How I Multiplied Visitor → Chat Conversion Rates 6x (0.1% → 0.6%) with Drift

I recently spent a few months with a client and one of my projects was to launch personalized chatbots with Drift.

Here’s the results of my work.

Website visitor → chat engaged moved from average of 0.1% to 0.6%, or ~2 chats a day to ~7 chats a day. This netted my client $600k in closed/won deals over the course of 3 months.

With great days hitting a +2.0% average conversion rate.

And some playbooks hitting +8% conversion rates (e.g. /contact pages).

At the end of this post you’ll know exactly how to build the right Drift chatbots to improve your conversion rate.

And, one note, this post is written for B2B companies with a high LTV.

Peter inspired me to write this up for his Modern Sales Pros group. Thanks Peter for the encouragement and the platform to share my learnings.

Average Visitor -> Chat % for all Playbooks after launching Drift Chatbots. Data.

And, if you prefer a video intro…

It’s me, on camera.


This post is a little bit dense, so here’s an outline to my process.

Validate if chatbots make sense for you (and how to calculate it). This section may lead to some reader drop-off! If you pass go, then…

  1. Drift setup 101 (getting started).
  2. Build your quick win bots (low setup, high reward).
  3. Build your all site visitors bot (this is the catch-all if no personal bot exists).
  4. Get to know you 1–1 targeting options (hyper-relevant bots!)
  5. How to research and prioritize your 1–1 bots.
  6. Do the work and write them :).

Validate The Opportunity

This is absolutely the most important thing you can do.

First you must know two things:

  1. That you’re getting enough traffic to run meaningful experiments
  2. That the traffic is of good quality (stay tuned to find out how).

For traffic estimation math

Conservative Math:

( Monthly Traffic * 0.001 * (your lead → close rate) * LTV ) - $18,00 (Drift yearly) = Way Worth It.

Optimistic Math:

( Monthly Traffic * 0.01 * (your lead → close rate) * LTV ) - $18,00 (Drift yearly) = Way Worth It.

This should give you an idea if it’s worth it.

For good quality traffic math

You can easily build a Google Analytics dashboard with Clearbit Reveal to find out:

Clearbit Reveal data that shows who is visiting the site. Amazing firmographic data.
  1. What companies visit your site
  2. What industries
  3. What technologies they use
  4. What countries they come from
  5. How many employees work at these companies

A separate subscription to Clearbit Reveal is key to know your playbooks… ping me if you want custom advice here.

But, you can also Keyword Hero or Google Search Console.

Here you can see what organic keywords people are searching, what pages, countries, etc.

This way you’ll at least know that there is intentful traffic coming to your site

(e.g. a search for Send Jordan Money is better than a search for Take Money From Jordan).

Shoutout to my Postcard Marketing company for providing live real data!

This can basically be napkin math, but if it doesn’t pencil out, don’t bother continuing :).

1. Setup Drift

I absolutely recommend a Drift Premium (I can get you setup with my favorite CSM and priority support). It’s so amazing to have a dedicated Drift CSM review everything… they will even edit your playbooks!

Plus you get Marketo integrations to add in super personalized targeting.

With a CSM you can give the setup below the good college try and have them come in and fix what you broke ;).

Before you get going you must make sure that all of your SDRs are on Drift AND have connected their Google Calendar.

The tiny green calendar icons mean good. If it’s missing, SDRs won’t be able to get calendar invites.

Then, setup your routing rules.

You can do chat assignments by at least a billion criteria. It’s best not to spend too much time here, just get something that makes sense for your sales team and move on.

SFDC can do lead routing, too.

Examples of the type of routing you can do.

Ideally you’ll also setup your integrations.

I personally didn’t muck with a Salesforce integration, but you should if you have the resources to do it right and your admin isn’t swamped.

Then, SFDC can do assignment rules, analytics, and a bunch of other cool stuff. But it’s not a walk in the park, even with a dedicated SFDC admin.

And other integrations like Outreach are one-click and done.

One thing you definitely SHOULD do is turn OFF Drift here:

Turn off the Drift widget. We’ll do Playbook targeting instead.

If you’re doing a Marketo integration — link to the setting page — (or Segment or whatever). You MUST make sure you can do attribution. B2B Attribution models are all busted 12 ways from Sunday.

So, if SFDC isn’t connected, I’d map UTM tags.

