, and different teams following different data practices, the quality of our data is a big mess. Until you clean your data, AI is off the table.
Don’t worry! Here are ten tips to get your data in order ASAP.
Before implementing AI across your tech stack, try a small pilot first. Pilots give you the chance to align cross-functionally on the desired outcome of an AI initiative. They also allow you to test sales rep adoption before you commit.
Asana ran their pilot with a competitive intelligence bot that was launched in Slack. Once the RevOps team confirmed that reps were finding value in the bot, they transitioned it out of Slack and into Salesforce. Now, when a rep adds a primary competitor to an opportunity, the bot leverages ChatGPT to deliver competitive intelligence immediately.
Jump to the clip to learn how Asana piloted their Competitive Intelligence bot.
“For our content library, where we have our competitive intelligence, we have a team that constantly updates the data within that folder. It’s not stale data. Metadata needs to be up to date and accurate.” - Maschal Malek
Start by identifying, across all of your different systems, what data you care about, where it lives, and whether you can connect it to a central data lake. In this case, you’re looking at metadata, which are the tables and columns that make up the objects and fields that your data resides in.
Next, evaluate the data for fitness. How often is data populated and is it stale or inaccurate? For larger organizations that have lots of teams, you might find that they all have different data habits. Two different systems may give you two different reports about pricing. You must decide which one is going to be your source of truth when there’s competing information.
You’ll definitely need multiple data providers to broaden your data sets since each vendor will have different strengths and relevancy. At Asana, Maschal found that ZoomInfo does really well for account data. It also has good person data, but only in North America, not in EMEA or APAC. Asana needed additional data providers to flesh out their database.
It can cost a lot to enrich every single person and account in your system so it’s important to find data providers that are cost-effective as well as accurate for the different regions that your prospects live in.
Jump to the clip to learn how Asana chose their enrichment vendors.
“There are two approaches to removing noise: proactive automation - like brushing your teeth, and a reactive/manual approach - like going to the dentist.” - Ernesto Valdes
Noise is basically any data that stops you from understanding who your customers are and the relationships and history you have with them. To reduce noise you need a two-pronged approach.
First, proactively automate data cleanup by setting up duplicate rules to dedupe your leads, contacts, and accounts upon creation. You’ll want to set up strict cascading match rules that identify people by first name, last name, and email for example, then automatically merge these matches and flag any issues for review.
Second, set up a cadence to manually review records that weren’t matched automatically. This can be on a daily, weekly, or monthly cadence depending on the data volumes that you deal with at your company.
Normalization helps you dedupe proactively because it standardizes and improves the quality of your data. Start by standardizing and identifying commonly used values across your systems.
Ernesto uses state and country picklists, as an example.
People make frequent errors when typing into forms, so rely on picklists instead. Create anomaly alerts to identify when any new information entered into forms or coming in as a lead is outside of the standard format.
Jump to the clip for a deep dive into how to normalize street address data.
A shared data dictionary is always a winner. At Asana, Maschal started in a spreadsheet before moving over to Asana. She pulled in every single field name from their systems and labeled the source. Was it from Snowflake, Salesforce, or Marketo?
Each source listed their definition of the field. Sales reps could see how different systems named their fields and defined the value of the field. This helped them understand the importance of the data in their workflows.
“Leads that are not matched properly can create data silos, inconsistencies, and can cause a lot of fragmented data.” - Maschal Malek
Typically, Salesforce doesn’t connect your leads to an account. You end up with a lot of duplicates across contacts and leads and have to go through a deduping process. Asana uses Traction Complete to identify accounts that leads are tied to based on set criteria.
This is critical for maintaining Salesforce cleanliness. Without clean data, you’ll really hamper your AI performance because you can’t see if you’ve spoken to this contact or account in the past.
Jump to the clip for a peek at how Asana uses Traction Complete to manage leads.
When you zoom out, you’ll also need to connect related accounts according to their legal family tree of companies. This gives you the full context of who your customer actually is. Are there existing contracts, SLAs, or discounts with the parent company that need to be respected? Have you already gone through the procurement process, allowing you to expedite conversations?
AI tools like Einstein Prompt Builder can help you craft much better introduction emails if you understand the account context. Without reference to past communications, deals, or relationships, AI-generated emails will feel like cold outreach instead of a warm connection.
Zoom out one more time to connect accounts to their sales territories, giving you a clear picture of how you go to market. This allows you to identify existing patterns associated with successful, stalled, or lost deals and the role those play in the territory.
An LLM can automatically identify territories where legal contracts are filled with red lines, marking those territories as potentially at-risk. AI can also pick up positive patterns, like talk tracks that work especially well in a region.
Jump to the clip for a discussion about how contract red lines are actually a territory data point.
“You want to avoid failure in your initial AI implementation. How are you going to request extra budget to grow that AI implementation unless you can build on existing success?” - Ernesto Valdes
You can't do an AI pilot without setting some sort of measure of success to see how the initiative progresses. We're not going to know if it's successful or not if we don't get feedback, run the right reports, and align with our teams.
Make sure that the metrics you choose directly align with business goals and RevOps strategies. For example:
Want more information on data preparedness best practices from Traction Complete? Or maybe you’re interested in learning more about Asana’s AI pilots? Reach out to Ernesto or Maschal via the RevOps Co-op Slack community or through LinkedIn.
Looking for more great content? Check out our blog and join the community.
.png)
