Finding ERP Buyers Before They Go Looking

So there I was, having just spent $8.63 scraping LinkedIn to find people who were unhappy with their ERP.

It got me 3 leads.

Which, to be fair, beats paying $500 a lead. But I had no idea yet whether those 3 were worth anything at all.

Some background

I was building an outbound system for a company that sells Acumatica, a modern cloud ERP. ERP is the software a company runs its money and its operations on. Accounting, inventory, jobs, payroll.

Acumatica is one of the better ones. It also competes with a lot of other good ones.

And people buy it once every 10 or 20 years, because replacing your ERP is enormously disruptive. So at any given moment, almost nobody is looking.

That is the whole problem. You are not competing for attention in a busy market. You are trying to find the handful of people in a very large market who have started to care.

Resellers have never had to solve this before, because the software vendor found the buyers for them for about thirty years. That is breaking now, and nobody has worked out what replaces it. One competing reseller spent half a million dollars on outsourced cold calling in a single year, at under 1% effectiveness.

LinkedIn signals

You can often find really good people through what I would call LinkedIn signals. That just means they are engaging with stuff related to your industry. Liking it, commenting on it, writing it.

It works well in a lot of markets. So we tried it here. We pulled the people posting about ERP, and the people engaging with posts about ERP.

Here is what came back.

Look at the last row. Post authors should be the warmest people on the list. They sat down and wrote about the pain themselves. Instead they are the worst of the three, and it is not close.

A post about "Dynamics GP end of life" is almost never a GP customer asking for help. It is a reseller marketing to GP customers.

So the population talking about ERP in public is mostly the industry talking to itself. That is not a data quality problem I can filter my way out of. It is a property of the topic. People sell ERP every day. They buy it once a decade.

There were a few real buyers in there. I never went back and reached out to them properly, which in hindsight was silly, and I would like to know what would have happened.

So we went to Clay

Not enough people, and relatively expensive per person. Time to try something else.

The idea was to stop starting from conversations and start from companies instead. Find companies that fit, then work out who to talk to at each one.

Clay is good at exactly this. You describe the shape of a company and it goes and finds them.

The other filter is the one that mattered most. Procore is project management software for construction. If a company runs Procore, that is a real, checkable fact about them, and it tells you they are big enough and organized enough to care about their systems.

About 1,087 companies came back.

Then: what ERP is each one actually running?

This is the step that made everything else work, and it is where the credits went.

I ran every company through a technographic lookup to find out what ERP they were on. Roughly 9 credits each.

About half came back with nothing on file. Of the rest, roughly 18% were on construction-specific systems, 16% on something at end of life, 7% enterprise, 6% modern, 2% already on Acumatica.

I geeked out over that chart more than I should have.

Job postings turned out to be the cheap way to fill in the gaps. A job ad that says "strong hands-on experience with Great Plains" proves that company runs Great Plains, and it costs about half a cent to read. 12 search terms, $2.80, 577 postings, 170 of them with a real timing trigger.

The part I liked most was accidental. The client had described the movable segment as companies where "the owner dies or the son takes over," and said it like it was unfindable. The words retire and succession appear in 57 of those postings.

The thing he thought was unfindable was a keyword.

Four segments, because a different system means a different reason to leave

Somebody on a tiny system leaves for reasons that have nothing to do with why somebody on NetSuite leaves. So what I segment on is not industry or company size. It is how big their current system is.

Four branches, each looking for something different.

End of life systems. Look for the new CFO. These have published expiration dates, so the clock is already running. What I want is somebody new stepping in who now owns a dying system, because a new CFO or ops leader is the person who gets to say we are not doing this again.

I thought that one was too obvious and I was suspicious of it. I was right to be. The first version went 0 for 21, because it only looked at senior titles.

Industry specific systems. Look for them outgrowing the vertical. These tools are very good at one thing, so the signal is a company moving outside that one thing. New types of work. Buying a company that does something adjacent. A tool that fit perfectly across one line of business starts creating silos the moment there are two.

Modern systems, badly implemented. Look for customer complaints. If the software is current and it still is not working, that is an implementation problem rather than a software problem. So I look for late orders and quality problems.

One disqualifier lives inside this segment. If the complaints are about leadership, that tells me nothing about whether they need a new ERP. It tells me people do not like their boss, and new software will not touch that. So lateness is a signal, management is an anti signal, and the company comes off the list. It fired on 43 of 186 companies.

No system on file. Assume that is wrong. I checked the first 10 by hand and 6 of them ran a real, named system sitting in their own job postings. A database saying no data and a company saying no system are not the same claim.

What I got wrong about scoring

My first model gave the highest score to companies showing signs of migrating.

The client pushed back. If a job post shows they are already transitioning off their ERP, is that not too late? He was right, and it made me rethink what I was actually looking for.

For a while I described that 0.2% as the thing I was chasing. A company that has started a project and not picked anybody yet is the textbook perfect prospect. But it is 0.2%. You cannot build a motion on it, and I do not think it was ever really the target.

What I want is the 86%. People sitting on a dying system who are starting to notice, and have not jumped yet. Awareness forming, before it turns into a project. That is much harder, because at that stage there is nothing to see. Nobody posts about it. Nobody has opened a search.

The 14% turned out to be good news, just not the kind I expected. Those companies already picked somebody. That is not a lead, it is a reason to skip them, and a signal that tells you who not to call is still worth having.

I flipped that score from +4 to -8. The data was fine. I had the direction backwards.

So what happened

Lower than I usually see. Personalized outbound like this typically lands 30 to 50% acceptance and about 30% reply on top of that. I am well under it, and I am not going to dress that up.

Total scraping spend across the whole thing was about $21.60.

Plus about 9,000 Clay credits to work out what everyone was running, which is a lot. Some of the enrichment steps ran over 40 credits a row.

I should also stop calling this intent data. I know their system, I know it is dying, I know the company and who owns the system internally. I do not know they are shopping. That is a materially weaker claim and I should have drawn the distinction earlier.

What I am trying next

More signals, and better ones.

Company reviews are the obvious next place. Somebody writing that the system is getting old and needs replacing is much closer to awareness forming than a job ad is. It is a person saying it out loud instead of a company advertising a role.

First party data is the other direction. Anything I collect myself beats anything I can buy, because nobody else has it.

Then the non obvious ones, which is really just me admitting I have not thought of the right signal yet. The bet is that some public fact sorts these companies, and I have found two so far. There are probably more, and the interesting ones are the ones nobody is looking at.

Age is not the obstacle here. Genericness is. A 40 year old ERP market has job ads, review sites, published end of life dates and hiring patterns going back decades. The market that looks hardest to break into is the one leaving the most public evidence lying around.

That is the part I believe. The part I have not proven is that any of it converts.

If you have run outbound into a market where people only buy once a decade, I would really like to know what worked.