I added custom UTM tags and mapped them to Marketo so that I could track which leads came from Drift. A SFDC integration is a BETTER way to do this.

Now that you’ve setup Drift, you’re ready to ACCOMPLISH SOMETHING!

2. Do The 101 Playbooks

These are instant wins that require almost no work.

101 Play 1: Target Your Contact Us Pages (8% conversion rate!)

This is super simple. Forms suck. And leads are PROBABLY getting lost in Salesforce. And they are slow. People want to talk now, let them.

Messaging for /contact
Conversion Rate for /contact

101 Play 2: Greet Your Target Accounts (0.9% Conversion Rate)

This is as simple as uploading a list of domain names and, with Clearbit Reveal built into Drift, there’s no targeting needed beyond that.

Messaging for ABM
Results for ABM

This puts a real person on the chatbot and allows key accounts to start a conversation.

That being said, if your LTV is super high and you only have a few hundred target accounts… I’d ditch this.

Here’s why…

Instead, I’d do 1 playbook per company where I provide super relevant information to each company.

It’s time consuming, but if your deals are $100k/year and you spend 1 hour per company on 100 companies and 1 converts, your time just paid off at $1k/hour :)

1–1 Playbooks are SO MUCH better for conversion. Photo by LinkedIn Sales Navigator on Unsplash.

101 Play 3 (Optional): / Outbound Email Integration (0% Conversion Rate)

This is a no brainer, when an SDR sends someone an outbound email, they should be greeted by that person if they click the link.

This one is as simple as connecting Outreach and writing a message.

Here is mine. I decided not to personalize it based on the email campaign, which the numbers show was a bad idea.

Messaging for Outreach Integration
Depressingly Crappy Results for Outreach Integration

Ok so I included this as a 101 play even though NO ONE chatted, but I think this represents the quality of the outbound process combined with my poor ass copy above.

And this one is SO easy to implement, I still would do it to see if your results vary. This experiment takes a few clicks.

If I was doing this over, I’d probably do this with UTM targeting and personalize the email accordingly based on campaign (more on that later).

Now that we have the low hanging fruit, we’re going to do a bot for everyone else.

This bot won’t be amazing, but it’s a start.

Then we’ll fill in what’s between with 1–1 targeting.

3. Build a Catch-All Qualifying Bot

Let’s recap. I want to make sure you know the best order of operations.

  1. First build your 101 Playbooks that give you personalization without much targeting work. That’s ABM,, and the Contact Us page.
  2. Then build your generic bot for (almost) all traffic.
  3. Then fill in the difference and build out 1–1 personalized bots (that’s after this section). This is the MAGIC. Stay tuned.

When building your catch-all qualifying bot you should do it this way…

  1. First capture interest
  2. Then validate their need
  3. Then provide proof
  4. Then qualify the lead

(your SDRs will say, I WANT CHAT WITH EVERYONE, don’t listen to them when they say they can sell sand to someone in the desert… QUALIFY the lead, it improves conversion if they trust your bot not to feed them to sales.)

Here’s what it looks like…

Messaging to get people engaged. Give the lead options!

The Hiya isn’t seen, only the first question (until you click into the bot)… so it’s kinda useless for conversion optimization. It’s a throw-away thing.

Here is what conversions look like for this play…

Results for All Visitors Messaging

Let’s dissect this bot! We’ll use this same logic and flow when writing content for all the other 1–1 bots. So, the end of this post will send you back here!

1. Capture interest

10/10 Firefighters suggest you don’t do this. Photo by Riley McCullough on Unsplash

This is where you should say something compelling. WHABAM!

I cheated and used Drift’s blog post to define my opening line.

Capture interest… this question should be a false-choice question where there are only two ways to Yes. I have a “chat with a human” here… but it’s okay if you just have Yes and Absolutely as options.

Give people something.

Just like good sales, you should give people something for clicking.

And, this catch-all bot play ONLY runs if none of the specific bot plays we will build out later apply to the visitor.

More on that later.

2. Validate their need

If your product has a ton of use cases, you should group them and ask the lead what problem they have.

Otherwise you’ll tell people about your fantastic toothpaste when they are looking to buy floss.

Don’t sell toothpaste when your leads want TO FLOSS!
Make sure their problem matches with the proof you’ll provide later.

I x’ed out that area on the right above because you SHOULD NOT allow free form responses… people panic and drop like flies.

Here’s how to kill that option.

