If your facility is stuck in the loop where the vendor blames the utility and the utility blames you, another round of phone calls will not get you out of it. What ends it is evidence, recorded inside your facility on your side of the meter, while the problem is actually happening.
Producing an answer nobody can wave off
Monitoring on your side of the meter
What it is. Temporary power quality monitors installed on your system, recording voltage, frequency, power factor, harmonic distortion, and current draw across real operating time, timestamped throughout.
The trade-off. It needs access to your electrical system and a window long enough to catch an actual event. What you get for that is a record of what the waveform did at the moment it happened, and which direction the disturbance came from.
An analyst with nothing to defend
What it is. An independent engineer reading the data, with no equipment in your building and no delivery contract at stake.
The trade-off. It costs money, where the vendor and the utility will both analyze it for nothing. The independence is what makes the finding credible to people who have spent six weeks passing this back and forth, which is the only reason it is worth paying for.
A willingness to accept an internal answer
What it is. Going into the measurement without a preferred outcome.
The trade-off. More often than people expect, the data shows the cause was inside the fence the whole time. That is an uncomfortable finding and a useful one, because a cause inside your facility is one you can do something about.
What changes once the source is settled
It turns a question of blame into a question of fact, so nobody has to be talked into anything. The data separates a utility-side problem from a facility-side one in a way neither party can wave off.
If the disturbance really did come from the utility, you have documentation to take to them. If a motor start inside your own facility sagged the voltage, the conversation with the utility is finished and you can go after the real condition instead of replacing more equipment that was never at fault.
Either way the loop is over, because the question that kept it going now has a measured answer instead of a stack of opinions.