Don’t let people type unless a human is around. It kills conversion rates.

3. Provide proof related to their problem

Proof DIRECTLY related to their ask.

Simple. Talk about how you’ve “crushed it” (as an SDR might say) for other customers with their problem.

This response should be dictated by their use case answer in #2.

4. (Dis)qualify the lead.
I’m going to get SDR hate mail for this…

Don’t waste the lead’s time if they aren’t right. The SDR team may think, “I can convert anyone.” But this is a trust building exercise between the bot and the lead. It’s about the customer, not your sales team.

Sorry, but not everyone SHOULD by your product.

Leads love to know they are a GOOD fit, so ask them.

I say a nice goodbye (above) and auto-close the chat if they aren’t a right fit.

And, if they aren’t right, close the bot and DISQUALIFY the lead. And name the goal so your metrics are slick.

If they are a right fit, I send them a note and then I ask if they want the same results as a reference customer I told them about for their use case…

Schedule a meeting or email you later?

This bot is ONLY if Drift Status is set to OFFLINE (which means no one COULD chat with them).

It’s such a shitty experience when the bot says, “let me get you an SDR” and it’s like, *waits 3 minutes* “oh sorry, it’s 9pm local time and no one is here… how about your email now so I can get back to you later? Even though I just lied to you once I promise I will send a good email and not spam you with marketing.”

Yea, right. Do it right the first time.

Add a conditional branch to drop a calendar if your team is offline and if they are online, offer them live chat.

Drift has CONDITIONAL BRANCHING… which you should use here based on Team Status (e.g. UK team vs. US team) or Drift status (is online) so you can either route to a live human if they are online or drop a calendar and not set broken expectations.

And, that’s it. Super simple.


GOOLLLLLLDDDD. Ok, it’s glitter. Photo by Sharon McCutcheon on Unsplash


A quick recap.

We’ve now launched our easy to implement bots that are personalized like ABM, the Contact Us page, etc. and we’ve built out our bot for the rest of our traffic.

Now we get to the HIGH converting stuff.

Think 5–10% from visitor → chat.

This is where you could spend so much time, making sure that each lead gets a message just for them.

And, the more personalized, the better it performs.

This is why companies like MutinyHQ exist, because the more personalized the message, the MUCH better it is for your conversion rates.

When looking for links, Drift noticed I was spamming their site, so they have a return visitor bot. Too bad I’m afraid of commitment.

This section will be a little bit different than the others, because your data, your business, and your circumstances will dictate different plays.

But, hold on.

Don’t peace out.

I have more useful data to share.

I’m going to build out a framework for HOW you should determine what 1–1 playbooks to build.

4. Know Your Targeting Options

Before you dive into your OWN data, you should know what’s possible to spark great ideas.

Here is a summary of the targeting buckets (full list here).
This is the order you should likely use to structure your next steps.

  1. Firmographics (if the Clearbit recognizes the IP you can do, employee count, revenue, rank, technology used, domain, name of company, etc.
My results for firmographic targeting

2. URL targeting (like /contact or /webinar)

My results for targeting a URL (e.g. /webinar). This is NOT the contact us page, which converts much higher.

3. Behavior (returning visitor, time on page, referrer, UTM).

My results for behavior targeting. We added UTM tags to an outbound campaign so people could click-in and chat. This is what a good email -> website targeting looks like.

4. Demographics (US, IP address, zip code, etc).

I wasn’t with the client long enough to do geo-targeting (e.g. folks in SF should get a coffee message OR targeting for events by zip).

5. Drift (Drift online/offline, SDR-based, outbound session, etc).

Online bots convert so much better. I’d use consider using conditional branching instead of dedicated playbooks, but it’s easier to see data this way.

There are also Marketo Smart Lists too. Consider using those too if you use Marketo (ideas below).

5. Research and Prioritize Your Plays

Boy, you’re in luck.

I built a spreadsheet to help your prioritize your plays.

Details on how to fill out this sheet below.

I would start with building out plays in THIS order. Firmographics, URL, Marketo, Demographics, and then Behavior.

1. Start with your Firmographic data from Clearbit Reveal.
This intelligence is so key to helping you speak to your (previously) anonymous visitors. As you saw above, you can get insane conversion rates using this intelligence.

You should build a dashboard in Google Analytics so you can easily see data like this.

It’s not shown here, but you can also do tag reporting to see what technologies they use.

I built a competitor bot that linked folks to the job pages (haha!). You could also do a tech bot that says, “I noticed you’re looking to switch from {{tech}}?”

Clearbit Reveal Custom Dashboard in Google Analytics.

This is where you need context to make great decisions.

If is a 100x better prospect than other companies here, then build a play JUST for Walmart. A super researched play. “Hey Walmart I noticed you don’t sell toothpaste…” or whatever.

If you have 10,000 people visiting from companies of 51–200+ employees and 100 visiting from 1000+ employees, then the 1000+ leads must be worth 100x to justify building a play for larger companies first.

I bet you don’t even know what this is. Photo by Dawid Małecki on Unsplash

You don’t have to get crazy, just sort of eyeball this and figure out what’s the best combination of tons of traffic and size of possible opportunity.

Pick the top 5 and then agree to do a customized bot on them.

If you don’t have a seperate Clearbit Reveal plan, it’s going to be hard.

Just buy it.

2. Then, move on to URL targeting
The best places to look here are pages that have serious intent (e.g. a page about the strength of your flossing game or a blog about flossing materials) and those that have UTM tags (e.g. a paid campaign from Google about the size teeth that work best with your floss).

URL targeting is the second most useful type of targeting because it’s pretty easy to determine WHAT the user wants (e.g. if you’re on /webinar it’s easy to write copy for them… “want me to send you a virtual copy of the webinar if you can’t make it?”)

Do a Google Analytics report showing top pages and UTM referrers

Do the same thing here, pick your top 5 with the best combination of clear purchasing intent and traffic.

3. Then do Demographics, Behavior, and Marketo Smart lists
I sort of lumped these into the last category because they are really business specific.


Don’t worry, it’s reclaimed wood. Photo by Brett Zeck on Unsplash

Demographic targeting is useful if you’re running events in an area, if you want to drop local messages from your SDRs offering a free lunch for locals, or if location is a huge part of your business.

Demographic targeting is a bit weird because it’s hard to infer someone’s problem by their location (unless your insurance and there was some event).

It’s mostly used for assignment rules (Michigan leads gets Michigan SDR).

You’re mostly optimizing to get people to click because of where they are (or if you’ve purchased IP data from like G2 Crowd… then you have intent and can build cool plays here).


Dang, good luck dude, you trapped. Photo by John T on Unsplash

Behavior is actually more interesting to me than demographics. If you think about referrer, I much prefer to do URL targeting with UTMs so you know EXACTLY the problem the person is having vs. blanket referral from Google.

Device may be interesting, “Did you know we have an iPad app,” but, again, this is super business specific.

Your mileage may vary.

Marketo Smart Lists

The Marketo Bot (just kidding). Photo by Owen Beard on Unsplash

This is a good option for folks already in the funnel, for doing targeting based on past behavior, or for routing support conversations for existing customers (e.g. John Smith is your CSM, here is his contact info).

6. Write Your Plays

This is sort of a trick section. Gotcha.

It’s just the same as above. Follow the same pattern.

I tricked you. This content is mostly repeated from above. Photo by Tine Ivanič on Unsplash

When building your bot you should do it this way…

  1. First capture interest
  2. Then validate their need
  3. Then provide proof
  4. Then qualify the lead
  5. Route the conversation based on online/offline

The only thing I will say is that you may not need to do 2, 3, or 4 depending on the page content or context of the user. I’d still follow this framework, though and evaluate on a case by case basis.


Hopefully this was useful. I’m totally down to talk about this in person or virtually and I’m my first name at if you want to email.

I do B2B consulting, postcard marketing, and I serve as an advisor to Job Portraits and Bouncer.


You read this far? Really?! Amazing compliment, thanks! Photo by Julian Lozano on Unsplash

Helpful Links

^ — PDF doc that explains key chatbot plays. Sara (below) made it.

My favorite person at Drift (okay, that’s a lie… EVERYONE I have chatted with I like… but Sara has so much hands on experience and she’s bright and helpful).

A note on Clearbit data…

This intro is solid. It addresses me by company name and it has their benefit. Go

^ — I like this bot. Clearbit thinks my company name is Cursive FWIW (it’s wrong — Clearbit is probabilistic data, not deterministic data). Just remember that it won’t be right for everyone, but it’s okay to get a few wrong I think. The bump in conversation is worth the occasional misstep.

I advise startups. And run a postcard marketing company at This is how I strive to live,

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store